A2

Power transformers
and reactors

Condition assessment of power

transformers
Reference: 761

March 2019
Condition assessment of power transformers
 Condition assessment of
power transformers
WG A2.49
 
 
Members
 
P. COLE, Convenor AU T.L. MACARTHUR, Secretary AU A. SBRAVATI US
C. BEAUCHEMIN CA L. LIDEN SE B. SPARLING CA
C. BECKETT AU M.A. MARTINS PT T.V. SRIDHAR IN
B.N. DE BHOWMICK IN T. MCGRAIL US E. TENYENHUIS CA
W. FLEISCHMANN DE N. MOODLEY ZA U. THIEß DE
M. FOATA DE A. MUDRYK UA M. VERMEER NL
Y. FUSE JP B. NEMETH HU D. WALKER GB
P. JARMAN GB P. PICHER CA J. WATSON US
M. KRUEGER AT T. PREVOST US R. WILLOUGHBY AU
J. ROUX FR S. MOHAMMAD IN

Other Contributor

L. MAUGAIN FR

Copyright © 2019
“All rights to this Technical Brochure are retained by CIGRE. It is strictly prohibited to reproduce or provide this publication in any
form or by any means to any third party. Only CIGRE Collective Members companies are allowed to store their copy on their
internal intranet or other company network provided access is restricted to their own employees. No part of this publication may
be reproduced or utilized without permission from CIGRE”.

Disclaimer notice
“CIGRE gives no warranty or assurance about the contents of this publication, nor does it accept any responsibility, as to the
accuracy or exhaustiveness of the information. All implied warranties and conditions are excluded to the maximum extent permitted
by law”.

ISBN : 978-2-85873-463-4
Condition assessment of power transformers

ISBN : 978-2-85873-463-4
 Condition assessment of power transformers

Executive summary
An asset manager may be responsible for a large fleet of transformers, each of which may have a
large amount of information related to the condition of the asset. Consequently, determining which
transformers most need attention can be extremely challenging.
Transformer Assessment Indices (TAIs) can be generated by calculating a Transformer Assessment
score for each transformer in the fleet, then using the assigned scores to rank the transformers.
Assuming the scores are calculated using an appropriate and sensible method, and using timely and
accurate data, the asset manager can easily identify those transformers which most urgently need
attention or intervention.
The focus of Working Group A2.49 was to investigate information used to derive Transformer
Assessment Indices, its’ consolidation, and the uses to which the output can be utilized.
Many asset managers currently use a ‘Health Index’ for this purpose. However, in many cases the
index does not provide any indication of how quickly the worst transformers on the list need to be
actioned, nor does it provide any indication of the type of action needed i.e. replace, repair or
refurbish. Many indices also fail to provide any indication of the confidence that the asset manager
should have in the assessments in the index.
Furthermore, these indices have generally been designed to provide an indication of each
transformer’s ‘health.’ Many asset managers use their ‘health index’ to help determine which
transformers in their fleet to replace. However, some “unhealthy” transformers can be (relatively)
easily repaired and therefore do not need to be replaced. A ‘health’ index may therefore not be the
ideal tool to determine transformer replacement.
Chapters 1 to 7 introduces a process that can be used to assess a transformer and to develop
Transformer Assessment Indices to suit the needs of the user. It also introduces the concept of the
scoring matrix, which can be developed and used to ensure that scores are allocated to each
transformer failure mode or mechanism being assessed in a consist way and with regard to the time
scale required for action. A number of different methods of generating both numeric and non-numeric
transformer assessment scores are also explained. Annex B provides a number of examples
illustrating how to generate different types of Transformer Assessment Indices and how to use the
different scoring methods. This section of the technical brochure also discusses:
▪ Options for dealing with the uncertainty in some information that will no doubt result for any
user with a large fleet of transformers
▪ How to utilize on-line monitors for assessment and in Transformer Assessment Indices
▪ Issues that an asset manager should consider before acting to correct transformers in poor
condition, and
▪ Methods that can be used to provide an estimate of the probability of failure.
The information and methodology presented in Chapters 1 to 7 can also be applied to assess and
develop and assessment indices for other assets.
Chapters 8 to 12 and Annex A deals with the sub-components of a transformer and discusses:
▪ Failure modes and mechanisms, and
▪ Methods of diagnosing the failure modes and mechanisms for each sub-component.
This section is mostly existing knowledge. Annex A provides tables of diagnostic information that are
formatted to allow assessment using the methods described in the technical brochure. The information
is from IEC & IEEE guides, CIGRE publications, CIGRE experts, and other industry experts. Users
can use these values as a starting point when assessing a transformer but should consider if the
values are suitable for their fleet of transformers, operating conditions, maintenance practices, and the
time scales used in the scoring matrix.

5
Condition assessment of power transformers

6
 Condition assessment of power transformers

Contents
Executive summary ...............................................................................................................5

1. Introduction - Fundamentals of transformer condition assessment..........................11

1.1 The problem for transformer asset managers ........................................................................................... 11
1.2 Transformer deterioration and failure modes ............................................................................................ 11
1.3 Condition assessment .................................................................................................................................12
1.4 Fundamentals of transformer condition assessment ............................................................................... 12
1.5 Role of insulating fluid in transformer condition assessment .................................................................13
1.6 Transformer assessment indices ................................................................................................................ 14
1.7 Operational feedback ................................................................................................................................... 17

2. Transformer assessment indices...................................................................................19

2.1 Using transformer assessment indices for asset management decisions ............................................. 19
2.2 Types of transformer assessment indices .................................................................................................20
2.3 Steps to develop a TAI .................................................................................................................................22
2.4 Methods of calculating a transformer assessment score ......................................................................... 26
2.5 Limitations of a transformer assessment score and index ...................................................................... 30
2.6 Examples – How to develop a transformer assessment index .................................................................31

3. Refurbishment index........................................................................................................33
3.1 Application of the transformer assessment index to identify eligible candidates for refurbishment ... 33
3.2 Application of the tai to estimate the potential benefits from refurbishment ......................................... 35

4. Dealing with uncertainty information.............................................................................37

4.1 Dealing with uncertainty within available information .............................................................................. 37
4.2 Dealing with unavailable information ......................................................................................................... 41
4.3 Options for showing uncertainty................................................................................................................. 45
4.4 Examples ....................................................................................................................................................... 45
4.5 Additional examples ..................................................................................................................................... 47

5. The role of on-line monitoring systems.........................................................................51

5.1 Condition assessment using on-line monitors .......................................................................................... 51
5.2 Use of on-line monitors with a transformer assessment index ................................................................ 52
5.3 Alarms from on-line monitors ..................................................................................................................... 52

6. Criticality and other condition considerations.............................................................53

6.1 Transformer assessment and consequence of failure .............................................................................. 53
6.2 Assessing criticality & developing a criticality index ................................................................................ 54
6.3 Economics of transformer management .................................................................................................... 54

7
 Condition assessment of power transformers

7. Estimating a probaility of failure....................................................................................55

7.1 Statistical methods ....................................................................................................................................... 55
7.2 Estimation techniques .................................................................................................................................55
7.3 Commentary on operational and external influences ............................................................................... 58

8. Transformer active part...................................................................................................59

8.1 Solid insulation degradation assessment .................................................................................................. 59
8.2 Dielectric assessment .................................................................................................................................. 65
8.3 Mechanical assessment ............................................................................................................................... 69

9. Bushings and cable boxes..............................................................................................77

9.1 Transformer bushings.................................................................................................................................. 77
9.2 Transformer bushing test and diagnostics ................................................................................................ 81
9.3 Failure mode assessment ............................................................................................................................ 83
9.4 Transformer cable boxes ............................................................................................................................. 84
9.5 Failure mode assessment ............................................................................................................................ 86

10. On-load tap-changer (OLTC) / de-energized tap changer (DETC).............................87

10.1 Failure modes ............................................................................................................................................ 88
10.2 Tests ........................................................................................................................................................ 91
10.3 Failure mode assessment ......................................................................................................................... 92
10.4 Scoring matrix for tap changers............................................................................................................... 93

11. Cooling system, transformer tank and ancillary components..................................95

11.1 Cooling system .......................................................................................................................................... 95
11.2 Transformer tank ....................................................................................................................................... 99
11.3 Ancillary components ............................................................................................................................. 100

12.1 Introduction .............................................................................................................................................. 105

12.2 Degradation and contamination ............................................................................................................. 106
12.3 Dielectric liquid testing and dissolved gas analysis ............................................................................ 111
12.4 Failure mode assessment ....................................................................................................................... 113
12.5 Scoring matrix for insulating liquids ..................................................................................................... 114

ANNEX A Transformer condition assessment tables................................................... 115

ANNEX B How to develop a TAI...................................................................................... 135
ANNEX C Literature overview ......................................................................................... 145
ANNEX D Principal international standards for insulating liquids .............................. 151
ANNEX E References ...................................................................................................... 153

8
 Condition assessment of power transformers

Figures and Illustrations

Figure 1-1 Relationship of Failure Modes to Assessment Indices and Diagnostic tools ...................... 15
Figure 4-1 Exponential distribution ........................................................................................................ 40
Figure 4-2 Exponential distribution with time dependent shape............................................................ 40
Figure 4-3 Distribution of bushing test results ....................................................................................... 48
Figure 4-4 Unknown parameter distribution .......................................................................................... 49
Figure 4-5 Colour misqualification example .......................................................................................... 50
Figure 8-1 Electromagnetic force .......................................................................................................... 71
Figure 8-2 Forced buckling distortion of coil circle ................................................................................ 72
Figure 8-3 Free buckling distortion of coil circle .................................................................................... 72
Figure 8-4 Conductor tilting ................................................................................................................... 73
Figure 8-5 Beam failure of conductor span between winding spacers ................................................ 73
Figure 9-1 Core insulation of bushing ................................................................................................... 77
Figure 9-2 Bushing failure modes.......................................................................................................... 80
Figure 9-3 Overheating of the top terminal seen using an IR/sensitive camera. .................................. 82
Figure 9-4 Phase to Earth Fault in liquid filled 11kV Cable Box............................................................ 84
Figure 9-5 Degradation and Crystallisation of insulating liquid in 11kV and 33kV Cable Boxes .......... 84
Figure 9-6 33kV Dry Type Cable Box Failure due to PD Activity .......................................................... 85
Figure 10-1 OLTC and DETC failure root cause analysis [6] ................................................................ 88
Figure 10-2 Major transformer failure due to OLTC Flashover - Root cause analysis [56] .................. 89
Figure 10-3 Examples of silver sulphide formation on OLTC contacts [37] .......................................... 90
Figure 10-4 Probability of OLTC minor failures [62] .............................................................................. 91
Figure 11-1: Particle build-up in water cooler ........................................................................................ 96
Figure 11-2: IR Image showing blocked radiators [66] .......................................................................... 97
Figure 11-3: IR image showing hot spot [66]....................................................................................... 102

Tables
Table 1-1 Condition Assessment Techniques ....................................................................................... 16
Table 2-1 Example of a Scoring Matrix ................................................................................................. 25
Table 2-2 Methods of calculating a Transformer Assessment Score ................................................... 26
Table 2-3 Example of a weighting matrix .............................................................................................. 29
Table 2-4 Examples of Scoring Methods .............................................................................................. 29
Table 4-1 Data validation methodologies ............................................................................................. 38
Table 4-2 Different options for managing of unavailability of information in assessment index design 42
Table 4-3 Example table of data quality indication ................................................................................ 45
Table 4-4 Colour intensity classification including required actions ...................................................... 50
Table 7-1 Estimated Probability Calculation.......................................................................................... 57
Table 7-2 Estimated Probability using Reduced Relative Probabilities ................................................ 57
Table 8-1 - Temperature rise limits ....................................................................................................... 60
Table 8-2 Scoring matrix for mechanical assessment .......................................................................... 76
Table 9-1 Types of transformer bushings.............................................................................................. 78
Table 9-2 Bushing assessment methods .............................................................................................. 81
Table 10-1 Failure detection capabilities of different OLTC tests [5] .................................................... 92
Table B-1 Filtering to develop a Refurbishment Index ........................................................................ 137
Table B-2 Refurbishment Scores ........................................................................................................ 140
Table B-3 Calculation of replacement scores .................................................................................... 140
Table B-4 Scores with weighting factors ............................................................................................ 142
Table B-5 Filtering to develop a Refurbishment Index ....................................................................... 143
Table B-6 Refurbishment Scores ....................................................................................................... 143

9
Condition assessment of power transformers

10
 Condition assessment of power transformers

Introduction – Fundamentals of transformer

condition assessment
Power transformers are expensive and critical equipment in power systems and play a significant role
in the transmission and distribution of electricity. Although transformers are generally reliable pieces of
equipment, failures do occur, and there are many degradation mechanisms operating in components
and sub-systems that will ultimately limit the useful operating life.
It is important for transformer users and asset managers to be adequately equipped to assess the
condition of a fleet of transformers in service in their utility/organisation as a basis for making critical
decisions about operations, repair/rectification of minor failures, refurbishment or replacement. Users
and asset managers need to understand all the failure modes of transformers to pinpoint the part of
the transformer affected, and to identify appropriate responses. Broadly there may be failures in active
parts of transformers or their accessories due to dielectric, mechanical or thermal breakdown. Some
sub-components also have their own unique failure mechanisms.
This brochure aims to set out the failure modes, the tests and diagnostic methods that can be used to
detect them, and methods of combining the available data into useful information in the form of
assessment indices that can form the basis for decision making and intervention prioritisation in
transformer asset management.

The problem for transformer asset managers

Whilst individual network and business requirements vary, transformer asset managers are generally
trying to achieve the required levels of safety and reliability from their fleet of transformers at minimum
cost. Knowledge of condition is therefore essential for efficient transformer asset management
decisions. Without this information only the most basic activities are possible; such as time-based
maintenance, replacement before end of life, or repair after failure.
For asset managers, determining the minimum required budgets for maintenance and replacement
and determining the most effective and targeted way of spending, is an important task. Often this must
be justified to stakeholders and regulators in an increasingly competitive environment.
Good knowledge of plant condition is an important step to good asset management. The relative costs
of replacement, failure and condition assessment will also need to be assessed to determine the most
economic approach. In general, for large transformers, the potential consequences of failure and the
high cost of replacement, make a sophisticated condition and risk-based approach very worthwhile.

Transformer deterioration and failure modes

To identify transformer condition, it is first necessary to identify the failure modes and failure
mechanisms; their consequences; and the interventions that can be implemented to change them.
Much work has been done in this area by CIGRE and others, and individual knowledge of the
transformer designs and operating conditions is necessary to identify weaknesses. In this technical
brochure, the failure mechanisms have been broadly categorised for the active part of the transformer
into thermal, mechanical and dielectric. On top of this, all of the other components of the transformer
necessary for satisfactory operation need to be considered; in particular bushings (and other
terminals), tap-changers, cooling systems, and the tank. The transformer dielectric fluid, (normally
mineral oil) has a special significance because it is not only critical to transformer operation and has
deterioration mechanisms, but it influences the rate of deterioration of other parts most notably the
solid insulation, and it provides a vector for much of the condition information.
Although it is not always the case, when assessing a transformer, it is useful to think of each failure
mode as independent and operating in parallel. Any one of the failure modes could, by definition, lead
to failure irrespective of the operation of the other failure modes. Additionally, interventions can be
thought of as affecting some failure modes and not others. For example, a deteriorated bushing could
cause a failure in an otherwise new transformer, changing the bushing affects the bushing failure
mode, but not the other modes. This concept is very important when developing transformer
assessment indexes.
Although transformers come in a very wide variety of types and sizes, it is possible to provide a
common set of tools and definitions to define and characterise the deterioration and failure
mechanisms that are important.

11
 Condition assessment of power transformers

Condition assessment
The essence of condition assessment is to identify the indications that can be used to determine (and
quantify where possible) the extent of the degradation of the components and sub-components of the
transformer. Much work has been done looking at individual mechanisms of degradation and failure,
together with methods for detection and quantification. This brochure aims to bring this work together
to provide a framework for overall assessment of transformer condition, which can then be used for
asset management. The appropriate condition assessment regime will very much depend on the size
and importance of the transformer, the availability of indicators, and the operating and business
environment. Where information for detection of degradation is not available, it may be appropriate to
use indirect indicators such as loading and age, and family related issues to assist with the
transformer assessment.
Although the assessed condition of a transformer will be critical in asset management decisions, other
factors are also generally considered. Refer to 6 Criticality and other condition considerations for an
overview of some of the other issues that should be considered before replacing a transformer or
proceeding with any other major interventions.
Users must also promptly respond to correctly set alarms and act upon bad, or suspected erroneous
test results or extreme observations. Ignoring a bad result may lead to catastrophic failure soon. The
user need not wait for full condition assessment to act if smoke is observed from transformers!
Fleet screening
A detailed assessment of a power transformer involves extensive testing (electrical as well as the
dielectric fluid), intrusive inspection and collection/compilation of other relevant data (operational as
well as maintenance history including commissioning data). This data is then analysed by an expert
based on relevant standards and expert knowledge. Before embarking on detailed assessment which
is time consuming in terms of outage & availability of transformers in the grid, users may consider a
screening assessment of the entire fleet to identify the problematic transformers for further testing and
assessment prior to considering major asset management decisions like repair, refurbishment or
replacement. Initial fleet screening may be based on a preliminary inspection, limited tests and other
factors like loading, age, known history of manufacturing defects etc. The fleet screening may be
automated or semi-automated to simplify the process.
It should be noted that an assessment based on limited available information is not complete, but it
may give insight to developing critical problems and is better than NO assessment.

Fundamentals of Transformer Condition Assessment

Many of the common condition assessment techniques used for transformers assess only the
transformer active part. For example, Dissolved Gas Analysis (DGA) from a dielectric fluid sample
taken from the main tank monitors the condition of the active part. However, all components and sub-
components of a transformer should be considered for a complete assessment. It should be noted that
unexpected failures of bushings and on load tap changers (OLTC) can frequently lead to catastrophic
failure of the transformer. Asset Managers should therefore consider on-line monitors or off-line testing
and inspection for the critical items. Design information, family history, maintenance records and visual
inspections can also be used to assess the components and sub-components of a transformer. Refer
to Chapters 8-12 for details of the sub-components of a transformer, their failure modes/mechanism
and how to assess them.
Assessment of Transformer Active part (magnetic core & windings) can be carried out in three ways -
thermal assessment, dielectric assessment and mechanical assessment. Condition assessment
techniques like DGA through testing at laboratory or on-line DGA monitors are used to analyse the
symptoms of an incipient fault being developing in transformers based on IEEE/IEC standards.
Reference may also be made to ISO 18095 - Condition monitoring and diagnostics of power
transformers [1].
Thermal failure of transformers basically depends on faster than anticipated ageing due to excessive
temperatures resulting from winding hot spot, metal hot spot and hot metal fault insulated with paper.
Dielectric Failure of transformers are a function of the degradation process of solid/liquid insulation
due to high thermal or electrical stress, moisture in paper/dielectric fluid, manufacturing defects, and
(semi) conductive particle (solid particles) contamination.
Mechanical Failure of transformers depends on the disintegration of the structure in core, windings
and solid insulation due to one of many reasons like strong electromagnetic forces from short circuit

12
 Condition assessment of power transformers

current flow through windings; mechanical deformation of windings due to shocks during
transportation, erection or during system faults etc; and degradation of solid insulating material.

Role of Insulating Fluid in Transformer Condition assessment

The insulating fluid plays a key role as an insulating medium that acts as a barrier between windings
and the tank and acts as a cooling medium that dissipates the heat generated within the windings and
core. It also functions as a preserving media that protects the windings and solid insulation from the
moisture and other physical contaminations.
The insulating fluid is often derived from hydrocarbon based “petroleum”. Hence it has a rich
hydrocarbon molecular structure. This molecular structure is subjected to electrical, mechanical and
thermal stresses during transformer operation. Hence the fluid decomposes and as a result generates
hydrocarbon-based gasses. These gases dissolve into the dielectric fluid, and therefore can be
measured. Some dissolved gases can escape if there is an escape path to the atmosphere via the
conservator without a sealing mechanism. Thus, the dissolved gasses may reveal a fault condition in
the transformer, if analysed properly by using techniques like DGA etc.
Materials like vegetable dielectric fluid, synthetic esters and synthetic materials like silicone fluid and
other materials, are also being used as insulating fluid. These materials also decompose when
subjected to electrical, mechanical and thermal stresses and generates certain type of gasses
depending upon their molecular structure.
Dissolved Gas Analysis
Dissolved Gas Analysis (DGA) is a powerful diagnostic technique used to analyse dissolved gasses
that are generated during the dielectric fluid and solid insulation decomposition process. Dissolved gas
analysis (DGA) has been an industry standard for the detection and determination of faults in
transformers for over 30 years. Developed in the late 1960s, DGA has been recognized worldwide as
the main tool to prevent catastrophic failures of power transformers.
Insulating liquids used in transformers, high-voltage bushings and other electrical equipment require
proper testing and maintenance to keep them in good condition. Liquid insulation that is not in good
condition may cause damage to equipment and become a hazard to personnel. Deterioration of
insulating liquid results from factors like oxygen, moisture, solid particle contamination, high thermal
stresses etc.
Effect of Oxygen on insulating fluids
Moisture contamination is one of the most common causes of deterioration in the insulating quality of
fluid. A less rapid but more serious deterioration due to the formation of acids and sludge, is caused
by oxidation. In free-breathing transformers, the oxygen supply is virtually unlimited and oxidative
deterioration is consequently faster than in sealed transformers. Leaky gaskets constitute a very real
hazard due to both oxidation and moisture contamination. The rate of oxidation also depends on the
temperature of the fluid; the higher the temperature, the faster the oxidative breakdown.
Moisture in transformers and insulating fluids
Water can be present in dielectric fluid in three states:
 In a dissolved form,
 ‘Bound’ water, where water molecules attach themselves to fibrous particles (including dust and
dirt), as well as acids formed during the oxidation process of the fluid. Once it is ‘bound’ to another
molecule aside from the fluid itself, it never moves, until filtration takes place; and
 As free water normally found at the bottom of the container holding the insulating fluid.
Demulsification occurs when the moisture dissolved in the fluid reaches the point of saturation. At this
stage the phenomena known as “raining” inside the tank occurs. The result is water sinks to the
bottom and forms a pool of free water. Emulsified water or water in the free state may be readily
removed by filtering or centrifugal treatment. The filtration process can partially remove dissolved
water if the filter papers are thoroughly dried before filtration and are replaced frequently.
The amount of moisture which can be dissolved in insulating fluid increases as the insulating fluid
temperature increases. Therefore, an insulating fluid treated at too high a temperature within a
transformer, may lose a large percentage of its dielectric strength on cooling because some of the
dissolved moisture (extracted from the windings at high temperature) forms an emulsion.

13
 Condition assessment of power transformers

Insulating fluid Deterioration in Transformers

In transformers, sludge can adhere to the surfaces through which heat is normally dissipated; the
sludge forms a resistance to the flow of heat from the insulating fluid to the cooling system and from
the core and coils to the (cooler) Insulating fluid. If allowed to continue long enough, the sludge may
even block the flow of insulating fluid through the cooling ducts. As a result, the transformer insulation
gets too hot and is damaged, particularly between turns of the windings. Deterioration of the insulation
may eventually lead to short circuits between turns and the breakdown of the transformer. When
oxidation progresses to the point where sludge is being precipitated, the first step should be to attempt
to remove the sludge from the transformer.

Transformer assessment indices

An asset manager may be responsible for a large fleet of transformers, each of which may have a
large amount of information. Consequently, determining which transformers need attention can be
extremely challenging. Many asset managers assign a score to each transformer based on some, or
all of the condition assessment information that they have.
Transformer Assessment Indices (TAIs) are generated by calculating a Transformer Assessment
score for each transformer in the fleet, then using the assigned scores to rank the transformers.
Assuming the scores are calculated using an appropriate and sensible method, and using timely and
accurate data, the asset manager can easily identify those transformers which most urgently need
attention or intervention.
What is the index for?
It is important to be clear about the question that the index or score is going to be used to answer. The
question, ‘which transformers on my system are more likely to fail?’ is very different to the question
‘which transformers should be considered for replacement?1’ The reason for the difference is that the
first question would generally take account of all the possible failure mechanisms whereas the second
question could exclude mechanisms that can be corrected by an intervention other than transformer
replacement.
The failure modes/mechanisms used to determine the transformer assessment score can be chosen
with an intervention in mind. For example, an assessment score used to determine if a transformer
should be replaced might not include the condition of bushings if they can be replaced separately. On
the other hand, a score used to determine the safety of personnel working in close proximity to the
transformer would certainly include the bushings.
Throughout the document a distinction is drawn between failure modes/mechanisms which are
underlying conditions that can lead to failure, and diagnostic results indications and other
measurements that are used to assess the underlying conditions that are not likely to impede
operating functionality. The following figure shows how diagnostic results are used to indicate the
progress of a particular failure mechanisms, and how scores for each failure mechanism can be
combined into different indices. The diagram is an example only, to highlight the possible relationships
between diagnostics, failure modes/mechanisms and indices. When designing an index, users should
determine the failure modes/mechanisms relevant to their index, and the information available for
assessment. Refer to Chapter 2. Transformer assessment indices for further information.

1
A TAI generally ranks transformers based on their assessed condition only. Refer to 6 for details of
other issues that should be considered before replacing a transformer.

14
 Condition assessment of power transformers

Figure 1-1 Relationship of Failure Modes to Assessment Indices and Diagnostic tools

Many TAIs are designed to consider and prioritise transformer replacement, i.e. a Replacement Index.
However, the same general approach can also be used to identify and prioritise other interventions.
Each TAI would use a different subset of the available condition information. Refer to section 2.2
Types of Transformer Assessment Indices for more information.
Developing a transformer assessment score
Once a user has determined the purpose of the TAI and the failure modes/mechanism to be included
in the index, test data and other available information can be used to assess the included failure mode
or mechanisms. It is important to distinguish between failure mechanism that can cause failure e.g.
partial discharge, and the indicators used to detect and assess them e.g. the level of a dissolved gas
in the dielectric fluid. Although many transformer assessment schemes go directly from the indication
to the score, it is hard to correctly interpret and objectively weight the indications without considering
the failure mechanisms they indicate.
It is essential that assessments are consistent, so users are encouraged to develop a scoring matrix
so that scores can be allocated to each failure mechanism of each transformer consistently. Refer to
section 2.3 Steps to develop a TAI for more information.
Once the indicators are attached to a failure mechanism and used to determine the likelihood of failure
in chosen timeframe, it becomes easier to combine the progress of different failure mechanisms into
an overall transformer assessment score.
A TAI or score is built up from the assessed condition of the failure mechanisms relevant to the index.
The condition assessment scores need to be combined in a way that, allows a transformer with a
single failure mechanism close to failure to be highlighted, but also for a transformer with multiple
small problems to be highlighted above a transformer with fewer small problems. This is not a
straightforward process because of differing levels of information and perceived importance for
different indications, and many schemes are used or have been proposed.

15
 Condition assessment of power transformers

Probably the most rigorous method of combination is to consider the probability of failure from each
failure mechanism and then multiply the survival rates from each failure mechanism together to arrive
at the overall survival rate. This is analogous to considering the failure of a chain knowing the failure
rate of each of the links. Unfortunately, the state of knowledge of the failure mechanisms and the
quality of the indications is not sufficient to apply this method with certainty, but the general approach
is effective in fulfilling the combination criteria above.
Weighted average schemes have been used to combine scores, but the choice of weightings is
difficult and arbitrary weightings can introduce significant bias. By the nature of the averaging process
the more failure mechanisms that are included, the harder it is to allow an individual bad component to
have the prominence it deserves overall. Refer to section 2.4 for further details of methods that can be
used to generate a transformer assessment score.
The assessment score will only be as good as the data and information that is used to assess the
transformer. In many cases, the data available for scoring will not be current. It is essential that the
effect on the score of this missing or low-quality data is understood. In some cases, it will make a
trivial difference to the overall score. In other cases, it could make a bad transformer look good;
resulting in the transformer being overlooked for critical repair or replacement. Similarly, the missing or
low-quality data can make a good transformer look bad. Implementing a “quality score” or some other
indication of the quality of the data used (typically called “significance”) may help to understand the
quality of data and the reliability or confidence level of a score. Refer to 4Dealing with uncertainty in
information.
Users should understand that many numeric scoring systems provide a score which is useful only for
ranking transformers. The absolute value of the score may not clearly indicate the overall condition of
the transformer or the urgency to address problems with the transformer. However, once a TAI is used
to highlight the worst transformers in the fleet, the assessment used to derive the assessment score
can be reviewed to determine appropriate actions and the time to act.
Assessment of Failure Mode/Mechanisms
Ideally, all failure modes/mechanisms would be assessed with the most precise and advanced
condition assessment techniques. However, in many cases this is not technically or financially
possible. Therefore, a selection needs to be made as to how a failure mode/mechanism is assessed.
There are multiple ways of assessment and each type of assessment has a corresponding level of
accuracy and uncertainty. In principle the following assessment techniques are available, in order of
increasing levels of uncertainty:
Table 1-1 Condition Assessment Techniques

A Advanced condition assessment Advanced measurements are used to gain information on

techniques the actual condition of a part of a transformer. Generally,
more advanced techniques provide more accurate
information than basic measurements. Chapters in this
brochure have identified a number of these techniques to
be used for each part of the transformer and each failure
mode.
B Basic condition assessment Basic measurements are used to gain information on the
techniques actual condition of a part of a transformer. It is expected
that basic techniques provide a reasonable indication.
Chapters in this brochure have identified a number of
these techniques to be used for each part of the
transformer and each failure mode.
C Inspections Visual inspections are used to gain information on the
actual condition of a part of the transformer. The
detection is limited to degradation that is visible, audible,
can be smelt or felt. The inspections are subjective to
human interpretations.

16
 Condition assessment of power transformers

D Logical reasoning Whenever a measurement or inspection is not available,

logical reasoning can be utilised, using indirect
information, to gain information about the most likely
condition of a part of the transformer, e.g.
• Using relevant experiences of brand / types /
materials used
• Using environmental information and knowledge of
physical processes
• Using statistical inference of known parameters in
other comparable transformers

In the following chapters, multiple different assessment techniques have been identified for different
failure modes/mechanisms. Preferably, and whenever possible, one or more advanced or basic
condition assessment techniques should be included. Whenever multiple condition assessment
techniques will be used in a failure mode assessment the confidence level of the assessment may
increase if all techniques come to comparable results. Similarly, trend analysis techniques using
multiple measurements through the transformer’s history may provide increased confidence in the
failure mode assessment.
Whenever multiple condition assessment techniques have been used to assess a given failure mode,
the most accurate data technique should be selected for use in the transformer assessment. A more
accurate indicator should overrule less accurate indicators, if indicating the same failure mode. No
double-counting should occur. When multiple indicators with the same accuracy have been for the
same failure mode, either an average or a worst-case approach can be used.
For example, a Partial Discharge problem can generally be detected from the DGA results, but other
types of Partial Discharge e.g. ultrasound measurement could also be used to identify the same
problem. The DGA result is typically used as the first indicator. Whenever partial discharging is
detected using DGA, a more accurate ultrasound measurement can be used to gain more accurate
information on partial discharging. In this case the ultrasound measurement should override the DGA
results for partial discharges.
Note however, that some condition assessment techniques may provide information on multiple failure
modes. Therefore, the same information can be used to assess multiple different failure modes.
Condition assessment techniques may not always be practical due to the cost or technical limitations.
Ideally information from logical reasoning should be included in the assessment of the failure
mode/mechanism. Examples of logical reasoning will be discussed in more detail in Chapter 4 Dealing
with uncertainty in information.
Users should also consider how often the assessment information needs to be updated. This will be
dependent upon the expected speed of degradation of failure modes/mechanism and the cost to
implement. Some failure modes show sudden very rapid degradation after prolonged periods of
normal operation e.g. degradation of bushings, whilst other failure modes show more gradual
degradation, e.g. paper degradation.

Operational feedback
It is crucial to the entire process that information on degradation and failure is gathered from
operational experience. The most valuable information can be obtained from post-mortem forensic
examination of transformers when they are scrapped. For most scrapped transformers, it is possible to
determine the extent of degradation by observation and testing of critical components, most
importantly the solid insulation. If they have failed in service, then the failure mechanism needs to be
understood. In any case, the information can be used to help to validate and calibrate the condition
assessment methods used and to determine whether there are any previously unrecognised failure
mechanisms in play. Please refer to CIGRE TB 735 [2] titled Transformer Post-Mortem Analysis.

17
Condition assessment of power transformers

18
 Condition assessment of power transformers

Transformer assessment indices

An asset manager may be responsible for a large fleet of transformers, each of which may have a
large amount of information. Consequently, determining which transformers need attention can be
very difficult. Many asset managers assign a score to each transformer based on some, or all of the
condition assessment information that they have. The transformers can then easily be ranked
according to their score.
This chapter discusses some of the different methods for developing a Transformer Assessment Index
(TAI) and details the advantages and disadvantages of the various methods discussed.
Guidance is also provided to allow the asset manager and the transformer specialists to determine
and develop the most suitable TAI for their fleet.

Using Transformer Assessment Indices for asset management

decisions
To manage a fleet of transformers an asset manager needs to determine which transformers would
most benefit from intervention or action, and then decide the most appropriate type of intervention or
action for each transformer.
For example, the asset manager may consider replacement of the transformer, refurbishment or
repair. During this process the cost of possible actions or interventions needs to be balanced
against the current performance and risks of the transformer, and the improvement(s) expected
from an intervention. A TAI is one method that may help identify the transformers which most
need intervention or action. A well-designed TAI which has been designed for a specific purpose
often suggests an appropriate action for consideration.
Furthermore, other factors need to be considered for each transformer before deciding on plans and
priorities for replacement or repair e.g. refer to Chapter 6 Criticality and other condition considerations
for further details of how these issues can be considered in conjunction with a TAI.
It is also essential that users of a TAI have some understanding of the level of confidence in each
assessment that has been undertaken prior to assigning a transformer assessment score to each
transformer. In any large fleet, some of the data and information required for assessment of some
transformers in the fleet will be missing, inaccurate or out of date, and the user must ensure that
assessments based on such low-quality data are highlighted. If the data used for assessment is
incomplete or incorrect, a transformer that needs urgent intervention may not be highlighted by the
TAI. Refer to Chapter 4 Dealing with uncertainty in information for a detailed discussion of this topic.
Users must also be aware that a TAI (particularly one that is poorly designed) may hide, or mask bad
results. For example, consider a transformer which is generally in good condition but has a bushing in
poor condition. In many TAIs, particularly those that produce a simple numeric output, the total score
could show that the transformer is in average or good condition. The bad bushing result has been
masked by the good scores assigned to other failure modes being considered.
Some users believe that the masking effect can be easily overcome by using weighting factors. If the
bushing score was heavily weighted, a bad score would result for a transformer with a bad bushing,
although other failure modes being considered have good scores. The problem is that the masking
effect is now exaggerated for other failure modes which have not been heavily weighted. Problems
(not related to the bushing) which may lead to failure and require urgent attention are further masked
by the weighted system. The urgency associated with a problem can be easily lost in a poorly
designed TAI.
Users need to carefully consider the effect of masking when choosing a scoring system and designing
their TAI. Furthermore, users must understand that a TAI is not intended to be used for “Alarm
Management.” When a bad test result is received, or there is any other indication that a transformer
needs urgent attention, the problem should be addressed regardless of the TAI score.
Fleet screening or detailed assessment
Some asset managers choose to use an index for Fleet Screening; a method where they use readily
available information to do a quick assessment of each transformer in their fleet and then rank the
transformers based on their scores. The assessment process may be automated or semi-automated
to simplify the process. The transformers that are identified through this process would then, most

19
 Condition assessment of power transformers

probably, be subjected to a furthermore detailed assessment before any major decisions were made
to confirm the outputs of the Fleet Screening.
Other asset managers may have a transformer specialist perform a very detailed and thorough
analysis of each transformer and then assign transformer assessment scores. Major decisions could
then be made based on the transformers’ score. If an appropriate and sensible scoring system is
used, a score derived through this more thorough method could help to support a justification to
management and/or a regulator to replace or refurbish a transformer.
Other methods combine both above-mentioned methods, where a basic screening is performed with
readily available information but may also use smart algorithms and additional data when it is
available. Although this may give a better result that a more basic fleet screening TAI, a full
assessment should generally be performed before major decisions are made.

Types of Transformer Assessment Indices

TAIs are generated by calculating a score for each transformer in the fleet, then using the assigned
scores to rank the transformers. If the scores are calculated using an appropriate method and using
timely and accurate data, the asset manager can easily identify those transformers which most
urgently need attention or intervention.
It must be noted that many numeric scoring systems result in a value that allows transformers to be
ranked; but the score itself is often arbitrary. In these situations, a user should not assume that a
transformer that has a score of 84 is twice as good (or bad) as a transformer with a score of 42. Nor
should they assume that the remaining life or the probability of failure of one transformer is twice that
of the other. In most cases, the number only provides a means of ranking the transformers.
However, some users will utilise statistical techniques to try to estimate the probability of failure or the
remaining life from the transformer’s score. In this case, the TAI may be scaled to represent the
parameter being estimated.
A TAI can be designed to identify the transformers most in need of the asset manager’s attention or
intervention. Such an index could be referred to as a Reliability Index, or a Health Index.2 When using
this type of TAI to identify transformers most in need of attention, the most appropriate type of
intervention should then be determined by further technical analysis, consideration of the effectiveness
and the costs associated with the interventions being considered, and detailed consideration of the
criticality and other issues. Refer to 6 Criticality and other condition considerations.
Alternatively, users can develop TAIs for specific purposes. Indices can be developed to determine the
transformers that could most benefit from repair, refurbishment, or replacement. The purpose of the
index would help identify a possible intervention for consideration. For example, refurbishment could
be considered for transformers highlighted by a “Refurbishment Index.” Note that criticality and other
issues still need to be considered as per Chapter 6 Criticality and other condition considerations.
Note that a well-designed index and scoring system will generally identify those transformers which
most urgently need attention. The timescale for action will be dependent upon the timescales assigned
in the scoring matrix used to assess the failure modes and failure mechanisms considered as part of
the index.
It is essential that users understand (and communicate to all other users) the purpose of the index,
and that the index is not used for any other purpose without careful consideration. For example, a
Reliability / Health Index3 is used to identify transformers that are most likely to fail. An asset manager
should not simply use their transformer replacement budget to replace the top 10 transformers on the
list. It is possible that some of the transformers on the list can be easily and economically repaired,
whilst other transformers further down the list need replacement.
Reliability / Health Index
This index is used to rank transformer based upon likelihood of failure or expected time to failure, and
those that have degraded from their original condition. This index ranks all transformers that require

2
Note however that Health Index is a commonly used term that has been used by many users to
describe indices, which may not directly relate to the health of the transformer.
3
Refer to Section 2.2.1, which explains why this type of Index does not directly relate to the reliability
of the transformer.

20
 Condition assessment of power transformers

any type of intervention, based on a sense of technical urgency. Users need to identify the most
appropriate intervention for transformers that are identified as needing attention.
Note that a Reliability or Health Index will not generally result in a score that directly relates to the
reliability or health of the transformer. If the index is well-designed, it will assist in ranking the
transformers so that those that are less healthy or less reliable are highlighted.
Replacement Index
This index is used to identify the transformers that most need to be replaced. The scores for each
transformer are based on:
 Irreversible failure modes and / or
 Single or multiple reversible4 failure modes which in combination, cannot be economically
repaired. This can only be achieved by some consideration of economic factors during the design
of the index.
Users will need to decide if they wish to include components such as bushings, tap changers and
insulating fluids in their Replacement index. These components can be repaired or replaced, although
some users may consider it more economical to replace the transformer, particularly if more than one
of these components is in bad condition.
Repair Index
This index is used to rank transformers that would most benefit from repair or non-essential
maintenance. The scoring of transformers is based upon reversible failure modes which can be
corrected by repair or by performing maintenance. Users may need to do a more detailed financial and
condition assessment on transformers highlighted by this type of index, before proceeding with repairs
to ensure that the transformer can be effectively and economically repaired. A Replacement Index
could also be used in conjunction with a Repair Index to ensure that transformers that have been
identified as needing repair do not also have irreversible failure modes or have not already been
identified as being uneconomical to repair.
Users could also consider using a Reliability index in conjunction with a Replacement Index to identify
transformers that need repair. Any transformer that is identified as being unreliable that does not need
replacement, should be considered for repair.
Refurbishment Index
This index is used to rank transformers based upon their need and / or technical feasibility for
refurbishment. This index focusses, typically on multiple reversible failure modes or failure modes
which involves excessive cost interventions and lifetime extension. It is possible that transformers
identified may not be economical to repair, and further analysis will lead to an economic decision to
replace these transformers rather than repair them. A well-designed refurbishment index will however
ensure that those transformers which have well developed irreversible failure modes are not identified
as being suitable for refurbishment. Refer to Chapter 3 Refurbishment index for information about how
to develop a Refurbishment Index.
Composite Index
This approach combines the outcomes of the replacement and refurbishment or repair requirements. It
includes a technical assessment and generic financial considerations to allow it to decide the best
solution, considering the maintainability of failure modes in the transformer. This approach combines
the different types of indices in one index, e.g. highlighting maintainable indices and non-maintainable
indices in one composite index. The index identifies the reversibility of a failure mode and suggests
replacement or maintenance. A composite index may also include generic financial considerations to
allow it to decide the best solution. These financial discussions are beyond the scope of this Technical
Brochure.

4
“Reversible failure modes” refers to failure modes that can be corrected or partially corrected by
repair or replacement of a component or sub-component. e.g. failure modes associated with a bushing
could be considered reversible as the bushing could be replaced. Failure modes which are corrected
by repair are often not fully reversed.

21
 Condition assessment of power transformers

Mitigation Index/Score
Whilst typical indices assess the state of an asset, the approach can be adapted to assess the
technical feasibility of a refurbishment or repair. When a repair or refurbishment score
𝑅𝑆𝑂𝑟𝑖𝑔𝑖𝑛𝑎𝑙 indicates that a transformer should be considered for refurbishment or repair, the user can
then calculate 𝑅𝑆𝑀𝑖𝑡𝑖𝑔𝑎𝑡𝑒𝑑, by using the same index and scoring system to calculate a score based on
the assumed condition of the transformer after the intervention has been successfully completed.
Some users may consider 𝑅𝑆𝑀𝑖𝑡𝑖𝑔𝑎𝑡𝑒𝑑 to help determine if the repair or refurbishment is justified i.e. the
proposed repair or refurbishment would not be considered unless 𝑅𝑆𝑀𝑖𝑡𝑖𝑔𝑎𝑡𝑒𝑑 falls below a
predetermined threshold.

A Mitigation score can also be calculated as follows: 5

𝑀𝑆 = 𝑅𝑆𝑂𝑟𝑖𝑔𝑖𝑛𝑎𝑙 − 𝑅𝑆𝑀𝑖𝑡𝑖𝑔𝑎𝑡𝑒𝑑

Ideally a higher value for MS indicates that the repair or refurbishment has made a more significant
difference to the condition of the transformer. However, the value of a mitigated score will be
somewhat dependent upon the scoring system used. Many numeric scoring systems result in a value
that allows transformers to be ranked; but the score itself arbitrary. Users must understand the scoring
system to understand the relevance and value of using mitigated scores.

Steps to develop a TAI

The basic steps to develop a TAI are listed below:
1. Determine the purpose of the Transformer Assessment Score and Index
2. Identify the failure modes to be included in the TAI
3. Determine how each failure mode will be assessed
4. Design a calibrated system for categorising failure modes (scoring matrix)
5. Calculate a TAI Score for each Transformer
It should be noted that some users, particularly those with smaller fleets, or those performing a
detailed assessment on a single transformer, may not want or need an index to allow transformers to
be ranked. In this case Steps 1-4 can be followed. This will result in the relevant failure modes of each
transformer being assessed and categorised in a consistent way.
Determine the purpose of the Transformer Assessment Score and Index
The design of a TAI is highly dependent on its purpose. Therefore, the first step in the development of
a TAI is to determine its purpose.
Users should decide which categories of transformers need to be identified and what business
decision need to be supported. For example, a user may want a TAI that identifies those transformers
that should be prioritised for replacement; and may wish to use the results from the TAI to support the
business case for replacement of these transformers. Users should also decide if the TAI is to be used
for Fleet Screening or Detailed Assessment. Refer to section 2.1.1
For examples of TAIs that may suit users’ needs, refer to section 2.2. Users may also choose to
develop multiple indices for different purposes e.g. a Replacement Index and a Refurbishment Index.
The purpose of the TAI should be carefully discussed and agreed with its users and asset managers.
The purpose of the index should generally be closely aligned with the decisions to be made by the
asset manager.
Identify the failure modes to be included in the TAI
When the purpose of the TAI is understood and agreed, the failure modes to be included in the index
can be determined. The failure modes to be included will generally be aligned to the relevant business
drivers. For example, a user may choose to include only information indicating the condition of non-
reversible failure modes of the main tank into a TAI developed to highlight the transformers that should
most urgently be replaced. This would be appropriate if the user generally chooses to repair or replace

5
This formula assumes that the scoring system assigns smaller numbers to transformers in better
condition i.e. the improved score after mitigation will result in a lower number.

22
 Condition assessment of power transformers

bushings, tap-changers or other components of the transformers in poor condition, rather than replace
the transformer.
Users must, however, very carefully consider the failure modes that are to be included in their TAI.
Users may choose not to include reversible failure modes in a Replacement Index. Although the user
would generally replace components or repair the reversible failure modes, it may not be economical
to do so if the transformer has multiple reversible failure modes, or if the transformer has a limited
remaining life. For example, a user may not find it economical to replace a tap-changer and HV
bushings, unless the transformer is otherwise in good condition.
There are two general options for dealing with this situation:
Design the Replacement Index such that reversible failure modes are considered, but only have a
significant effect on the assessment score when:
 There are multiple reversible failure modes that need to be addressed; or
 A reversible failure mode is obviously not economical to repair; or
 The overall condition of the transformer is poor (i.e. low residual value) so that addressing
reversible failure modes is not economical
This can be easily achieved if assessments are being performed manually by a transformer specialist.
However, this would be more difficult in an automated system. Weightings of scores would probably
be required, which can lead to other problems.
Do not include reversible failure modes into the replacement index; but otherwise identify
transformers that have active reversible failure modes that require intervention and determine if
these interventions can be economically justified. Transformers that require interventions that
cannot be economically justified should be added to the list of transformers that are prioritised for
replacement.
Although this example deals with problems associated with a Replacement Index, similar problems
may occur when developing other TAIs. In general, this can be resolved by developing a more
complex TAI e.g. a composite TAI or using one or more simple TAIs to develop a preliminary result
which may need further analysis.
Determine how each failure mode will be assessed
Ideally all failure modes included in a TAI would be assessed with the most accurate and advanced
condition assessment techniques. However, for technical and financial reasons, this is generally not
possible. Therefore, users need to determine how each failure mode is to be assessed.
Users should design the TAI to ensure that relevant failure modes are not assessed multiple times.
When multiple sources of data and information provide information about a failure mode, the user can
use the multiple sources to assess that failure mode, generally with a higher level of confidence in the
assessment. The multiple sources of data should generally not result in multiple assessment scores
for a single failure mode.
Assessment techniques each have a corresponding level of precision and uncertainty. Note however
that measurements of high accuracy or precision are not always necessary. For example, touching the
radiators of a heavily loaded transformer and finding that one radiator of the bank is much cooler than
the others, is a good indication of a cooling system problem. Although such a measurement is
imprecise, it gives a good indication of a problem that should be corrected. Such qualitative
information can be used on its own or used in conjunction with other information when determining a
score for some failure modes associated with the radiators, if the scores are to be used to generate an
index for Fleet Screening.
Chapters (8-12) identify the possible condition assessment techniques for each failure mode and have
included a calibrated scoring system. In many cases, different assessment techniques have been
identified for a failure mode.
Whenever multiple condition assessment techniques will be used in a failure mode assessment, the
confidence level of the assessment may increase if all techniques come to comparable results.
Similarly, trend analysis techniques i.e. using multiple measurements through the transformers history,
may provide increased confidence in the failure mode assessment.

23
 Condition assessment of power transformers

Whenever multiple condition assessment techniques are used to assess a failure mode, the TAI
design should select the most accurate data to be used to assess the failure mode. Additional
information can be used to gain more confidence in the failure mode assessment, but “double
counting” of the same failure mode should be prevented.
For example, a Partial Discharge (PD) problem may be detected from the DGA results. When PD is
detected from the DGA results, users may choose to perform a more accurate UHF measurement to
confirm the presence and the severity of the PD problem. In this case the UHF measurement would
probably be used to assess the failure mode. The user however has a higher level of confidence in PD
assessment, since PD problem is also shown in the DGA result. This will affect any quality score
assigned to the overall assessment.
As another example, the transformer loading guide (IEE C57.91 [3] or IEC 60076-7 [4]) can be used to
estimate the loss of life of paper due to its historical thermal load. Similarly, furan analysis can be used
to assess the paper loss of life. Both techniques still have elevated levels of uncertainty. Therefore,
both assessment techniques can be used in the failure mode assessment either using an averaging or
worst-case approach. Knowing that the furan result and the loading information both indicate loss of
paper life, the user has a higher level of confidence in the assessment.
The TAI therefore needs to be carefully designed. Generally, the following principles should be
followed:
 The more accurate indicator should be used as the primary indicator for the assessment of a
failure mode. “Double counting” should not be allowed.
 When multiple indicators with similar accuracy are available for a single failure mode, either an
average system or a worst-case approach can be used. The choice will be dependent on the
purpose of the TAI.
 Users need to also determine how to manage situations where multiple indicators for a failure
mode result in a very different assessment. The worst-case result could be used, but the
assessment would probably be accompanied by an indication that the user should have a low
confidence in the result; due to the uncertainty around the assessment. A poor assessment might
result in the transformer being highlighted to the asset manager, but the indicator of low
confidence needs to result in the asset manager investigating further before reacting.
Note: Condition assessment techniques may provide information on multiple failure modes e.g. DGA,
therefore the same information can be used to assess multiple different failure modes. In such cases
the user will need to carefully consider how to include the information in the assessment to prevent
double counting.
Design Information, family history, loading information, data from site inspections, and inductive
reasoning may be useful if more precise or accurate information is not available. Any assessment
based on this information would ideally be based on experience and would be subjective. Users
should ensure that these subjective assessments are consistent, so a guide or scoring matrix should
be developed to ensure consistency. Refer to 4 Dealing with uncertainty in information. This
information can also be useful to improve the confidence in an assessment where other information is
used for the primary assessment. Again, double counting should be avoided in the design of the TAI.
To achieve a reasonable level of confidence in the TAI, the availability of condition assessment
information for the transformer population should be considered. There may be little value in including
a failure mode into the TAI if reasonable indicators of the failure mode are not available for a
significant proportion of the transformer population. This will again be dependent upon the purpose of
the TAI and its design. For example, if a user has recently introduced a new test technique, and all
transformers in the fleet will be progressively tested over the next 5 years, the failure mode indicated
by the new test could be included in the TAI. Users would however need to determine how to assess
those transformers that have not yet been tested with the new technique.
Design a calibrated system for categorising failure modes (scoring matrix)
Each failure mode being considered should be assessed and the effect on the transformer, assuming
no remediation is carried out, should be determined. It is essential that the scoring of failure modes is
consistent, so a scoring matrix should be developed prior to allocating scores. By using the scoring
matrix, all the failure modes that are considered will have a similar score if they are likely to have a
similar effect on the transformer. The scoring matrix can have any number of levels, as long as each
level has a label and a definition. Levels will often relate to the urgency that problems should be

24
 Condition assessment of power transformers

addressed to prevent failures. However, given the lack of precision associated with the assessment of
most failure modes, a matrix with more than five levels would probably be of little value.
An example scoring matrix developed by the working group is used throughout this brochure. This
matrix effectively has five levels. (The 6th level - labelled F - is not used when generating a TAI but is
noted to consider very-short term failure criteria) The five levels, which relate to five different condition
levels, are each allocated a colour from Green to Red. Letters (A to E) or any other label can also be
used.
The user must also carefully consider and define the time scales that they wish to use in their
matrix. The purpose of the matrix is to ensure consistency when scoring failure modes/mechanisms.
It is therefore essential that the terms used in the example matrix, such as “as soon as possible,”
“short term,” “medium term” etc. are defined by the user. An engineer who determines that a failure
mode/mechanism needs to be addressed within 2 years, needs to understand if the failure mechanism
should be assessed as Pink or Red.
Note that if any failure mode is assessed as Level F, the Transformer should be removed from
service. A TAI is not intended to assist in short-term decision making.
The mitigated score (2.2.6 Mitigation Index/Score), however, could be calculated to assist in
determining if the transformer should be scrapped or repaired.
Table 2-1 Example of a Scoring Matrix

F De-energize as soon as possible. Don’t return to service until problem is repaired. Component is at end
of life.

E Very Poor condition – high likelihood of failure. Component is near end of life. Repair or replacement
as soon as possible is recommended. De-rating or restricted operation of the transformer may be
appropriate, and operation under extreme conditions may not be appropriate until replacement is
possible.
D Poor Condition. Repair or replacement should be considered within the short term. Reliable operation
may be impaired or compromised. Performance or component may be causing deleterious effects.
Consider review of rating and operating condition.
C Acceptable condition with significant signs of aging or deterioration. Reliable operation expected for
medium term – but consider condition-based maintenance if applicable.

B Good condition. Some signs of aging or deterioration are evident. Reliable operation expected for a
lengthy period.
A Minimal Signs of ageing or deterioration. As new condition.

User can use the example matrix developed by the Working Group (after defining the time
scales), or they may also develop their own matrix with any number of levels. It is however
essential that there is a clear definition of each level or category to ensure that scores for each failure
mode being assessed can be applied consistently.
The matrix above has been used throughout the document. However, it has been varied in
some sections so that the descriptions are more specifically aligned to the failure modes and
mechanisms associated with some of the sub components of the transformer.
Note that to calculate a total transformer assessment score it may be necessary to assign a numerical
score to each category. Refer to section 2.4
Calculate a TAI Score for each Transformer
Once all relevant failure modes have been assessed and assigned to a category on the scoring
matrix, a method is required to calculate an overall Transformer Assessment score. There are multiple
ways to generate an overall score. The method chosen will depend on the purpose of the TAI and the
individual user’s needs.

25
 Condition assessment of power transformers

When designing the scoring system, the following points should be considered:
1. The scoring system should allow all transformers in a fleet to be ranked, such that those which
are the highest priority for action or intervention are easily identified.
2. The scoring system should ensure that transformers with a failure mode that requires urgent
attention are easily identified. That is, a bad failure mode should not be masked.
3. The scoring system result should be easily interpretable by any user, with reference to the
purpose of the TAI.
4. The scoring system should be transparent and reproducible.
Designing a method for calculating transformer assessment scores such that the above conditions are
achieved can be challenging. If for instance an averaging scoring system is used where the overall
assessment score is the average or weighted average score of all failure mode scores, an urgent
failure mode may be hidden if the transformer is otherwise in good condition. A non-linear scoring
system (i.e. low scores for good and average conditions) will help minimise the effect of masking.
The following section provides an overview, with advantages and disadvantages, of scoring systems
discussed in the working group.

Methods of Calculating a Transformer Assessment Score

There are multiple ways to present the final result of a TAI. Many users expect their TAI to produce a
single number to allow transformers within the index to be easily ranked. Other users prefer the output
presented as simple colour to indicate the overall state of each transformer. The colour code enables
quick fleet assessment and dashboard functionalities. The output from a TAI may be both a number
and a colour; perhaps a number that represents the “average condition” of the transformer and a
colour to indicate the worst-case failure mode assessment.
A number of methods used for calculating a score are described Table 2-2.
Table 2-2 Methods of calculating a Transformer Assessment Score

# Description Advantages Disadvantages

1 Summation of individual failure mode scores: Simple algorithm Poor condition assessments
𝑁 Transparent may be masked so that score
𝑇𝐴𝐼 = ∑ 𝑆𝐹𝑀𝑖 Weightings can be added if may provide an optimistic
𝑖=1
required. indication of transformer’s
SFM is the score of an individual failure mode. condition.
A simple set of linear or non-linear scores Using non-linear scoring may
help to prevent masking
would generally be added to the scoring
Scores do not generally
matrix. Weighting can also be added as per reflect urgency.
method #2.
2 (Weighted) average: Transparent As above
∑𝑁𝑖=1 𝑊𝐹𝑀𝑖 ∙ 𝑆𝐹𝑀𝑖 Weighting allows some Weighting of failure modes,
𝑇𝐴𝐼 = failure modes to be may mask further mask
∑𝑁
𝑖=1 𝑊𝐹𝑀𝑖
SFM is the failure individual mode score highlighted. some failure modes.
WFM is the weighting per failure mode
N is the total number of failure modes
A simple set of linear scores or non-linear scores
would generally be added to the scoring matrix.
Weighting factors are applied to failure modes
that the user wants to highlight.

26
 Condition assessment of power transformers

# Description Advantages Disadvantages

3 Non-linear mathematical approach: Masking of worse scoring More complex scoring
𝑘−1 failure modes is prevented. system
𝑇𝐴𝐼 = ∑ 𝑥𝑛 𝑖 𝑛 The scoring results can be
𝑛=0 more difficult to interpret
i is the number base / radix, which is equal to or If weighting factors are also
greater than the number of failure modes used, a slight modification of
included in the TAI the formula would be
xn is the number of failure modes per category needed to prevent masking.
k is the number of categories included in the
failure mode assessment
n is the counter in this formula.
Refer to the example later in this section for
further explanation.
4 Numerical Score using estimated failure The TAI score can be scaled if It is generally only possible
probabilities required to estimate a very
A probability of failure, based on test results data Highlights single advanced approximate probability for
and other assessment information, can be failure modes and properly each failure mode
estimated for each of the failure mode. A score combines several less Method might tend to
can then be calculated advanced failure modes for indicate an unjustified level
TAI = 1 – ((1-est. PoFFM1) x (1-est. PoFFM2) x overall comparison of precision.
(1-est. PoFFM3) x……. (1-est. PoFFMn)) Works well provided each
failure mode probability, or
score is on the same scale
even if it is not a true
probability.
5 Worst case approach Simplest algorithm Weighting of failure modes
𝑇𝐴𝐼 = 𝑤𝑜𝑟𝑠𝑡(𝑆𝐹𝑀 ) Transparent only possible if the number
SFM is the score of an individual failure mode Worst case failure mode is of failure modes assessed
It is also possible to indicate the number of failure highlighted. with the worst-case score is
modes which have been assessed with the worst- included.
case score.
Score = Red 3 (transformer has 3 failure modes
that have been scored as Red).
5a Hybrid Score Combines two simple scoring A transformer with a single
The worst-case score can be used in conjunction methods advanced failure mode,
with one of the numerical scoring methods The simple numeric score cannot be distinguished from
described above. gives an indication of the a transformer with multiple
For example, a simple summated score can be overall condition of the advanced failure modes.
combined with a Worst-Case score. Scores would transformer, and the worst-
be of the form: case score highlights the
Tx 1 = 64 Red worst failure mode of the
Tx 2 = 64 Orange transformer.
It is clear that Tx 1 needs urgent attention,
although its numeric score is the same as Tx 2.
6 Count per category Very good visibility of the The TAI is not a single
The TAI is shown as a set of numbers, rather than total transformer health number but a set of
an individual scalar value. The number of failure assessment score numbers. This may make
modes assessed as being in each category is Masking of worse scoring representing the result on a
shown. failure modes is prevented dashboard, or in other
For example, using the five-colour matrix shown Weightings are possible if simplified forms more
in Table 2-1, a score for a transformer with 12 required, and will not mask difficult.
failure modes assessed might be as follows: problems.
3 5 3 1 0

27
 Condition assessment of power transformers

# Description Advantages Disadvantages

7 Machine learning: These techniques may find Complex algorithms are
The index does not use a predefined formula to new correlations between required
calculate the assessment score, but instead uses condition indicators and Large volumes of data are
modern data analytic techniques, where smart failure modes, potentially required (including failures)
algorithms e.g. neural networks, analyse the enhancing the index quality. Not transparent.
condition data and failure data. Results would need to be
validated to ensure that the
machine has ‘learnt’
correctly.

Use of weighting factors

Users often apply weighting factors to ensure that problems of greater consequence are highlighted
over lesser problems. For example, a dielectric fluid leak and a HV bushing may both urgently need
attention, but it is likely that the asset manager would want to see the bushing problem highlighted
above the dielectric fluid leak. This could be achieved by applying a lower weighting factor to the
dielectric fluid leak, and a higher weighting factor to the bushing.
The problem with weighting any failure mode assessment score is that it may further mask or hide
problems; although this will be dependent upon the scoring system chosen. For example, if a
transformer in generally good condition has a bad dielectric fluid leak, and a simple summation of
failure mode scores is used to calculate the transformer assessment score, it is probable that a good
score will result for the transformer; if the dielectric fluid leak has a low weighting factor. The urgency
to address the bad dielectric fluid leak is lost.
For an index that considers only the failure modes that are the critical drivers to the decision process
(e.g. an index that considers only the critical failure modes that will be used to determine replacement)
weighting factors may not be necessary.
Weighting factors may become relevant when an index includes failure modes that are not critical to
the decision process, but still need to be considered. For example, some users may choose not to
include bushings in a replacement index, as bushings, tap-changers and dielectric fluid leaks are not
generally critical drivers to transformer replacement, since these problems can be mitigated by the
repair or replacement of the component. However, it may not be economical to replace the bushings
and tap-changer and to repair dielectric fluid leaks. Therefore, these failure modes may be included in
the replacement index with a weighting factor applied, so that only their cumulative effect will push the
transformer further towards the top of the replacement index. Note that with weighting factors applied,
a transformer with a bushing that urgently needs replacing, may not be highlighted for replacement. It
is therefore essential, that this failure mode is identified using another index, or through direct
interpretation of test results, and addressed where appropriate.
If weighting factors are used, the user should take care when assigning values. Even with a non-linear
scoring system, a heavily weighted good score could hide a problem that requires urgent attentions if
that failure mode has a low weighting. If a user believes that some failure modes have a significantly
lower consequence than others, it may be appropriate to eliminate the less significant failure modes
from the index. Alternatively, a scoring system that does not hide failure modes should be considered.
If weightings are to be applied, they should be applied consistently. Some examples of aspects to be
considered when determining weighting factors include failure consequence, cost to repair, and time
required to restore the transformer to working condition. A matrix may be developed to assist in
determining the most appropriate weighting factors for each failure mode being considered. The
values in Table 2-3 are arbitrary but allow weightings to be applied consistently.

28
 Condition assessment of power transformers

Table 2-3 Example of a weighting matrix

Weighting Factor Criteria

1.5 This failure mode could result in a failure that will take more than 3 months to repair
or cost more than 50% of the replacement cost of the transformer.
1 This failure mode could result in a failure that will take between 1 week and 3
months to repair with an estimated cost of 10% - 50% of the replacement cost of the
transformer.
OR
This failure mode will lead to accelerated ageing of the transformer which result in
premature replacement of the transformer or a premature failure.
0.75 This failure mode could result in a failure that will take less than 1 week to repair or
cost more less than 10% of the replacement cost of the transformer.

Examples of Scoring Methods

In the following examples it is assumed that a simple index is based on 4 failure modes: FM1 - FM4.
Each failure mode has been assessed using a scoring matrix with the following non-linear scores
assigned for use in examples 1 – 3.
Note that the scores allocated to the five categories are only for generating a numeric assessment
score for the transformer and to allow ranking of the transformers. A non-linear scoring system has
been chosen to help ensure that transformers with one or more failure modes / mechanisms assessed
as “Red,” have a higher transformer assessment score. Users can assign their own linear or
non-linear scores to the matrix.

10 4 2 1 0
Red (E) Pink (D) Orange (C) Yellow (B) Green (A)
FM1 FM4 FM3 FM2

Table 2-4 Examples of Scoring Methods

Method Calculation Interpretation

1 Summation TAI = FM1 + FM2 + FM3 + FM4 The failure mode that is red should be
TAI = (1 × 10) + (1 × 4) + (1 × 2) addressed urgently, but this is not
+ (0 × 1) + (1 × 0) obvious from the score, since a
= 16 combination of less urgent failure modes
TAImax = 40 could give the same score.

2 Average (10 + 0 + 2 + 4) The failure mode that is red should be

TAI = =4
4 addressed urgently, but this is not
TAImax = 10 obvious from the score, since a
combination of less urgent failure modes
could give the same score.
3 Non-linear TAI = 1 ∗ 40 + 0 ∗ 41 + 1 ∗ 42 + 1 ∗ 43 This method ensures that bad results
(ignore + 1 ∗ 44 cannot be masked. A score above 256 can
scores TAI = 1 + 0 + 16 + 64 + 256 = 337 only be achieved with at least 1 failure
assigned in TAImax = 4 ∗ 44 = 1024 mode assessed as being red.
the matrix)

29
 Condition assessment of power transformers

4 Numerical Estimated Failure Probabilities (for this Although not immediately obvious, TAI
Score using example)6 are: scores for transformers with one or more
estimated FM1 – Red = 0.1 (10%) failure modes that are Red, will have a
failure FM4 – Pink = 0.03 (3%) higher score than transformers which do
probabilities FM3 – Orange = 0.01 (1%) not have Red failure modes.
FM2 – Green = 0.002 (0.2%) Although the methodology is the same as
that used to calculate a Probability of
TAI=1 – (1-0.1)x(1-0.03)x(1-0.01)x(1-0.002) Failure, it cannot be assumed that the
TAI = 0.1375 score is indicative of the expected PoF,
since the probabilities for each failure
mode were only estimates. This
methodology is only intended to allow
ranking of transformers.
5 Worst case TAImax = 𝐑𝐞𝐝 𝐨𝐫 𝐄 The score highlights that at least on
Failure mode has been assessed as being
5a Hybrid TAI = 1E or 1 Red in the Red category
Score This score indicates the number of failure
modes assessed as Red
6 Count per The score clearly shows the number of
category failure modes in each category.
1E 1D 1C 0B 1A Scores can be easily ordered by the
number of failure modes in Category E,
then D etc.
7 Machine Not provided
learning

Limitations of a Transformer Assessment Score and Index

Users should be aware of the following potential problems when using scores and indices.
The score may hide major problems
If the transformer’s total assessment score is either the sum or average of the scores from each of the
transformer’s failure modes, the overall transformer assessment score may mask a bad failure mode.
If most of the failure mode scores are good and only one, or a small number of, failure mode scores
are bad, the overall assessment will hide or mask a problem in the transformer that needs to be
addressed more urgently. For example, a transformer in overall very good condition, except for a
single oil impregnated paper bushing which is in poor condition, may have a good overall assessment
score.
If this type of scoring system is used it is essential that the major transformer failure modes are
identified and assessed by analysis of the individual test results or condition information, as the overall
assessment score may mask problems which need to be addressed urgently.
Another option is to use a “worst case” scoring system in conjunction with the average score, to
ensure that major problems are highlighted. Of course, this method may also have limitations.
The score has limitations
It is essential that users understand what information is used to determine the transformer assessment
score. A user may develop a scoring system which scores only the active part of the transformer. This
means that components such as bushings and tap-changers would not be included in the assessment.
All users must understand the limitations and have some other method to ensure that those
transformers with bushing or tap-changers in poor condition are identified. It is essential that the
purpose and limitations of the model are clearly defined and communicated to all users of the index.

6
The estimated failure probabilities have been chosen only to illustrate the methodology. Users should
estimate their own failure probabilities based on knowledge of their fleet and historical data. To
improve the accuracy of the ranking users should try to ensure that the relativity of the probabilities
assigned are realistic.

30
 Condition assessment of power transformers

Quality of the data

The quality of the assessment score will not only be a function of the design of the TAI, but will also be
a function of the accuracy, completeness and timeliness of the data. In many cases, the data available
for scoring will not be current. For large fleets, the testing of transformers may be spread over a period
of several years. For example, dielectric fluid samples for DGA may be taken annually, but bushings
might only be tested every 5 years during a scheduled outage. This means that some test results may
be several years old7. For some transformers, the last test may not have been completed as
scheduled e.g. transformer not available for an outage. In this case, the data will be even older.
Furthermore, some tests results may be incorrect and need to be discarded.
It is essential that the effect on the score of this missing or low-quality data is understood. In some
cases, it will make little difference to the overall score. In other cases it could make a bad transformer
look good; resulting in the transformer being overlooked for critical repair or replacement. Similarly, the
missing or low-quality data can make a good transformer look bad.
Implementing a “quality score” or some other indication of the quality of the data used (typically called
“significance”) may help to understand the quality of data and the reliability or confidence level of a
score. Refer to 4 Dealing with uncertainty in information.
If users implement any kind of scoring system (particularly an automated scoring system), it is
essential that they consider data quality. The system must include data validation to ensure that data
that is obviously incorrect is not used and that missing data is treated appropriately, as detailed in 4.
Furthermore, the automated system should also calculate a quality or confidence level score. A
transformer with a good assessment score based on low quality data, might actually be a transformer
with a major problem!

Examples – How to Develop a Transformer Assessment Index

For detailed examples showing how to develop a TAI refer to Annex B.

7
Users who want a high level of confidence from their TAI should consider the frequency of testing
and inspection necessary to ensure that the data used for assessment is current. Quality scoring
should reflect the rate of change of degradation of the failure mode being considered.

31
Condition assessment of power transformers

32
 Condition assessment of power transformers

Refurbishment index
Asset Managers will sometimes decide to perform maintenance or repair work on a transformer that is
not immediately necessary for the continued reliable operation of the unit. In this chapter, this work is
referred to as refurbishment. This chapter describes how indices can be used to help Asset Managers
determine which transformers should be considered for refurbishment work. Previous chapters have
described how indices can be used to identify transformers that are in poor condition and should be
considered for replacement or maintenance. These indices are not appropriate for determining which
transformers should be considered for refurbishment work.
Some users believe that an intervention, prior to significant degradation or major component failure, is
the most economical option in asset availability, cost control and residual value.
This work is normally done for one of two reasons.
1. To eliminate the need to perform this work in the future. This will possibly eliminate the need to
arrange another outage of the transformer and the requirement to have a work team return to the
site. Refurbishment is indeed an opportunity to perform works on transformers on one occasion
rather than numerous maintenance interventions.

2. To extend the life of the transformer. For example, if a transformer has a high moisture level and
acidic dielectric fluid, it is unlikely that these problems will cause an immediate failure of the
transformer. However, if left uncorrected, they will accelerate the degradation of the cellulose
insulation, possibly reducing the transformer’s life. The Asset Manager (after an economic
assessment) may choose to remove the excess moisture and to arrange replacement or treatment
of the dielectric fluid as part of a maintenance strategy that could extend the transformer’s life.
It is not within the scope of this Technical Brochure to determine what maintenance interventions
should be performed during a refurbishment or to determine the economic justification for the
refurbishment. Users should refer to CIGRE Technical Brochure “Guide for Transformer Maintenance”
(TB 445) [5] as well as C57-140™ IEEE Guide for the Evaluation and Reconditioning of Liquid
Immersed Power Transformers [6], for information regarding maintenance tasks that should be
considered.
Some failure modes of a transformer can be readily corrected. For example, a failed cooling fan can
be easily replaced. However other failure modes are not economical to correct. For example, badly
degraded cellulose insulation could only be corrected by replacement of the windings. Similarly, many
internal faults would require de-tanking, removal of the top yoke of the core and possibly replacement
of part of the winding. This work would generally not be considered economical; particularly if the
transformer has other defects that need to be corrected. Some failure modes may need more detailed
consideration. For example, major repairs or replacement of a tap-changer may be relatively simple if
suitable parts or a replacement unit is readily available. However, in some cases, major modifications
to the tank may be required to fit a replacement tap-changer if a similar unit is not available.

Application of the transformer assessment index to identify eligible

candidates for refurbishment
A simple way to identify transformers suitable for refurbishment consists of using a Transformer
Assessment Index (TAI) Index and appropriate filters.
Filters can be applied to failure modes that cannot be easily repaired. The filters can be used to
exclude transformers from the ranking list that score pink or red colours for these chosen failure
modes. This approach assumes that a consistent and correct categorisation of the failure modes has
been applied to the transformers. A single well advanced and non-repairable failure mode that is not
correctly detected can be sufficient to nullify the potential refurbishment benefits and will be a major
reliability risk for the transformer following the refurbishment.
Refurbishment is typically applicable to transformers that have not revealed any significant
abnormality of non-repairable components. Both technical and economic aspects need to be
considered.

33
 Condition assessment of power transformers

Failure modes to be used as a Filter

Filtering of specific failures modes should be consistent with the fleet characteristics and user’s
objectives. When a transformer reveals critical defects for the chosen failure mode then it should not
be considered for refurbishment. Deriving a complete list that satisfies all identified defects is
impossible, but the user can refer to the guidance below to select failure modes that should be used
for filtering to prevent a transformer being considered for refurbishment:
 Failure mode has no repair solution available
 Repair of failure mode is unlikely to be cost-effective
 Damage is irreversible
A fundamental driver for the decision to refurbish is the active part condition, especially the cellulose
and solid insulation that often determines the remaining life of the transformer. Dielectric, mechanical
and thermal defects in the winding and core typically require too much effort to be cost-effectively
repaired. Most users would not consider a refurbishment if the active part was in poor condition and
the transformer therefore had a limited remaining life.
A list of common failure modes to filter would include: design issues, compatibility issues of old
components, core insulation, core fault, winding fault, dielectric fault, thermal fault, advanced paper
degradation, winding movement, corrosive sulphur and major sludging issues that are likely to affect
the flow of insulating fluids. There are numerous others that the user can add to the list drawing on a
user’s experience from their own fleet.
Failure modes to include in the Refurbishment Index
Final candidates for refurbishment are generally transformers that:
 require significant maintenance work (particularly in relation to accessories)
 show normal or reduced ageing and degradation of the active part compared to expectations
based on loading and operating conditions
 do not have other failure modes that are likely to significantly limit the life of the refurbished
transformer.
This approach assumes a consistent determination of the repairable conditions. Deriving a complete
list that satisfies all identified cases is difficult, but the user can refer to the guidance below to select
relevant failure modes:
 failures modes related to components and accessories that can be replaced cost-effectively
 degradation that can be reversed or corrected (in contrast to irreversible damage)
A list of common failure modes that should be considered during a refurbishment include: failure
modes related to bushings, pumps, fans, tap-changers, control equipment, moisture, clamping
pressure, degraded dielectric fluid, corrosion, paint, dielectric fluid leaks and gaskets.
Repairable failure modes can be divided into two categories; those that will degrade or age the
transformer, and those that may have an effect on the operation or function of the transformer but do
not necessarily accelerate ageing.
Other failure modes need only to be considered during a refurbishment if they are scored as red (i.e.
need attention in the short term to minimise the risk of failure)
A TAI can be used for refurbishment decisions to identify transformers for refurbishment based on
their technical assessment. Criticality, economic considerations, losses etc. should also be
considered. Refer to 6 Criticality and other condition considerations. Furthermore, a detailed economic
assessment which considers the estimated cost of repair (refer to CIGRE TB 248 Guide on Economics
of Transformer Management) and the expected remaining life of the refurbished unit should be carried
out before a transformer is refurbished.

34
 Condition assessment of power transformers

Application of the TAI to estimate the potential benefits from

refurbishment
After obtaining a list of eligible candidates ranked by their overall condition and derived using the filters
detailed above, a mitigated index ranking can be established considering all of the failure modes that
can be repaired and effectively corrected. This index can be compared to the original TAI (before
refurbishment work). This comparison enables the estimate of interventions required and the derived
benefits from the refurbishment and allows the user to identify the transformers where the
difference/gap between these two indices is high. However, the value of a mitigated score will be
somewhat dependent upon the scoring system used. Many numeric scoring systems result in a value
that allows transformers to be ranked; but the score has no real meaning. Users must understand the
scoring system to understand the relevance and value of using mitigated scores. Refer to section
2.2.6.

35
Condition assessment of power transformers

36
 Condition assessment of power transformers

Dealing with uncertainty in information

All assessments of transformers include a non-avoidable level of uncertainty. This level of uncertainty
is based on the fact that degradation may not act in a predictable manner and that transformer
degradation is not perfectly assessed by the known condition indicators. The user of a transformer
assessment or index should always consider the basic level of uncertainty of their assessments. This
level of uncertainty is influenced by the data quality used in the assessment. If too much confidence is
given to the assessment result, incorrect decisions might be taken, e.g. a bad assessment result may
have led to a replacement decision where in reality additional information could have resulted in a
better solution.
Unavailability of data or incorrect and/or obsolete data will adversely affect the assessment results and
increase the level of uncertainty. Therefore, an indication of the uncertainty should be part of the
decision-making process.
It would generally be preferable if a transformer is able to be assessed and an assessment score
generated, even if some data is missing or does not match the required attributes. However, a quality
score or confidence score should be generated with the assessment so that the Asset Manager
understands that they should have limited confidence in the score.
In some cases, it may be appropriate to terminate an assessment if quality data is not available.
However, it is essential that transformers that have not been assessed are highlighted to the user, so
that they can arrange for quality data to be obtained (e.g. transformers could be re tested) where
appropriate.
When a Transformer Assessment Index (TAI) is designed, or an assessment methodology
determined, it should initially be assumed that quality data will be available for the assessment of the
transformer fleet. However, as it is likely that quality data will not be available for all transformers in a
fleet, the user should determine how the TAI manages uncertainty due to inaccurate data or
unavailable data.
This chapter discusses techniques to manage unavailable and/or poor or obsolete data. The more
accurate the data that is available to feed into the TAI, the more likely the Index will be useful. The
designer of the TAI should be aware that lower value assets often have less data available for
assessment, leading to a higher level of uncertainty in the TAI result.
A reference to the levels of uncertainty should generally be included with the TAI. The uncertainty
requires its own scoring methodology and can be presented with the TAI score. It could also be
presented as separate dimension of the TAI score.
Uncertainty can be expressed using an uncertainty index, a colour code, or the TAI score can be
shown as a range to represent the uncertainty i.e. falling between a minimum and maximum. If the TAI
score is presented using a colour code, the colour intensity can be also used to show the levels of
uncertainty.

Dealing with uncertainty within available information

Uncertainty within available data may have different causes. The following three causes will be
discussed in more detail in this section:
 Incorrect data entry, or erroneous or questionable test results
 Uncertainty in the condition assessment
 Aged data
Incorrect data entry
Specifically, if data is entered manually, typing errors or spelling errors may cause incorrect data. To
enable correct TAI-results it is recommended that the data entry is validated.

37
 Condition assessment of power transformers

The following table shows types of data validation.

Table 4-1 Data validation methodologies

Type of solution Purpose Level of Pros and Cons

in data complexity
validation
Comparison of Compares data to a known data set; e.g. Low Pro: Validates within a preset
data / table look raises a flag if an entered substation of valid options
up name does not exist. Con: small typing errors or
language issues may cause
unwanted validation failure
Data type / unit Checks if for example the data in a date Low Pro: always good format of
field is entered as yyyy-mm-dd. data
Con: no value checking, only
the type of data format
Digit check Checks the number of digits or Low Pro: simple to install
characters Con: Limited validation
capabilities
Limits and Raises a flag if the entered data is Medium Pro: good check for mistyped
ranges outside the possible range, e.g. numerical values, when fall
manufacturing date older than 1886 out of predefined ranges
(first commercially available Con: no validation within the
transformer) is considered incorrect, or predefined ranges.
the make and brand is spelled correctly Predefined ranges need to be
according to a predefined list, or designed carefully. Outliers
measurement values are within a need to be managed.
certain range, e.g. power factor must be
between 0 and 1.
Consistency If the asset under consideration is a Medium Pro: Only relevant data is
transformer, the data attributes should accepted
only allow transformer related data Con: No content validation
Chronological Check if for example an inspection data Medium Pro: Very useful in date and
order is not before the manufacture data time sensitive data
Con: Only valid in case of
dates and time
Cross reference Checks if data is consistent with other Medium Pro: Very useful as an
data attributes. For example, the actual enhanced range validation
load of the transformer cannot exceed a Con: requires the reference
given range depending on voltage class. data attributes to be
accurately available
Incremental Checks if incrementally entered data is Medium Pro: Very useful as an
data not exceeding an expected value. A enhanced range validation in
maximum number of operations can be repeatable data attributes
calculated based upon historical data. If Con: requires historical data
the counter reading of a new reading is
exceeding the expected number a
validation warning could be triggered

Uncertainty in the condition assessment

The accuracy of condition assessment techniques varies between the different techniques. This
means that each condition assessment technique does have a certain level of uncertainty. The user
and designer of the TAI should be aware of this effect. Small uncertainties may be ignored, having
little effect on the decision. But higher uncertainties may affect the decision. In the case of high
uncertainties, the decision maker needs to be informed, enabling them to adjust the intervention
decision. Failure modes can be assessed using multiple techniques, as discussed in later chapters
which describe the different failure modes and condition indicators. The likelihood of detection may

38
 Condition assessment of power transformers

vary based on the method. Therefore, the techniques with a higher likelihood of detection will have a
lower level of uncertainty. If needed, additional techniques should be used to perform a more accurate
condition assessment.
A few examples are:
 Partial discharges may be detected by DGA, however DGA will most likely not provide detailed
information on the location and type of PD-activity. Therefore, an intervention is usually not
performed on DGA results only. A second, potentially more accurate, PD-measurement (e.g.
UHF) is recommended whenever DGA shows PD-activity.
 Manually performed inspections will have a subjective interpretation with inherent levels of
uncertainty. The inspection regime should be designed to minimize the uncertainty, by calibration,
training, use of qualified inspectors, and approved 3rd party service providers such as
laboratories. Still, in some cases, a level of uncertainty could be included in the TAI for subjective
condition indicators.
A simple approach calculates the TAI with the minimum, maximum and average value of the uncertain
input parameter and then shows these results to the decision maker e.g. by using a boxplot. This
works fine in situations with only one or two uncertain parameters. In the case where more parameters
have uncertainty a more advanced TAI is suggested using statistical techniques to show the levels of
uncertainty. These techniques interpret data attributes as distributions, rather than single data points
(scalars), and use more advanced techniques to calculate the final results. An example that can be
used is Monte Carlo simulations, as discussed in section 4.2.5. and 4.5.2.
Uncertainty due to old data
A condition assessment provides information of the asset’s condition at the time of measurement or
assessment. Degradation is however continuing after the assessment. It should be recognized that a
test report being old, does not inherently mean it is out of date. It only becomes out of date if the
underlying circumstances change so that the test result no longer reflects the current status of the
transformer. However, it is generally impossible to determine this without repeating the test. Therefore,
the assumption is made that over time, the test result becomes less and less likely to represent the
current condition of the transformer. In other words, the uncertainty of the assessment result increases
in time. Whenever using older condition assessment data in the TAI the decision maker should
preferably be informed, that older data is used, so they can take appropriate actions.
Bouzeghou and Peralta [7] Introduced terminology for describing the different aspects of timeliness.
They differentiate between currency (the time since the measurement was made) and obsolescence
(how much the underlying value has changed since the measurement was taken). With regard to the
whole database, they define a freshness rate, which is the percentage of records which are not
obsolete. Finally, they define a timeliness period as being an application-specific length of time in
which a piece of data may be considered not obsolete.
Sparling and Catterson [8] show how this work can be applied in the cases of a transformer where
there is uncertainty about whether or not a test result is obsolete. The currency can be calculated from
the known date of a given test, but the challenge lies in determining whether that test result is obsolete
given the time elapsed since it was taken. The timeliness period (as defined by Bouzeghou and
Peralta) can be used to help assess the likelihood of obsolescence.
As an example, two DGA tests a year apart may show low absolute values of key gases and very little
change within the year. In this case, it would be reasonable to assume that the test report is unlikely to
become obsolete within the next year, and that the likelihood of the test result becoming obsolete
increases with each subsequent year. However, the timeliness of a report is highly dependent on the
test type and test result. If the two DGA reports show a rapid rate of change over the year, the
timeliness period would shorten significantly, and the chance of obsolescence one year later is high.
Table 4-3 suggests a method of scoring or evaluating the uncertainty in the TAI score. If multiple data
is missing or out of date (obsolete), it should result in a low level of confidence in the TAI. If weighting
factors have been used in the TAI, similar factors should be used in the Data Quality Score, as used in
the TAI.
Whenever using a more advanced TAI, statistical techniques, such as Monte Carlo analysis, can be
used to show the levels of uncertainty due to old data. Here data attributes are interpreted as
distributions rather than single data points (scalars). In this technique, the uncertainty in the
distribution (spread) increases with the age of the condition assessment result. To illustrate this
technique, an example is provided taking the DC resistance of a transformer winding.

39
 Condition assessment of power transformers

Example: The actual DC winding resistance, measured during the routine test, is likely to stay
constant in time, but may increase due to degradation mechanisms. Contacts in the winding circuit
potentially degrade leading to higher contact resistance, or connections may come lose. Given this
degradation mechanism, the likelihood of highly increased winding resistance is low, and the
winding resistance will not decrease in time. For this example, an exponential distribution can be
used when estimating the actual DC winding resistance if the winding resistance measurements
are old. Figure 4-1 shows the exponential distribution’s shape, providing a shift parameter and the
shape parameter β. In this graph the shape parameter is fixed. The shift of the exponential
distribution is to be determined by the actual DC winding resistance measurement result. The
shape parameter β has to be determined by the level of uncertainty. This can be done by pattern
fitting from other available similar data, or by experts’ design. The level of uncertainty of the
measurement increases with the age of measurement. If a proper DC winding resistance has just
been measured the level of uncertainty is rather low, however if many years have passed since
this measurement, the level of uncertainty is high. Therefore, the shape parameter β from the
exponential distribution describing the measurement data is time dependent. Figure 4-2 shows a
three-dimensional graph with the age of the measurement on the third axis. Please note that this
distribution describes the most likely behaviour of the input parameter (measurement in the
example). It is not related to the failure probability of the transformer.

Figure 4-1 Exponential distribution

Figure 4-2 Exponential distribution with time dependent shape

40
 Condition assessment of power transformers

Incorrect data
In some cases, data validation shows that the data is simply incorrect, while correct data is not
available. Data enhancement actions are preferred prior to running the assessment tool. However, this
may not be feasible in all cases. Whenever incorrect data is found, it is recommended to discard the
incorrect data and treat it as missing data, with the options provided in section 4.2.
As older data is not necessarily incorrect, it may be appropriate to use older where more recent data is
obviously incorrect. As discussed in section 4.1.3, this data can be used in conjunction with a level of
uncertainty depending on the relative age of the data (age of the data compared to average
degradation speed).

Dealing with unavailable Information

Once a transformer assessment index has been developed, the required information is determined by
the condition indicators that are included in the design of the TAI. It is unlikely that an engineer will
have all the required information readily available to generate a transformer assessment index for an
entire fleet of assets. In order to deal with any unavailable information and to still generate a
meaningful TAI, the TAI should be robust to prevent it from providing no answer or a false answer. If
the index stops and does not provide an answer, the decision maker is still not informed on the correct
status and there is a high likelihood that the decision maker is ignoring this transformer, risking
unknown bad conditions.
Preferably the assessment index provides a most likely answer including an indicator showing that
information is missing and providing an output range or level of uncertainty due to the unavailability of
information. This prevents over-confidence in the TAI result when information is missing. When
information is missing sensitivity analysis is recommended to assess the impact of the missing
information on the assessment index score. This section describes some techniques which can be
used if required information is missing.
To perform any analysis at all, a minimum set of information is required. If the available information,
related to the transformers under study is too limited, the analysis will not result in a sensible
transformer assessment index result. The techniques described in this section, requires sufficient
reference information from similar assets or parameters to perform some manner of statistical
analysis. It may be that little or no information at all is available or will become available for some
parameters. In this case, the engineer should re-affirm the need for the parameter to be included in
the assessment index.
The methods mentioned in Table 4-2, describe a number of ways in which missing input parameters
can be estimated. Using estimates however, includes a level of uncertainty, i.e. estimations give an
indication of the possible and even likely values of an input parameter, but the actual value is not
known. Therefore, this estimate influences the transformer assessment index result and confidence
level. Where a single parameter is not known, the determination of the transformer assessment index
and uncertainty of the result can be easily derived using statistical or algebraic calculations. Where
multiple parameters are unknown, algebraic calculations often become too complex. Using an iterative
method of uncertainty modelling, such as Monte Carlo analysis, may be preferred.
The level of complexity of the design of the transformer assessment index is dependent on its purpose
and the level of information at hand. Whenever an asset manager has only a limited number of
transformers, it most likely is not worthwhile to pay lots of effort in making a complex transformer
assessment index. However, if an asset manager is responsible for vast numbers of transformers, it
becomes rather relevant to design a transformer assessment index, which is accurate and provides
sufficient granularity even in situations where information is partly not available. Bigger granularity
might be required whenever larger transformer fleets are assessed.

41
 Condition assessment of power transformers

Table 4-2 Different options for managing of unavailability of information in assessment index design

Type of solution for Purpose Level of Pros and Cons

managing complexity
unavailable
information
1 Stop assessment, or Low number of assets e.g. less Extremely Pro: Simple to design,
ignore missing data than 10. The assessment index low low cost. Easy to
and manually assess algorithm cannot be fully identify missing data.
automated, and the engineer must Con: No decision
assess the health of the support, risk of missing
transformer manually. bad transformers.
Manual assessment
required.
2a Use a default value Low to medium number of assets Low Pro: Simple to design,
e.g. less than 100. The default low cost.
value can vary e.g. Best case, Most Con: No decision
likely case, or Worst case. It support, risk of missing
enables an index result at all bad transformers. No
times, but may be inaccurate. indication when data is
missing.
2b Use a default with a A default value is provided Medium Pro: Index result is
range including a possible range of that provided, range is
value. The index result now provided (level of
provides a range of possible certainty).
results, e.g. min, max, and average Con: More complex to
or median. design. The range may
not be tuned to local
circumstances.
3a Use of statistical High number of assets, e.g. more High Pro: Level of
inference on limited than 100, with limited uncertainty is provided;
number of input unavailability of data. a range is provided. The
parameters The level of uncertainty is range can be adopted
provided for 1 or 2 input to local circumstances.
parameters. The level of Con: Method is only
uncertainty is mathematically applicable if only 1 or 2
calculated and shown in the index input parameters are
result. missing. Limited
capabilities in adjusting
towards individual
transformer
circumstances.
Statistical data is
required.
3b Use of statistical High number of assets with a High Pro: Level of
inference on many varying or high number of uncertainty is provided;
input parameters unavailable data. The level of a range is provided. The
uncertainty is provided for many range can be adopted
input parameters. The level of to local circumstances.
uncertainty is calculated Capable of handling
numerically (e.g. using Monte many missing input
Carlo simulations) and shown in parameters.
the index result. Con: Limited ability to
adjust for individual
transformer
circumstances.

42
 Condition assessment of power transformers

Statistical data is
required.
4 Imputation using High number of assets with a High Pro: Level of
external and local varying or high number of uncertainty is provided,
circumstances unavailable data varying with local a range is provided. The
circumstances. range can be adopted
Known physical influencing factors to individual
are considered in the analysis, transformer
allowing a statistical assessment circumstances.
imputing the likely value of Con: impact of local
unavailable input parameters, influencing factors and
including levels of uncertainty. statistical data is
required.
5 Machine learning Machine learning can be used to Highest Pro: The systems are
assess the missing input flexible and can provide
parameters. promising results and
potentially makes new
correlations and
insights
Con: Requires high
volumes of information.
Costly to implement.

The above-mentioned methodologies 2 – 5 enhance the dataset and fill in the missing gaps in case
data is missing. Options 3, 4 and 5 include capabilities to calculate and show the levels of uncertainty.
Option 1 and option 2 are not able to do so. For those options a confidence indicator can be simulated
by using a data completeness index. This index calculates the percentage of data points provided. If
needed, weights can be added per datatype, allowing higher weighting factors for more critical data
and therefore having a bigger impact in the data completeness index. Further details of these methods
are given below.
Stop assessment, or ignore missing data and manually assess
These methods are mostly used whenever the number of assets to be studied is relatively low.
When data is missing the engineer has the option to:
 Stop the generation of the transformer assessment index, providing no result for the transformer
in question. In this case, the missing data would need to be obtained to allow the assessment to
be completed, or
 Manually completing the TAI assessment without the missing data. This may be possible for some
TAI designs. For example, a worst-case score can still be generated without data to assess one
or more failure modes/mechanisms. However, for other scoring systems the scores would need to
be adjusted if all failure modes/mechanisms are not assessed e.g. where the TAI score equals the
sum of individual failure mode assessments scores,
This option is only recommended when the fleet comprises a small number of transformers, and the
engineer can verify the status of the transformer.
Use a default value
An automated algorithm can always provide an assessment when input parameters are missing by
using default values. However, the answer may be misleading, and the use of default values may not
be visible to the decision maker when they are analysing the results of the assessment index.
In this option, the engineer needs to choose the desired default value. For the case of a best-case
default (“Innocent until proven Guilty”), the assessment index will always be biased towards a better-
condition score, with a high likelihood of overestimating the quality of the transformer. Winding
deformation can be used as an example. Typically, specialized tests like SFRA are used to assess
winding deformation. Most utilities will not perform this periodically, therefore this information is not
available. As a default, the user can choose “no deformation” until an SFRA measurement results
proves differently. When using a worst-case approach (“Guilty until proven Innocent”) the assessment
index result will always be pessimistic in the case of missing information, providing an early warning.
This approach has a potential risk that the asset manager is going to ignore the early warnings, if the

43
 Condition assessment of power transformers

warning turns out to be false too often. To overcome this the best solution is using the most likely
value (expected value, median or average). This however needs an assessment of the input
parameter in general to indicate what the expected value would be and may need further
consideration of local circumstances and basic statistics. When using this approach, the type of
default should be carefully considered.
Use a default with a range
Like the previous option, however, the default is now a range of values providing the minimum,
maximum, and if possible, the average or median. The assessment index is now calculated for all
values and the results are presented as a range, providing an indication of the possible results and the
impact of the unavailability of the information. In a smart design of the assessment index, the ranges
can vary based on the type of data and the type of transformer, allowing better alignment of the
assessment index default values. This approach is very useful but increases the complexity of the tool.
The user needs to consider that this approach requires statistical information. The way the results are
distributed within the range is unknown, but this approach can provide a median, minimum and
maximum value providing high level insights into the level of uncertainty.
Use of statistical inference on limited number of input parameters
Statistical inference can be used in situations where it is known which input parameters are potentially
missing and if the number of missing input parameters is limited to 1 or 2. In these situations
straightforward mathematical algorithms can be used to calculate the impact of the uncertainty of the
input parameter in the result of the transformer assessment index, provided that the input parameters
distribution is known or estimated. In this methodology, the missing input parameters are estimated by
a distribution, providing the likelihood of the parameter value. These distributions can be of various
designs, e.g. a discrete distribution when only a limited number of values are possible, a normal
distribution, a Weibull distribution or exponential distribution. The type of distribution selected is
dependent upon the behaviour of the unavailable input parameter, and can be assessed by statistical
inference, provided that the information used for the inference is valid for the type of transformer and
the input parameter. Please note that this type of analysis cannot differentiate between different local
circumstances for individual transformers.
Use of statistical inference on many input parameters
This method uses the same methodology as the previous option, but now the potential number of
available input parameters is too high for algebraic calculations, e.g. more than 1 or 2 of unavailable
input parameters. In these cases, it becomes too complicated to calculate the impact of the
uncertainty to the TAI result. To manage this Monte Carlo simulations can be used. This is a numerical
methodology, using random sampling techniques and multiple calculations, to calculate the final
assessment index results. This is a more complex methodology, but it provides significant insights into
the level of confidence and the expected assessment index result distribution. The value of these
more complex tools can be found in situations with many transformers and many or a variable number
of missing input parameters.
Imputation using external and local circumstances
This method can be used whenever multiple input parameters are missing and local circumstances for
individual transformers are varying and may have a high impact on the transformer assessment index
result. In those cases, it is preferred to use imputation techniques to assess the likelihood of an input
parameter’s value based upon known degradation factors or accelerators. In these cases, individual
transformer circumstances are included in the assessment of the input parameter. As an example, if
one needs to know the level of corrosion of a transformer radiator, but the inspection information is not
provided; one can include the location information to assess the likelihood of corrosion, e.g. outdoor
transformers are more susceptible to corrosion if located near the sea shore or polluting industries. It
is recommended that this type of data imputation includes levels of uncertainty and is used in
combination with the techniques mentioned in option 3.
Machine learning
This method uses modern algorithms e.g. neural networks, to learn the behaviour and failure
characteristics of transformers. These techniques may find some new correlations and insights with
possibly very positive results, however, to achieve reliable results, a large amount of data is required
and potentially a long learning curve is required. In this brochure, these techniques are not discussed
in detail.

44
 Condition assessment of power transformers

Options for showing uncertainty

It is essential that the end user of a TAI has an understanding of the level of uncertainty before making
any decisions based on the TAI. There are multiple ways of showing levels of uncertainty, some of
which have already been discussed.
The output of the TAI could be expressed as a range to show the uncertainty. The range could show
the minimum, maximum and possibly the average value, or the output could be shown as a boxplot.
Refer to example 4.4.2.
A separate index can also be provided to give an indication of the level of uncertainty. The index could
be presented as a numeric or alphabetical score, a colour code (e.g. Green, Orange Red) or the
uncertainty could be shown as a colour shading intensity where the TAI output is already expressed
as a colour.
To ensure that any measure of uncertainty is assessed and represented consistently, user should
consider developing a data quality scoring matrix. A simple example is provided below.
Table 4-3 Example table of data quality indication

Evaluation of the transformer data quality

Quality
Results
Indicator

All the tests related on the main indicators were performed with the normal
10
frequency and the results obtained were reliable.

One or more of the tests were performed more than 6 months ago and less
than 24 months ago, after the date corresponding to their normal frequency 7
and the results obtained were reliable.

One or more of the tests were performed more than 24 months ago and less
than 36 months ago, after the date corresponding to their normal frequency or
some of the results are not available or are questionable. 4

One or more of the tests were performed more than 36 months ago, after the
date corresponding to their normal frequency, or there is no available results, or 0
many of them are questionable.

Examples
For the following examples, which illustrate the methods discussed in Table 4-2, a standard case is
used. It is assumed that a simple index is based on 4 failure modes: FM1 - FM4. Three failure modes
have been assessed using a scoring matrix. Input for one of failure modes is however missing. We
assume the TAI is designed using a linear summation principle: 𝑇𝐴𝐼 = ∑4𝑖=1 𝑆𝐹𝑀𝑖 . The maximum score
of individual failure modes is 25, hence the maximum TAI-score for a good performing transformer is
100.
 FM1: Paper degradation: SFM1 = 25 (DP = 850)
 FM2: Dielectric strength: SFM2 = 20 (DGA Values OK, but BDV=55 kV)
 FM3: Bushing: SFM3 = 15 (tan delta = initial value * 1.5)
 FM4: Mechanical structure: SFM4 = unknown (No SFRA measurement)

45
 Condition assessment of power transformers

Stop assessment, or ignore missing data and manually assess

When using this methodology, the ideal situation would be to assess the mechanical structure so that
data was available to properly assess the transformer. However, for this example it is considered not
feasible to collect that data. If not corrected for the missing data, the score of the TAI would now be 60
(from a maximum score of 75). The scoring needs to be recalibrated such that the maximum score
sums up to 100 again. For this example, the corrected score would be
TAI = 60 * 100 / 75 = 80
Use a default value
When using a default for the missing information, a standard value will be used instead of the missing
data. For this example, the SFRA data is missing. The SFRA result could either be “Good condition”
(“Agreement with benchmark, or acceptable phase comparisons”) or “Very poor” (“Signature
differences detected”). When using the “innocent until proven guilty” approach the default will apply
the value “Good condition” and the score for FM4 will be 25. The TAI would score 85. When using a
“Guilty until proven innocent” approach the value “Very poor” and the score for FM4 would be 0. The
TAI would be 60
When using an average scoring as a default, the designer needs to know the mechanical deformation
distribution. They can calculate that by counting the total operational transformers of the population of
transformers not having mechanical deformation and compare that with the number of transformers
with mechanical degradation. This analysis should be performed with a timeframe long enough to
collect sufficient data e.g. 10 years. Assuming that within this timeframe the total transformer
population was 25 and two transformers showed mechanical deformation. The average score for FM4
would then result in 23 and the TAI would be 83.
When using these methods, the decision maker is potentially unaware if data was missing, and
overconfidence in the assessment result may occur. By using a default value with a range, the score is
not a single value. In this case the range could be provided with the following parameters:
 Min: “Very poor” with SFM4 = 0
 Average: SFM4 = 23
 Max: “Good condition” with SFM4 = 25
The TAI now provides a scoring range which can be shown in a boxplot:
TAImin = 60, TAIaverage = 83, TAImax = 85
Using statistical inference
When using this option, the inference is automatically applied by the TAI-algorithm. Using this
methodology, the TAI will look into the database of similar asset types and checks the values for the
data point in those assets. Consider the example that in an asset base of 100 similar transformers,
80% of the transformers have relevant SFRA results. Ten transformers show differences in SFRA
signatures. This will mean that using statistical inference a discrete distribution can be used with the
following results:
 Very poor: 10/80*100% = 12.5%
 Good: 70/80*100% = 87.5%
This results in a TAI result, also presented as a discrete distribution:
TAI=85 (87.5%); TAI=60 (12.5%)
When only one or two parameters are missing (and it will always be the same parameters that are
missing), these calculations can be performed algebraically in the TAI. But if the number of missing
data points is varying or higher, then this becomes more complicated. It is then recommended to use
iterative algorithms like Monte Carlo solutions. For this example, it will result in the same result. But
whenever multiple inputs are missing, the expected value is calculated, including the complete TAI
distribution, providing the level of uncertainty and a sensitivity analysis.
Imputation using external and local circumstances
When using imputation in this example, the likelihood of mechanical deformation can be assessed
using information of known mechanical degradation principles, e.g. earth quakes and through faults.
Let’s assume that the transformer is not located in an earth quake sensitive area, so this will not play a
factor for our TAI-design. Through faults can however cause mechanical deformation if the dynamic

46
 Condition assessment of power transformers

forces exceed the design forces of the transformer. Therefore, the number of through faults and the
magnitude of the short circuit current can be used to assess the likelihood of mechanical deformation.
Assuming for this example that the rated maximum dynamic short circuit current Ik’ = 50 kA. For
transformer 1, one through fault occurred in its lifetime with a measured short circuit current
Ik’ = 25 kA. This short circuit current is well below the rated value and it can be expected that the
likelihood of mechanical deformation is low.
For transformer 2 with the same ratings, two through faults have occurred in its lifetime. The first had
an amplitude of Ik’ = 45 kA and the second had an amplitude of Ik’ = 55 kA. The actual short circuit
currents were close to and exceeding the ratings, therefore this transformer has a higher likelihood of
mechanical deformation. Applying a discrete distribution for the two transformers, with an engineer’s
judgement of the level of uncertainty could provide the following results:
Input Mechanical deformation Transformer 1 Transformer 2
“Very poor” 5% 80%
“Good condition” 95% 20%

Whenever more knowledge is available more accurate estimations can be provided for the likelihood
and the levels of uncertainty.
Machine learning
Machine learning and data analytics techniques can provide good results for this type of analysis;
however, a detailed discussion of these techniques fall outside the scope of this brochure.

Additional examples
Single unknown parameter estimation using statistical inference
If a transformer assessment index is missing one single parameter and the impact of this parameter
on the transformer assessment index result is known, relatively simple methods can be used to
estimate the transformer assessment index result by using statistical inference.
To use statistical inference on this missing parameter a batch of similar transformers under similar
conditions should have known values for the same parameter. This provides a distribution of this
parameter. The expected value of this distribution can then be used to calculate the most likely
transformer assessment index result. The spread of the distribution can provide the level of confidence
of the most likely transformer assessment index result. The following example shows a simplified
transformer assessment index with only 1 unavailable parameter, which is assessed using a normal
distribution.
 A utility has a fleet of 132/11 kV transformers, assessed using a 4-parameter non-linear
transformer assessment index. The transformer assessment index comprises four attributes (core
and coils, OLTC, bushings and oil).
 In this fleet, a 40-year old transformer is being assessed. Three out of the four attributes are
assessed to be in acceptable (‘Orange’) condition (Score =10), but the 132 kV RBP (Resin
bonded Paper) bushing condition is unknown. Thus
 HI = 10 +10 + 10 + ? = ? (In this simplified example the assessment index is a sum of four
attributes)
The utility has tested a number of similarly aged bushings, by measuring the dielectric dissipation
factor (DDF), and has used the data to develop a distribution of similar results, shown below. The
results are represented as a normal distribution with parameters μ = 1.2% and σ = 0.35%.

47
 Condition assessment of power transformers

Figure 4-3 Distribution of bushing test results

From the above distribution, the utility can deduce (from the normal distribution) that the most likely
bushing assessment index score is 10, with less than 20% likelihood that the bushing score is 25 or
greater.
Bushing Condition Colour Code (Score) Likelihood Overall Transformer HI
As-new Green (0) 28.4% 30
Acceptable Orange (10) 52.0% 40
Poor Pink (25) 18.4% 55
Poor Red (50) 1.2% 80

The utility is now able to develop a transformer TAI, and articulate a degree of uncertainty around the
score, e.g. the transformer assessment index score is most likely to be 40 with a 52% certainty.
Example: Multiple unknown parameter estimation
When there is more than one unknown parameter, and the unknown parameters may each be
represented by a non-linear distribution, a computational method such as Monte Carlo simulation can
be used.
The Monte Carlo simulation takes the following steps:
1. For each unknown variable in the transformer assessment index, use available data from other
assets or parameters to develop a suitable statistical distribution.

2. Select the suitable statistical distribution and identify the parameters.

3. Randomly choose input values from the default values and uncertainties for all input parameters
from their distributions. A Monte Carlo simulation tool will be managing this.
4. Determine the transformer assessment index results from this dataset through the transformer
assessment index model.
5. Randomly choose a second dataset from the distributions and repeat the sequence 3 and 4, and
so on.
6. Collect all iteration results and plot it as a distribution, then use the most likely value of the median
as the assessment index result and use the spread of the distribution as the uncertainty indicator.
Step 1 in this process is crucial: instead of having a single value for a data-point, a data-point should
be represented by a statistical distribution. This distribution should match the expected behaviour of
the data-point. Types of typically used distributions are: Normal, Weibull, Lognormal, Exponential or a
discrete distribution. The distribution parameters are set by the expected behaviour and the spread is
determined by the level of uncertainty. Preferably these parameters are calculated from applicable
reference data. If suitable reference data is not available, the engineer’s judgement can be used.

48
 Condition assessment of power transformers

Steps 2 to 4 are repeated many times (typically thousands of times). The result of this simulation is a
distribution of outcomes for the transformer assessment index results. The expected value of this
transformer assessment index result distribution provides the most likely transformer assessment
index value. The spread in this distribution represents the level of uncertainty in the transformer
assessment index result, due to the uncertainty in the unknown input parameters. To make the
simulation work all uncertain input parameters will require a distribution rather than a single scalar
value. These distributions can be created using inductive reasoning or statistical inference.
For example, a utility has a fleet of 132/11 kV transformers, assessed using a 4-parameter non-linear
transformer assessment index. The transformer assessment index comprises four attributes (core and
coils, OLTC, bushings and oil).
The utility is trying to assess the assessment index for an aged power transformer with only two
known conditions as the bushing and OLTC condition is unknown. From similar test results, the utility
knows that the bushing DDF reading can be accurately represented by a normal distribution
μ = 1.5% and σ = 0.35%,
and that based on the operating experience of similar transformers, the OLTC has a 70% likelihood of
being in acceptable condition, and 30% likelihood of being in poor condition. The core and coils and
the transformer oil are both assessed as being in good condition.

Figure 4-4 Unknown parameter distribution

Using these distributions, a Monte-Carlo simulation can be performed to estimate the Transformer
assessment index, as well as the possible variance. From the simulation, a median assessment index
of 35 is obtained, which the utility classifies as being in ‘Good’ condition. However, 32% of the
simulation results indicate a possible assessment index result of ‘Acceptable’ or ‘Poor’. The
transformer expert may then decide to assign a ‘Good’ assessment index to the transformer, with a
caveat that there is a 32% chance of the assessment index being worse; the transformer assessment
index is misqualified 32% of the time.
This is quite a high level of uncertainty due to half of the input variables to the assessment index being
unknown. The utility would need to undertake more testing and assessment to increase the certainty
of the TAI result.
Using colour intensity indicators
In many cases transformer assessment index results are presented in colour schemes. If using colour
coding, one can use the colour intensity to indicate the level of uncertainty. In the following example,
Monte Carlo analysis is used to determine the most likely TAI result. The colour intensity code is
determined by the number of Monte Carlo iteration results not matching the colour code of the most
likely TAI result.
Figure 4-5 shows the transformer assessment index result distribution as “Remaining lifetime” and
shows the transformer assessment index results as either Red, Orange or Green.
The concept of misqualification is illustrated by Figure 4-5. Each Monte Carlo iteration result is plotted
in the transformer assessment index result distribution, and each result is assigned a colour coding.
The expected value now represents the most likely result of the Transformer assessment index, in this

49
 Condition assessment of power transformers

case green. The non-green iteration results are now an indication of the level of uncertainty of the
colour green. The larger the number of non-green results in the total distribution, the less certainty of a
green TAI score. In the example of
Figure 4-5 the misqualification is approximately 15%. Therefore using Table 4-4, no action is required.
The level of misqualification will be presented to the engineer though the intensity of the green colour.

Reference
Critical time
Expected
time
Value
Number of simulation results

% misqualification

Remaining lifetime
Figure 4-5 Colour misqualification example

Table 4-4 shows that for the 15% misqualification, the certainty is considered “medium,” and this
determines the colour intensity of the green result. The final transformer assessment index colour
indication is highlighted in the table below with a red outline.
Table 4-4 Colour intensity classification including required actions

Misqualification Certainty Colour & Intensity

<5% Very High No No More Urgent ImmediateAction Immediate

action action Testing Action Action

5% - 10% High No No More Urgent Urgent Action Immediate

action action Testing Testing Action

10% - 25% Medium No More More More Urgent Testing Urgent

action testing Testing Testing Action

>25% Low More More More More More Testing Urgent

testing testing testing Testing Testing

50
 Condition assessment of power transformers

The role of on-line monitoring systems

The increasing capability and decreasing cost of on-line monitoring technology for transformers and
many of their components, means that it is now feasible to cost-effectively monitor the condition and
operation of power transformers in near real time. For the purpose of this Chapter, “On-Line
Monitoring” refers to the sampling or measurement, and where appropriate analysis of the information,
from a device on or near the transformer, whilst the transformer is operating. For example, on-line
systems for dielectric fluid analysis, bushing monitoring or partial discharge measurements.
It is now common practice for many medium and large sized transformers to be fitted with one or more
on-line monitors. The complexity of the monitoring system is generally dependent upon the size and
criticality of the transformer, and the user’s budget.
Some transformers will be fitted with only one monitoring device to monitor a combination of critical
gasses. More complex gas monitors that analyse up to nine dissolved gasses and moisture are
commercially available and common on larger or more critical units. Other types of monitoring systems
allow the monitoring of other parameters such as moisture content, load, temperature and partial
discharge. On-line monitoring devices are also available for bushings and tap-changers.
CIGRE TB 630, “Guide for Transformer Intelligent Condition Monitoring (TICM)” published in
September 2015 [9], and IEEE C57.143-2012 (currently under revision) “Guide for Application for
Monitoring Equipment to Liquid-Immersed Transformers provides” the reader with in-depth information
and recommendations on the application of these on-line monitoring technologies.

Condition assessment using on-line monitors

On-line monitoring devices can be a valuable tool for the condition assessment of power transformers.
Assessment of a transformer fitted with on-line monitors should generally be conducted in accordance
with this Technical Brochure. However, the assessment can be updated more frequently as updated
information becomes available from the monitoring devices.
Many on-line monitoring devices use common analysis methods to detect faults and assess the
condition from the continuously measured data. For example, some of the more complex gas monitors
will not only provide information about each of the measured gasses but will also analyse the results
using methods such as the Key Gas method or the Duval Triangle. The Key Method is a graphical
interpretation technique found in IEEE C57.104-2008 [10], to compare certain key gases found in a
sample to faults conditions where these gases are predominating. The Duval Triangle method is
another graphical interpretation technique used to identify the nature of a fault [11] [12]. Once the
user is satisfied that the methods used by their monitor are appropriate for their transformers and sets
any threshold levels, these devices can greatly simplify the process of transformer condition
assessment.
It is however essential that the user understands that a complete assessment of a transformer should
consider all failure modes. It is therefore necessary to consider the capability, and the limitations of the
online monitoring system being used. Not all transformer components can be assessed using an on-
line monitor, and for components that are monitored, the monitor may not be able to analyse all
condition metrics, and therefore may be unable to detect all failure modes. For example, an on-line
DGA monitor that can measure all key gasses is not able to measure particle count or the dielectric
strength of the dielectric fluid.
A complete assessment of a transformer can only be conducted using on-line data in combination with
other test data and information. The data from the on-line monitor that is used in the assessment can
be regularly updated, based on the latest information from the on-line monitors; whilst the other data
used for the assessment will be updated only when new data is available, e.g. following the next
scheduled transformer outage.
Some critical failure modes of power transformers can be measured and assessed using on-line
monitors. Although an assessment based only on the results of on-line monitors will never be
complete, it is likely to include some of the more critical failure modes of the transformer; and may still
be a very useful tool provided that users are aware of the limitations of such an assessment. For
example, a user should be aware that an assessment based on the results of an on-line gas monitor
does not give any information relating to any critical failure modes of transformer bushings.

51
 Condition assessment of power transformers

Use of on-line monitors with a transformer assessment index

Users of on-line monitoring systems who wish to develop indices to help prioritise transformers within
a fleet for refurbishment, replacement or other purposes, should develop the index in accordance with
this Technical Brochure. In most cases, information about failure modes that are not monitored
continuously will also be included in the index so that the on-line data will be used in conjunction with
other information about the transformers. The data from the on-line monitors used in the index can be
regularly updated, whilst the other data used in the index will be updated only when new data is
available. Refer to Section 5.1. This method ensures that the assessments used within the index are
updated as the information about some of the critical failure modes becomes available.
If a user has a large group of transformers, each fitted with one or more on-line monitors, an index
could be developed using only the data from these monitors. Such an index could be continuously
updated. Although an index based only on the results of on-line monitors will cover only some of the
failure modes of the transformers, it is likely to include some of the more critical failure modes; and
may still be a very useful tool. Again, the user must be aware of the limitations of the index.

Alarms from on-line monitors

One of the major advantages of on-line monitors is the high sampling rate of the parameter being
measured (relative to the alternate test methods). This gives the user significant volumes of data to
trend and analyse, new insights into the performance of the plant, and the ability to observe high rate-
of change events, including rapid deterioration of condition. On-line monitoring systems can be
configured to alert the asset operator when incipient failure modes are observed within the transformer
that necessitates rapid decision making and possible action. If users respond promptly to these
alarms, major failures can often be prevented.
It is important for asset owners and operators to understand the distinction between short-term and
long-term monitoring of the transformer. Transformer Assessment Indices are intended to be used to
assess the transformer over medium to long time periods, whilst on-line monitors are also useful for
monitoring the operation of the transformer in real-time. If an on-line monitoring system brings to
attention a condition measurement that indicates a critical failure could be imminent, users should
respond quickly and appropriately e.g. de-energising a transformer after receiving a critical gas alarm.
Users should not wait until an index or an assessment of the transformer is completed to respond to
an alarm (or other information indicating an imminent failure) from an on-line monitor. However, a
more detailed assessment of the transformers using all available data, may be appropriate before the
asset manager decides how to deal with the problem transformer.
Some users of on-line monitoring devices will also use alternate tests to periodically measure the
same parameters. For example, transformer owners may continue with periodic sampling of dielectric
fluid for dielectric fluid quality and DGA assessment; even where an on-line monitor is installed. If the
different test methods are producing comparable results, the user’s confidence in the data increases.
However, the different techniques, technologies and accuracy of these assessment methods may lead
to different results, and therefore to different interpretations of the defect type or severity. Users would
then need to understand the accuracy of the test and sampling methods used to determine which
measurements are the most accurate; and then use the more accurate test results for critical
decisions.

52
 Condition assessment of power transformers

Criticality and other condition considerations

A transformer asset manager needs to make decisions about when and what type of interventions are
required, e.g. replacement, refurbishment or additional maintenance. During this process, the asset
manager needs to balance the cost of the interventions and the anticipated performance improvement,
against the current risk of the transformer. Therefore, the consequence of a transformer failure will
often need to be considered in conjunction with the condition assessment. Some of the consequences
of a transformer failure are listed below. The list is not intended to be comprehensive but is intended
only to highlight some of the issues that may need to be considered.
The consequence of transformer failure is influenced by
 Possibility of consequential damage or injury from fire or explosion: Fires or
explosions resulting from a transformer failure can damage other transformers and equipment in
the substation, but also result in damage to property outside the substation. This is a major issue
if the substation is in a high-rise building or in a built-up area. If staff are regularly required to work
near the transformer, or if the substation is located near an area with heavy pedestrian traffic, the
possibility of severe injury or death should be considered.
 Possibility of environmental damage: When a transformer is located near a waterway or
some other environmentally sensitive area, the risk of insulating fluids escaping into the
environment must be considered. The consequence will depend on the availability of oil
containment and the type of insulating fluids in the transformer.
 Size and type of load interrupted: Some loads such as hospitals and high-rise buildings with
elevators, transit systems etc. are more important than some other types of loads. The number
and type of customers likely to be affected will also be an issue. This will potentially affect revenue
and the company’s reputation.
 Duration of the possible interruption: In some circumstances, more critical loads have
alternate sources of supply or are provided with back-up generators and uninterruptible power
supplies (particularly ITC applications). In other circumstances, there is adequate redundancy in
the system to supply the interrupted load, particularly if nearby network assets can be operated at
an emergency rating. However, in some other situations, supply cannot be restored until the
transformer is repaired or replaced. The duration of the outage will also potentially affect revenue,
expenses and the company’s reputation.
 Time to repair or replace the transformer: Repair will not only be a function of the extent of
the damage, but of the availability of spare parts, skilled labour and suitable workshop facilities.
Replacement will be determined by contractual arrangements, transformer suppliers and of
course, if funding is available. The ability to replace a transformer may also be affected by limited
access to the site and transport limitations. Although supply may have been restored to
customers, the reliability of the system will generally be compromised until the transformer is
returned to service.

Transformer assessment and consequence of failure

The condition assessment of a transformer or an entire fleet would not normally consider the possible
consequences of a transformer failure, as these consequences are related to the location and
application of the asset, not to its condition. As stated in Chapter 2, a Transformer Assessment Index
(TAI) would not normally consider, or include, the consequence of potential failures.
However, the consequence of a potential failure should generally be considered in conjunction with
any condition assessment8. If two transformers were in a similar condition, the asset manager would
more likely improve the condition of the transformer that would have more serious consequences if it
were to fail.

8
If a user chooses to use weighting factors when developing a TAI, they may be based on the relative
consequences of failure modes being considered e.g. failure modes related to Oil Impregnated Paper
bushings generally have a higher consequence than some other failure modes, and this would
probably be reflected in the weighting factors chosen. These weighting factors would be applied to all
transformers affected by these failure modes regardless of the location or application of individual
transformers.

53
 Condition assessment of power transformers

How to assess the consequences of a potential failure and utilise these factors to make asset
management decisions is beyond the scope of Working Group A2.49. However, the Working Group
agrees that a Criticality (Impact) Index could be developed to help identify those transformers whose
continued safe operation is most critical.

Assessing criticality & developing a Criticality Index

Assessing criticality can be challenging. For example, most users would assume that a transformer
supplying a hospital would be very critical. However, the hospital may be supplied from a network with
adequate redundancy to carry the load in the event of a transformer failure and the hospital may also
have its own backup generation. Users must therefore consider the capability and redundancy of the
network. To ensure that the criticality of each transformer is assessed accurately and consistently it is
recommended that users develop a scoring matrix prior to any assessment.
Any of the scoring methods used for the development of a TAI (refer to Chapter 2) can be similarly
applied to the development of a Criticality Index. The method selected will be dependent upon the
user’s preferences.
Further details of how to assess criticality or develop a criticality index are beyond the scope of the
Working Group A2.49 and this document.

Economics of Transformer Management

The TAI is a method to identify candidates for consideration, of repair/refurbishment/replacement
and/or scrapping, and as highlighted by this section, there are many other considerations in making
asset management decision. CIGRE TB 248 “Economics on Transformer Management” describes a
methodology that could be used in addition to the TAI, to arrive at a final decision.

54
 Condition assessment of power transformers

Estimating a probability of failure

Transformer assessment indices (TAI) are a useful tool to condense a large amount of technical data
and result analysis into a score that easily conveys information about overall asset conditions, and to
prioritise the selection of assets for further investigation or investment. The subsequent decisions
upon where and when to act are often made including an estimate of the risk associated with the asset
not being able to perform as a result of the estimated condition.
In order to assess the risk, knowledge of two elements are required; the consequence of failure of the
asset and the probability of failure (PoF) of the asset at a given point in time. The consequence of a
failure is generally well understood by Asset Managers who have detailed knowledge of the role of the
asset in the system, the substation design, levels of redundancy and availability of spares, and the
safety, economic and social consequence of a fault or failure. Determining the likelihood of failure
remains difficult to quantify.
Many transformer asset managers follow the generally accepted practice to proactively repair impaired
components once detected, or replace transformers prior to failure, based on the assessed criticality in
the system and the consequence should an in-service failure occur. This practice has prevented the
study and collection of failure data based on condition, and as a result, the actual remaining life is
rarely known.
The TAI alone can be used to derive a risk score that is not a quantified probability but can be
sufficient to identify assets that should be prioritised for risk mitigation. For example, identification of
assets that a pre-determined reasonable allocation of capital should be assigned to, based on a
combination of condition and criticality [13]. However, if a systemic multi-asset investment process is
used, a PoF is likely to be required; the TAI could be used to categorize like assets in the fleet into
groups, and then assign a risk profile or PoF for each group, which can then be analysed further.

Statistical Methods
Commonly used statistical methods and models to analyse failure rates such as Weibull Analysis [14]
are only suited to population studies, while more sophisticated methods (e.g. accelerated failure time
models) could be used to study assets as complex as transformers. Unfortunately, it remains difficult
to determine the probability of an individual unit with a specific condition, due to the difficulty in
gathering data for transformer behaviour and condition. In addition, whilst the overall probability of
failure is a function of all the individual failure modes, linking an individual test result is challenging, as
many condition assessment results may not be attributed to a single root cause or condition.
If commonly used statistical methods are to be applied, the analysis should be limited to similar
transformers i.e. transformers that are of similar designs, component selection (thus having similar
failure modes), with similar operating and environmental conditions [6]. This will derive a failure rate
for a subset of a transformer fleet, but it will not be possible to differentiate between transformers in
the subset. However, transformers are often bespoke designs, so most transformer fleet operators will
not have sufficiently large populations of similar transformers to analyse failure data using commonly
used statistical methods.
Where failure statistics are only available for an entire fleet, it should be noted that transformer design
tools, construction practices and the materials and components have evolved over time. These
nuances and product-specific issues that alter the expected lifetime should be taken into consideration
when considering the likely failure rates for subsets of transformers within the fleet. More sophisticated
methods are thus needed to estimate a particular transformer’s probability of failure.

Estimation techniques
Some failure mechanisms are a function of time (e.g. cumulated thermal, dielectric or mechanical
stresses) whilst others are random with possibly constant or predictable probability in nature (e.g. a
severe transient that instantly creates a dielectric failure). However, as power transformers are
generally repaired or retired prior to failure when the condition assessment indicates a failure is likely
in the near future, the likelihood of failure can be difficult to quantify.
The assessed relative conditions of a fleet of transformers (e.g. from a TAI or PoF) can help identify
the different families of transformers that have a higher probability of failure. This information can be
useful in determining repair and replacement plans, and the number of spare transformers for
reliability and budgeting purposes.

55
 Condition assessment of power transformers

A number of utilities have developed methods to estimate the probability of failure utilising the TAI of
their fleet of transformers and the expected failure rate for their fleet. Note that these methods could
also be used with failure rates derived from other utilities failure statistics or international surveys.
However, users need to ensure that their fleet, their operating conditions and their operating
environment are similar to those used to derive the statistical failure rates that they intend to use.

The conversion of a TAI into an estimated Probability of Failure, will only be able to give the likelihood
of an event, and is not able to distinguish the possible consequences of the various failure modes.
Further work to assess a likelihood of consequence may be required.
However, in the absence of relevant data needed by an appropriate statistical method, an estimated
PoF obtained from available data and a commonly used statistical method may assist the Asset
Manager in making an informed decision.
Two methods to estimate the PoF are included below. Whichever method is used to assign a
probability of failure, the result should correlate with the historic experience of the utility. Any
estimation model should be ‘calibrated’ using actual failure data where possible [15] and updated as
more data becomes available. It should be made clear that the specific failure rate is an estimation,
and a sensitivity analysis on the relevant values should be conducted.
Method 1 – Adjustment of Failure Rate
For large utilities with hundreds of similar types of transformers and a known annual failure rate, a
fleet-wide relative ranking of the transformers by condition can be useful in estimating transformer
failure probabilities.
This method assumes that the fleet has been consistently operated and maintained for several years,
and the asset manager has historical data (with assessed conditions) for the number of in-service
failures, and transformers proactively removed from service. Using this information, a reasonably
accurate prediction of the number of failures in a future time period can be estimated. In other words, if
the fleet has suffered 5-8 failures each year for the past 5 years, and there have been no significant
changes in operational or maintenance practices, it is reasonable to expect the same range of failures
next year.
In the absence of probability data for each condition category, a relative ratio (based on the
experience and judgement of the engineer, as well as considering the construction of the TAI) of the
likelihood of failure between categories must be estimated, based on how the condition categories are
allocated, and the failure modes covered. A Relative PoF (RPoF) can be estimated from this
information.
The Estimated PoF for each condition category can then be approximated by apportioning the
expected failures for the fleet across the condition categories relative to the number of transformers in
that category and the relative probability (RPoF) of that category resulting in a failure. The PoF for the
ith condition category can then be solved by the equation
𝐴𝑛𝑛𝑢𝑎𝑙 𝐹𝑎𝑖𝑙𝑢𝑟𝑒𝑠 × 𝑅𝑃𝑜𝐹𝑖
𝑃𝑜𝐹𝑖 =
∑𝑛𝑘=1 𝐴𝑠𝑠𝑒𝑡𝑠𝑘 𝑅𝑃𝑜𝐹𝑘
Where there are a total of n condition categories in the TAI.
Consider an electric utility that has 1000 medium and large power transformers and a history of
averaging 8 failures in service each year over the past 10 years (annual failure rate of 0.8%). Each of
the transformers in the fleet is assessed, and then assigned to one of five condition categories that are
used in the utility’s TAI. A series of relative probabilities of failure are assumed, based on the condition
thresholds used in constructing the TAI, and the condition of assets at failure (in this example, it is
assumed that the probability of failure rises exponentially as condition deteriorates). Applying the
equation above allows for the estimation of a probability of failure in a given time period based on
condition.

56
 Condition assessment of power transformers

Table 7-1 Estimated Probability Calculation

Estimated Probability of
Condition # of Transformers Relative Probability9
Failure per annum
As-new 120 0.5% 0.1%
Good 320 1% 0.2%
Acceptable 380 3% 0.6%
Poor 150 10% 2%
Very Poor 30 30% 6.1%

In this example, the utility has assumed that the likelihood of a ‘Poor’ transformer failing is 10 times
that of a transformer in ‘Good’ condition. After calibration based on the utility’s experience, the
estimated PoF for a transformer in ‘Poor’ condition is 2% per annum.
The outcome of this method is dependent on the assumed relative probabilities. If the utility decides to
change the relative probabilities, the Estimated Probability will change. Table 7-2 Estimated
Probability using Reduced Relative Probabilities includes an example using the same fleet data, but
with a different set of relative probabilities; in this case, a linear increase in probability based on
condition is applied, rather than an exponential increase used in Table 7-1 Estimated Probability
Calculation. Note that the estimated PoF for transformers in Very Poor condition has decreased from
6.1% to 1.5%
Table 7-2 Estimated Probability using Reduced Relative Probabilities

Estimated Probability of
Condition # of Transformers Relative Probability
Failure per annum
As-new 120 0.5% 0.3%
Good 320 1% 0.6%
Acceptable 380 1.5% 0.9%
Poor 150 2% 1.2%
Very Poor 30 2.5% 1.5%

It can be seen by comparison of Table 7-1 and Table 7-2 that the results can vary significantly
depending upon the initial assumptions as there is an infinite number of solutions for this problem.
Available failure data should be analysed carefully and used to ensure that the initial estimates of PoF
are as realistic as possible. Furthermore, users of estimated PoFs should be aware of the possible
significant inaccuracies in the estimates.
Method 2 – Failure/Replacement Probability versus Apparent Age
Using statistical mathematical analysis tools, it is possible to create a PoF vs. age curve using the
following available data:
 major failures in service, and
 condition-based preventive retirements.
Recent CIGRE reliability studies (e.g. CIGRE TB 642 [16]) indicate that failures are random in nature
and there is no correlation of the PoF with age. However, power transformers are not generally 'run to
failure' as they are generally repaired or retired preventively when the condition assessment indicates
a failure is likely to occur in the near future. When preventive retirements are included in the
calculation, the analysis demonstrates an increase of the PoF with age. Published studies in Canada
[17], and Australia [18] have all concluded that PoF is increasing with age when preventive retirements
are included in the analysis. Including the preventive retirements in the estimation calibration means
that this PoF vs age will mean that the result represents the probability of replacement or failure vs
age. The reader should understand that such a probability does not mean that the failure will actually
happen in service.
However, a probability of failure estimated from the transformers age alone does not take into account
variations of assets of the same age. In order to increase the probability of failure estimation precision,
transformers’ condition is used [17], [19], [20]. The condition of each transformer from a subset of

9
These assumed initial category failure probability values are for example only and not intended to
define any standard values for this approach

57
 Condition assessment of power transformers

transformers of a similar age, would generally be different. It is assumed that transformers in poorer
condition would have a higher PoF than those in a better condition.
Using the results of transformer condition assessments or a TAI, it is possible to compare a subset of
transformers of the same vintage. Transformers that are in a worse condition than the average of the
units of the same age, are assigned a higher “apparent age,” which would be higher that their
nameplate age. Similarly, transformers in better than average condition are assigned an apparent age
which is lower than their nameplate age. A PoF can be estimated for each transformer using the
statistically derived age vs. PoF curve, and the assigned apparent age of the transformer.
Transformers of the same vintage are compared in the second step of this approach, so by
construction it takes into account the technology advances in design tools and materials. It also links
the condition assessment data with the nameplate age inherently including all environmental and
electrical stresses that the transformer may have experienced with time.
Of course, the statistically derived Age vs. PoF curve will be incorrect if retired transformers (assumed
to be approaching end of life) are actually in good condition. Ideally a detailed forensic analysis should
be carried out on retired transformers to ensure that their condition was poor and failure was
imminent, to validate inclusion in the statistical analysis.

Commentary on Operational and External Influences

Users should be mindful of the complexity of calculating or estimating a probability of failure of a given
apparatus and of the potential impact of other actions that can be taken by the network operator on
the probability of failure in a given period. Transformer operators, upon identification of an asset with
poor condition, often apply prudent risk management techniques that influence the probability of
failure. In some cases, transformer operators may apply operating restrictions with the objective of
mitigating the risk. In these cases, management of system events by other measures will alter the
probability of failure of the transformer.
Power systems are complex systems operating in many different conditions across the world. These
different network designs and operating environments can influence failure rate; to make things even
more complex, these external influences may change over time, including rate of exposure to and
magnitude of external events, or change in system fault levels. Whilst not proven, utilities can
experience failures when multiple conditions are met (for example, poor insulation condition combined
with a voltage transient; high winding moisture content subjected to high loading on a hot day). This
should be considered when estimating a probability of failure; the final result will ultimately be
determined by the experience, knowledge and judgment of the transformer engineers.

58
 Condition assessment of power transformers

Transformer active part

The transformer winding (active part) is subject to many mechanical, thermal and electrical stress
cycles during its operating life. Accordingly, this chapter is broken into three distinct parts; thermal,
dielectric and mechanical assessment. Each section focusses on condition assessments that are
indicative of winding degradation due these stresses.

Solid insulation degradation assessment

The main thermal-related failure modes of the transformer active part considered in this section are
cellulose aging, hot spots (winding, and paper-insulated metal) and bubbling.
It should be noted that some of these factors also affect the likelihood of dielectric and mechanical
failures (but in different ways) and are discussed in the next sections. As well, some of these factors
are indirectly impacted by the cooling system (i.e. cellulose aging) but the cooling system is separately
discussed in a later chapter.
Dissolved gases in oil analysis (DGA), oil quality analysis, electrical tests and furans are the main
indicators in the assessment of thermal failure. [21], [22]
Information on the transformer design may also give an indication of the increased risk of thermal-
related failure.
Thermal-related failure modes
The main thermal-relative failure modes are listed below.
Accelerated cellulose aging
This condition has a DGA signature of elevated or high dissolved carbon oxide gasses, Carbon
Monoxide (CO) and Carbon Dioxide (CO2). Other dissolved gasses may also be present like
hydrogen and methane.
Cellulose aging is a complex function of temperature and the presence of oxygen, moisture and
sludge. Cellulose will naturally age since the normal transformer operating temperature is well within
the cellulose thermal decomposition range.
Cellulose aging is normally measured by degree of polymerization (DP) [23] which is essentially a
measurement of the general cellulose hydrocarbon chain length. It is generally accepted by many
transformer specialists that cellulose has a DP value of approximately 1200 when new, an
approximate value of 1000 after the initial factory high temperature drying and approximate value of
100 - 200 when it has reached end of life.
International standards (IEC & IEEE) define the nominal temperature rise values for power
transformers, which should result in transformer cellulose life expectation in the range of 35 to 40
years, depending on the loading and ambient temperatures.
For IEC 60076-2 [24] , the “conventional insulation system” is formed by plain kraft paper immersed in
mineral oil, which is assumed as thermal class 105 and whose nominal life temperature is 98°C. As
the average ambient temperature is assumed to be 20°C, the resultant nominal temperature rise for
the hottest point of the windings, so called “hotspot”, is 78K (98°C - 20°C).
For IEEE the “conventional insulation system” is considered to be thermally upgraded kraft paper
immersed in mineral oil, having a thermal class of 120 and nominal life temperature of 110°C. Since
the considered average annual ambient (highest monthly average) is 30°C, the resultant nominal
temperature rise for the hotspot is 80K (110°C - 30°C).
The average permissible winding temperature rise is directly derived from the hotspot limit. In both
cases, despite considering different solid insulation materials and different ambient temperatures, the
conventional limit for average winding temperature rise is 65K, which is aimed to limit, indirectly, the
hotspot temperature.
The top oil limit is defined based on the oxidation stability of mineral oil. For IEEE, where the
transformer is supposed to have an oil preservation system limiting the contact of the insulation liquid
with the ambient air, the nominal limit for top oil temperature of 105°C. Since the maximum ambient
temperature is 40°C, the temperature rise limit is defined as 65K. For IEC, considering that some
transformers may be of free breathing type where the insulation liquid is permanently in contact with

59
 Condition assessment of power transformers

ambient air, the nominal limit of top oil temperature is defined as 100°C, subtracting the maximum
ambient which is also assumed to be 40°C, the resultant top oil temperature rise limit is 60K.
These two sets of temperature rise limits are assumed to grant the nominal life expectancy for the
transformers:
Table 8-1 - Temperature rise limits

IEEE Standard IEC Standard

Thermal class of solid materials for the conventional 120 105
insulating system
Hotspot nominal temperature rise limit 80°C 78K
Average winding temperature rise limit 65°C 65K
Top oil temperature rise limit 65°C 60K

Assuming an operating condition with a maximum of 0.5% wt/wt of moisture in the solid insulation
system and with constant loading that results in the maximum permissible temperature rise and
constant maximum ambient temperatures, the Arrhenius curve will report an insulation end of life
expectancy after approximately 180,000 hours or 21 years [4]. In practice however, variations in daily
ambient temperatures and transformer loading will result in a much longer insulation life expectancy.
The insulation life expectancy for a typical substation transformer in a moderate climate can be
assumed to be closer to 40 years, while other types of transformers such as generation plant units
with constant loading at or near the nameplate value and in hot climates, or at the other extreme, very
lightly loaded transformers located in cold climates may have insulation life expectancies that are
significantly lower or higher than a typical substation transformer.
Accelerated cellulose aging is then the case where the cellulose is aging at a faster rate than the
considered for the nominal life expectancy, which can be a result of a higher temperature (i.e.
overloading, higher ambient temperature than planned, higher actual winding hot spot temperature, or
higher temperature due to degraded/failed cooling), presence of moisture in the cellulose and
increased acidity of mineral oil.
Transformers are required to be designed with a maximum winding hot spot temperature rise of 80°C
or 78K rise over ambient temperature. Although transformers are tested with a temperature rise type
test, the winding hot spot temperature is not directly measured (unless fibre optic probes are inserted).
If the winding hot spot was not calculated accurately then it is possible that the winding hot spot
temperature rise is significantly greater than 80°C rise and this could lead to localized accelerated
aging of cellulose in that region. Deviations from the temperatures measured by fibre optics are also
frequently reported, since the hottest point of the windings (where the fibres should be located) is a
theoretical concept, whose precise location is difficult to be predict.
When cellulose has reached the end of its life, paper becomes very brittle and is highly susceptible to
failure due to any significant mechanical movement (inrush current, short circuit fault). The aged paper
will also have reduced dielectric withstand, making it more susceptible to dielectric failure due to
overvoltage or transient voltage or, depending on the degradation level, even at nominal voltage
levels. Transformers with a low DP value (less than 200) and high chance of short circuit faults or
dielectric stresses will have the highest risk of failure due to cellulose aging.
Different solid insulation materials and the combination of the cellulose based materials with different
insulating liquids may result in different thermal classes and, consequently, different sets of
temperature rise limits. These values may be obtained from the standards IEEE C57.154 [25] and IEC
60076-14 [26], which both cover transformers using high temperature insulation systems.
Hot metal
This condition has a DGA signature of elevated or high levels of dissolved Methane (CH4), Ethane
(C2H6) and Ethylene (C2H4), the “hot metal gasses.” Dissolved Hydrogen (H2) may also be present
in elevated quantities due to thermal issues but can be caused by other conditions. Elevated or high
levels of carbon oxide gasses may also be present if the hot metal is in contact with cellulose
insulation.
International standards (IEC & IEEE) dictate that metal in transformers should have a maximum
allowed temperature of 140°C [3], at the normal cyclic loading and for conventional insulating system.

60
 Condition assessment of power transformers

If a metal temperature is greater than 140°C, the risk of some gas generation in oil must be
considered.
Most causes of hot metal gas generation are not harmful to the transformer however, the gas
generation can mask other issues (or the hot metal has paper involved which is the next discussed
failure mode) so it is imperative that hot metal gassing issues are investigated and resolved.
Hot metal issues can be due to improperly controlled leakage flux (heating of the clamp structure, tank
or core outer packet), improperly isolated/grounded metal parts, excessive circulating currents, or
overheated tap-changer contacts. All of these issues can have temperatures ranging from 150°C to
700°C and each temperature range will show a differing signature of hydrocarbon gases.
There are cases of “stray gassing oils” which can produce gases in transformers at temperatures from
105°C. This was observed as production of H2, CH4 and C2H6, very often in practice seen as rise of
C2H6 from 50-100 ppm, up to few hundreds of ppm. There is also a special case in moderately
overheated transformer cores that generate small amounts of hydrogen gas at core hot-spot
temperatures as low as 110-120ºC. Higher core hot-spot temperature will have higher rates of
gassing. This special case has been shown to occur in overexcited transformers in high ambient
conditions [27], IEC 60599 Section 4.3 [28] and [29].
Hot metal fault covered by paper
This condition has a DGA signature of hot metal gases (mainly ethylene, methane and ethane) and
carbon oxide gases. If a metal hot spot is covered by paper, this is a far more serious condition and
heavily increases the likelihood of a transformer failure. This could occur in an overheated bushing
connection, lead joint, winding exit lead or in the winding. Each of these conditions could burn up the
paper and cause increased risk of dielectric failure.
Bubbling
Bubbling occurs with the faster increase of temperature of solid insulation in comparison to the oil may
lead to a non-equilibrium of the moisture in the materials eventually resulting in the formation of micro
water bubbles in the windings. This can cause partial discharge activity and eventual dielectric
breakdown of the insulation. The bubble evolution temperature is a function of the oil gas content
(worse if saturated) and moisture content (worse with higher moisture content). Thus, with increased
load, ambient temperature, moisture content and hot spot temperature, the bubble evolution
temperature may be less than the permissible hot spot temperature.
Test and diagnostics
Below are the main tests and transformer information to diagnose the thermal-related failure modes.
Dissolved gas in oil (DGA)
The below is a very limited overview of DGA and other more detailed standards should be referred to
IEEE C57.104 [10] , IEC 60599 [28]. The main gas signatures are:
 Carbon Oxides - As mentioned above, as cellulose ages under normal conditions, it will produce
moderate levels of dissolved Carbon Oxide gasses but if the carbon oxides are very high then this
can indicate excessive cellulose aging. Per IEEE C57.10410 [10], the normal carbon monoxide
and carbon dioxide limits are 350 ppm and 2500 ppm respectively. In addition, the CO2 to CO
ratio should be in the range of 5 – 10 for cellulose aging (lower ratios would indicate paper
involved in a fault). The general levels of dissolved Carbon Oxide gasses in oil are a function of
the relative quantities of cellulose materials such as paper insulation, wood and pressboard used
in the construction of the transformer as well as the quality of oil. If possible, it is best to compare
similar manufacturers and types of transformers to determine what Carbon Oxide gas levels are
normal or elevated.
 Hot Metal Gassing - This is indicated by hydrocarbon gassing (methane, ethylene and ethane)
with ethylene being the key gas. However, methane or ethane might be produced in greater
quantities depending on the temperature of the hot metal. Acetylene is normally not produced in
hot metal gassing as it requires a temperature close to 1000°C.11
 Carbon Oxide gassing + Hot Metal Gassing - This is indicative of a hot metal covered in paper
which as noted above is the most serious thermal risk of failure.

10
IEEE C57.104 is planned for revision to significantly change these carbon oxide limits and ratios.
11
Acetylene can be present in low concentration.

61
 Condition assessment of power transformers

 Core gassing – This has a special gas signature of hydrogen and methane produced in the ratio
of approximately 7 (H2 / CH4) [27].
For different insulating liquids (i.e. ester oils), the standard IEEE C57.155 [30] should be verified,
which may require some adjustments on the values of ppm previously mentioned.
Furans analysis
Furans are a group of chemical compounds that includes 2-Furfuraldehyde and other related
substances that are formed when cellulose degrades. A high furans content or a high production rate
at the same or lower operating temperatures12 may indicate a high rate of paper degradation. The
furans value can be used to infer the Degree of Polymerization (DP) value and hence estimate the
remaining life in the cellulose, provided sufficient time has passed (up to 6 years), following any oil
treatment or replacement. – see CIGRE TB 323 and 494 [22], [21]
Another emerging chemical indicator of cellulose remaining life is methanol, see References [31] [32]
[33]. In the same manner as furans, the methanol is a by-product of paper degradation and its
evolution is correlated with reduction of the paper life expectancy.
Oil quality properties
Oil quality properties that can be used to diagnose the thermal risk of failure are:
 Acidity and Interfacial tension (sludging tendency) – The higher the acid number, the more acid
that is in the oil. Acid will also enter the cellulose and increase the aging of the cellulose. The
reduction of the IFT is caused by the presence of polar contaminants in mineral oil, typically by-
products of oil oxidation. A combination of high acidity and low IFT is critical for mineral oil. For
esters the limits of acidity will be different due to the lower corrosiveness of the longer molecular
weight acids. Due to the higher polarity of ester liquids in comparison to mineral oils, the IFT test
gives no meaningful information, presenting low to no variation along with fluid degradation. In
general, the sludge formation in esters is negligible.
 Oxygen – High oxygen in oil (saturated oxygen in oil is approximately 25,000 ppm) will accelerate
cellulose aging. The paper degradation rate in oxygen rich liquids may be increased in 2 to 5
times. [22], [34]
 For ester liquids, polymerisation can occur, leading to a rise in viscosity. This can be severe when
oxygen is available.
Moisture in cellulose
It has been shown that the rate of paper degradation (aging), will double with every 1% increase in
moisture content in the paper insulation above 0.5% wt./wt., which is the optimum moisture content in
paper [35]. The moisture in cellulose thus affects the aging of the celluloses and also the bubbling
evolution temperature (high moisture lowers the temperature). The moisture in cellulose can be
estimated / measured per below:
 Inferred from moisture in oil – The most common method used to estimate the insulation moisture
level is to calculate the relative moisture saturation value of the oil using the tested moisture PPM
and the measured sample temperature with relative saturation formulae or curves such as the
Oommen curves [36]. The PPM levels of moisture dissolved in the oil can vary over a wide range
due to changes in the moisture equilibrium between cellulose insulation and oil which are affected
by changes in ambient and oil temperatures.

The relative saturation method can provide more consistent results especially when the
transformer is maintained in a constant thermal condition for several days. It should nonetheless
be noted that this method must be used with caution due to the absence of thermodynamic
equilibrium in an operating transformer. Sensors must be carefully located in an area of the
transformers where there is an adequate flow of oil.
 Dielectric Frequency Response – The measured frequency response can be used to estimate the
average moisture content in solid insulation between windings. It can also measure between
windings to ground but accuracy and reliability of estimation depends of type of tested
transformer, number of windings and chosen measuring connection.

12
The partitioning of furans between paper and oil changes with operating temperature.

62
 Condition assessment of power transformers

 Sample blocks – Pressboard samples can be removed from the transformer and the moisture in
cellulose can be measured in a laboratory using one of the established methods.
 Dew Point – Dew point measurement can be used to estimate the moisture in cellulose using
curves of dew point temperature versus ambient temperature. As the method measures the
moisture content of the surface of the materials, it is more representative of the thin insulation
regions such as the conductor insulation.
Insulation Power Factor (PF) (or Dielectric Dissipation Factor (DDF) or tangent delta
The DDF / PF of the major (main) insulation can indicate moisture in the windings. Values below 0.5%
are typically considered normal for mineral oil transformers. However, the most valuable information
comes from the power factor variation which is the comparison of the measured values to the earlier
or initial values, since different transformers may have a characteristically higher insulation power
factor value than compared to typical cases.
Winding DC resistance
High winding resistance can indicate a bad lead joint or OLTC contact resistance variation which could
be causing hot metal gassing. The measured winding resistance should not differ more than about 1%
compared to the factory test report, if the winding temperature at measurement conditions is corrected
to the factory conditions. Difference between phases should usually be less than 2-3%. [5]. The
windings should be tested in all tap positions to detect high resistance joints in tapping leads. Prior to
making a conclusion though, the measurement circuit and the contact resistance for the measurement
devices must be carefully assessed to ensure reliable and comparable results.
Thermal scans of the transformer
Transformer thermal scans can indicate tank, bushing connections or bus duct overheating which
could be the cause of hot metal gassing. Thermal scans can also show if there are cooling issues
which could be causing higher oil temperatures.
Core and core structure grounding check
For transformers equipped with external core and/or core structure grounding, the testing of the
insulation resistance should be verified. The failure mode of this insulation component can be detected
with the measurement of insulation resistance.
Corrosive sulphur
The presence of potentially corrosive sulphur compounds in the transformer oil under appropriate
thermal conditions and winding materials (bare or un-enamelled copper) can lead to the development
of copper sulphide [37] [38]. The presence of corrosive sulphur compounds in the transformer oil can
lead to the decrease of dielectric integrity of the active part, by the formation of metallic sulphides in
operating transformers. Depending on the type of sulphur compounds in the oil, different thermal
conditions and construction metals present, silver, copper, and/or winding paper will be affected by
sulphides deposition.
Design information
Paper type
The cellulose paper type will have a large impact on the rate of cellulose aging. In particular, kraft
paper has a much higher rate of cellulose aging than thermally upgraded cellulose for a given
operating temperature. Thermally upgrading of paper may reduce the ageing rate by a factor of 1.5 to
3. The ageing acceleration from the presence of water is seen to be strongly reduced for upgraded
paper, while ageing with access to oxygen seems to be less influenced. It also appears that thermally
upgraded paper does not produce furanic compounds to the same extent that normal kraft paper
does. In addition, it can be important to note if the windings do not have cellulose paper covering and
hence would not have a cellulose end of life. This could occur if the windings are insulated with
Nomex, netting tape or enamelled bare copper.
Oil preservation system
The type of oil preservation system (sealed or free breathing) has an impact on the DGA interpretation
(gases will dissolve differently) and on the amount of moisture and oxygen in the oil (which affects the
cellulose aging and fluid thermal degradation).
Temperature profile of the transformer

63
 Condition assessment of power transformers

Historical transformer temperature data (top oil or winding temperature) can be correlated with the
instantaneous loading and ambient temperature l to assess the proper thermal performance of the
transformer. This can show if the transformer temperature has exceeded the standard limits which
could then be assumed to have had an effect on the cellulose aging.
Core overheating calculation
For the special case of core overheating causing hydrogen/methane gassing, this risk can be
determined by calculating the core hot spot temperature. This is a function of the core excitation (note
that it may be greater than 110% for generator transformers), number of core cooling ducts, core
geometry and top oil temperature. If the core hot spot temperature is greater than 130°C in certain
conditions, than core gassing can occur. See IEEE Transactions on Power Delivery [27].
The core overheating is characterized by 1 – 3 ppm per day of gas generation (H2 and CH4) with an
H2/CH4 ratio in the 6 – 8 range.
Transformer winding hot spot recalculation
If the transformer design is available, the winding hot spot temperatures can be recalculated using
present day transformer design tools. This can especially show more accurate winding hot spot
temperatures (and higher values) if a 3D Finite Element Method (FEM) calculation of the leakage flux
was not available at the time of the original design. See CIGRE TB 659 “Transformer Thermal
Modelling” [39] for more information.
Temperature rise type test report
This report has winding average temperature rise which is based on many assumptions with indirect
measurements. The hot spot temperature is not measured (unless fibre optic sensors are installed in
the proper windings locations).
Failure mode assessment
Dielectric condition can be assessed using the tests listed above, supplemented by family history,
results of inspections and inductive reasoning as appropriate and in conjunction with the scoring
matrix. The tables in Annex A can also be used as a starting point when assessing a transformer, but
users should consider if the values are suitable for their fleet of transformers, operating conditions,
maintenance practices, and the time scales used in their scoring matrix. User should refer to Chapters
1 to 7 before assessing their transformers.
Below is a summary of the information recommended for a basic fleet assessment for the transformer
active part - solid insulation degradation.
 Loading profile and ambient temperature registers
 Dissolved gas in oil test
 Oil quality results including moisture and acidity
 Estimated or measured moisture in cellulose
 Furans analysis
 The DDF / PF of the major (main) insulation
 Measured DC winding resistance
 Corrosive Sulphur in oil test
Below are optional items that can be included if possible:
 Calculated winding hot spot temperature
 Calculated bubbling evolution temperature

64
 Condition assessment of power transformers

Scoring Matrix for thermal Assessment

The generic scoring matrix illustrated in 2.3 has been adapted to more easily allow the assessment of
thermal related failures.

Scoring matrix for thermal assessment

F Transformer should be removed from service due to significant and serious test results
indicating actual breakdown of insulation.

E Indications show end of life aged insulation with multiple tests showing very high risk of failure

D Indications show significantly aged insulation with additional test results showing high risk of
near term failure

C Indications show aged insulation with increased risk of dielectric failure

B Indicators show small deviation from new, however follow normal degradation behaviour.
Normal operation is expected for long term.

A Normal - All test results normal: no trace of acetylene; normal levels of other gases and no
indication of problems from electrical tests.

Dielectric assessment
The transformer active part dielectric risk of failure is based upon the partial or complete breakdown of
the insulating material of both the solid and liquid insulation. A transformer’s primary function is to
change the voltage from one circuit to another circuit. The transformer will have windings and
components at different voltages including ground voltage. This results in dielectric stresses which
must be controlled using insulation materials. When the dielectric system fails there will be partial
discharges and faults. Partial discharge is defined as a localized electrical discharge that only partially
bridges the insulation between conductors and which can or cannot occur adjacent to a conductor
[40]. A fault occurs when the insulation completely breaks down and the voltage between the
electrodes collapses to zero and fault current flows until the transformer is disconnected from the
system.
Insulation systems in power transformers can be broken into four categories:
Major insulation – This is the insulation between windings and from windings to ground.
Minor insulation – This is the insulation within a winding.
Core insulation – This is the insulation within the core and insulation from the core to ground.
Insulating liquid – This is the dielectric quality of the insulating liquid. Refer to the oil chapter for more
information on dielectric quality measurements.
Dielectric failure modes
The main dielectric failure modes are listed below:
Failure of minor insulation
Failure of the minor insulation is likely to cause serious damage to the winding and is likely to require
transformer replacement.
 Turn to turn failure in winding
The individual turns in a transformer winding will be at a different power frequency voltage than the
adjacent turns based on the voltage rating of the winding and the number of turns. Under impulse
conditions, the turn to turn voltage will be based on the voltage distribution in the windings. The
impulse voltage distribution can be improved by the use of static rings at the winding ends, winding
geometry and special winding designs such as interleaving and floating conductors which increases
the series capacitance of the winding. A turn to turn insulation failure will cause short circuit current in

65
 Condition assessment of power transformers

that winding location and the whole winding will be destroyed by either electromagnetic forces or by
consequential insulation breakdowns due to gas generation in the failure location. A turn to turn
insulation failure will also change the voltage distribution in the remainder of the winding often leading
to complete breakdown of the insulation between adjacent windings or between the winding to ground
potential.
 Disc to disc failure in winding
Disc to disc failure in a winding is typically found in the high voltage windings where groups of turns
are wound together as a disc and each disc is separated by insulation spacers to provide dielectric
clearance as well as space for oil flow for cooling purposes. Both ends of the disc will have a
crossover where the conductor is brought from one disc to the next to continue the winding. There is
very little disc to disc stress at this point, however at the opposite end of the disc there will be a high
voltage potential between adjacent discs which could lead to partial or complete breakdown.
Failure of major insulation
Failure of the major insulation or tracking/partial discharge is likely to require transformer replacement
or major work.
 Inside winding to core
Normally the inner winding of a transformer wound with concentric winding construction is the lower
voltage winding. This minimizes the insulation space required to isolate the inner winding from the
core leg which is at ground potential. However, there is a potential difference between the inner
winding and the core which can cause the insulation to break down partially or completely. This failure
mode often follows a through fault event in which the inner winding buckles at some point reducing the
clearance to the core leg.
 Winding to winding (high to low) failure (major insulation)
Windings in transformers are at different voltages. The voltage stress between windings is controlled
by a combination of solid insulation barriers and oil ducts created by insulation spacers. The dielectric
stress at the ends of the windings can be higher due to the stress enhancement at the edges of the
windings. This is often controlled by using contoured insulation material. The winding to winding
insulation system must be designed for the voltage stresses found during normal operation at line
frequency and under impulse and switching surge conditions. Winding to winding insulation can
partially breakdown (partial discharges) or completely breakdown.
 Winding to top yoke failure (major insulation)
Transformer windings must be insulated from the grounded core material. Many windings have the
highest potential at one end of the windings which results in a voltage stress between the end of the
winding and the core and its associated clamping structure. Just as for the winding to winding space,
this insulation system is often made from solid insulation barriers separated by insulation spacers.
Typical failure modes are creep discharges along the surface of the barriers and spacers or oil
breakdown between the barriers. These result in partial or complete breakdown of the windings to
ground.
 Winding lead to ground failure (major insulation)
The winding lead will be at the highest voltage. This lead needs to be insulated dielectrically as well as
supported mechanically. Often for extra high voltage transformers, a series of concentric insulation
barriers separated by spacers is utilized as the “lead exit” insulation system. As in the winding to
winding and winding to ground insulation systems, there will be dielectric stress both along the surface
of the insulation (creep) and in the oil space between the barriers which can either partially breakdown
or breakdown completely.
 Phase to phase failure (major insulation)
For three phase transformers there will be a potential difference between the winding of one phase
and the winding of an adjacent phase. This will create a dielectric stress that must be accounted for.
Typically, a solid insulation barrier is placed between phases. For higher voltage transformers, a
system of barriers and spacers is utilized. As with other insulation systems, there will be creep and oil
gap stresses. Creep stresses are particularly high at the ends of the windings.

66
 Condition assessment of power transformers

 Winding to tank failure (major insulation)

For both single phase and three phase transformers there will be a potential difference between the
outmost winding and the grounded tank wall. This will create a dielectric stress that must be
accounted for. Depending on the voltage of the outer winding and the design, a solid insulation barrier
is placed between the winding and the tank. For higher voltage transformers, a system of barriers and
spacers is utilized. As with other insulation systems, there will be creep and oil gap stresses. Creep
stresses are particularly high at the ends of the windings.
Failure of core insulation
Failure of or arcing/sparking/partial discharge in the core insulation may require transformer
replacement, major work or corrective maintenance depending on the nature of the fault.
 Core to clamp insulation
In order to minimize circulating currents within the core, the core is typically grounded at only one
single point. An insulation material is placed between the steel clamping material and the core.
Although the potential difference is small in comparison to the major insulation spaces (high to low and
to ground), a dielectric stress does exist which can breakdown partially or completely. Complete
breakdown of core clamp insulation normally does not cause a transformer to trip offline however it
can lead to carbon generation (which can lower oil and winding insulation dielectric strength) and high
rates of gas generation. Circulating currents can burn the core lamination insulation leading to an
increase in circulating currents and eventually can lead to complete breakdown due to the high
generation of gasses and subsequent lowering of the strength of the insulation system.
 Core lamination insulation failure
Each lamination in the core has insulation material on the surface which provides electrical insulation
to adjacent laminations. This limits the circulating currents within the core. This insulation material can
breakdown which increases the circulating current resulting in higher losses. Failure of the core
lamination material does not result in a transformer fault except in extreme cases (see failure mode -
core to clamp insulation) but will increase core losses and gasses.
Failure of liquid insulation
The liquid insulation in a power transformer must withstand the dielectric stresses subjected to it
during normal and abnormal operation. When the dielectric strength of the liquid becomes too low due
to moisture, contamination or aging, it can either partially or completely breakdown. It is important to
test the liquid insulation to assess its condition as well as to identify any gasses that could have been
produced by partial or complete breakdown. The liquid should also be tested for stray gassing, as the
increased production of gasses can interfere with transformer diagnostics. The one advantage of the
liquid insulation system is that it can be processed or replaced. Thus poor liquid insulation quality is
not included in this section and instead it is considered in Oil Quality in Chapter 10.
Both IEC and IEEE have transformer oil maintenance guides which give recommendations for oil
reprocessing or replacement based on analytical tests of oil removed from the transformer [41] [42].
Test and diagnostics
Dissolved Gas Analysis (DGA)
DGA is an important indicator of dielectric breakdown failure (considered essential). The main gas
signatures are:
Hydrogen- Hydrogen is the key gas for partial discharges in oil or gas. Typically there will be methane
gas present as well.
Carbon Oxides – Dielectric breakdown inside the solid cellulose insulation material will generate
carbon monoxide and carbon dioxide.
Sparking Gasses: Partial discharges of the sparking type (IEC 60599 D1 [43]) will create hydrocarbon
gasses
Arcing Gasses: Complete breakdown of the insulation material, particularly from winding to winding or
winding to ground will result in an electrical arc with high levels of acetylene as well as other
hydrocarbon gasses and hydrogen.

67
 Condition assessment of power transformers

Oil physical properties:

Dielectric Strength –The dielectric strength of the oil needs to be sufficient to maintain the proper
insulation strength of the oil/paper system [44].
Dissipation factor –The oil should be tested at 25 and 100 degrees centigrade to ensure that it has no
contaminates and will provide sufficient insulation strength [45].
Moisture Content – The insulating performance of the oil is strongly influenced by the moisture levels
[46].
Dielectric dissipation (power) factor The dielectric dissipation (power) factor of the main insulation
between windings and windings to ground can indicate moisture and/or contamination in the winding
to winding or winding to ground insulation system. Measured values below 0.5% are considered
normal however the measured values should be compared to earlier measured values as some
transformers can have a characteristically high insulation power factor value. Generally, measured
values that are greater than 0.5% are considered deteriorated insulation and values greater than 1%
are considered very poor. (Section similar to above, Insulation power factor (or Dissipation Power
Factor or Tangent Delta)
Excitation current
Excitation current is a measure of the core characteristics but can also be influenced by shorted or
open turns. If the excitation current is less than 50 mA, the difference between the two higher currents
for a three phase transformer should be less than 10%. If the excitation current is greater than 50 mA,
then the difference should be less than 5%.
Turns ratio
Turns ratio is a measure of the effective voltage transformation from one winding to the other. The
turns ratio should be within 0.5% of the calculated turns ratio based on the nameplate [47]. Turns ratio
problems can be an indicator of shorted or open turns.
Dielectric frequency response (DFR)
DFR is similar to an insulation dissipation factor test except that it is measured over a wide frequency
range. The purpose of this test is to estimate the moisture content in the solid paper insulation. The
conductivity of the oil as well as other contaminants in the insulation system will also influence the
results.
Insulation resistance
Insulation resistance is a DC test sensitive to high moisture or contamination in the solid insulation.
The DC insulation test is also used to confirm that there is only one core ground (the intentional
ground is disconnected for the test and a very high insulation resistance should be measured between
the core or clamp and ground). The polarization index is determined by the ratio of the resistance after
ten minutes to the resistance after one minute. This ratio can also indicate moisture and/or
contamination.
Partial discharge
When the dielectric system fails there will be partial discharges and faults. Partial discharge is defined
as a localized electrical discharge that only partially bridges the insulation between conductors.
Because partial discharges are caused by dielectric stress they need to be measured when at a
minimum, operating voltage is present. This test is done off-line by applying voltage to the transformer
by an external power supply or on-line at operating voltage.
Frequency response of stray losses (FRSL)
The leakage reactance is measured over a frequency range of 400 to 15 Hertz. The plot of leakage
reactance to frequency is compared for all phases. If one phase is different, this can be caused by
shorted strands in a multi-strand winding.
Bubble evolution temperature
The temperature at which a bubble in the oil can evolve is calculated with curves against moisture in
cellulose (See C57.91.2011 [3] and CIGRE TB 349 [35]). Essentially high moisture in cellulose will
give lower temperatures at which a bubble may form which effectively limits the overload capability of
the transformer. If this bubble evolution temperature is exceeded by a higher hot spot temperature in

68
 Condition assessment of power transformers

the transformer, then the transformer has a higher risk of dielectric failure caused by the bubble
evolution.

Failure mode assessment

Dielectric condition can be assessed using the tests listed above, supplemented by family history,
results of inspections and inductive reasoning as appropriate and in conjunction with the scoring
matrix. The tables in Annex A can also be used as a starting point when assessing a transformer, but
users should consider if the values are suitable for their fleet of transformers, operating conditions,
maintenance practices, and the time scales used in their scoring matrix. User should refer to Chapters
1 to 7 before assessing their transformers.
Scoring Matrix for thermal Assessment
The generic scoring matrix illustrated in Section 2.3 has been adapted to more easily allow the
assessment of dielectric related failures.

Scoring matrix for dielectric assessment

F Severe partial discharge fault in main tank, likely to lead to imminent failure. Electrical tests
indicating actual breakdown of insulation.

E Indications show arcing/sparking/tracking or partial discharge fault of the major insulation is

getting worse and indicating failure on short term.

D Indications show steady arcing/sparking/tracking or partial discharge fault but not getting
worse with time, indicating potential failure on medium term.

C Indications show dormant or intermittent arcing/sparking/tracking or partial discharge fault e.g.

Core insulation deterioration

B Indicators show small deviation from new, however follow normal degradation behaviour.
Normal operation is expected for long term.

A All test results normal: no trace of acetylene; normal levels of other gases and no indication of
problems from electrical tests.

Mechanical assessment
A mechanical assessment of the transformer active part considers the structural integrity of the core
and coil assembly, its alignment and position within the tank, and its ability to withstand the
electromagnetic forces exerted within that structure when the transformer endures a severe
overcurrent event. A mechanical assessment of the active part should not overlook the following
componentry where present:
 Leads and connections to terminations and tapchangers (including electrostatic shields, shrouding
barriers, supporting cleat work, barrier boards)
 Auxiliary equipment such as voltage transformers, current transformers, fault limiting reactors,
preventive autotransformers
 Internal surge arrestors (Non-linear resistors / ‘metrosils’)
 Flux equalising windings on yokes or non-wound limbs
 Flux collectors and flux shunts
 Devices used for supporting, clamping or anchoring the internal assembly within the tank
Before making a mechanical assessment of a power transformer, it is necessary to know whether it is
core-form or shell-form construction. The inherent high short circuit withstand capability of shell-form

69
 Condition assessment of power transformers

construction together with its enveloping, rectangular core build (secured or clamped within the
transformer tank structure) means that a mechanical assessment for shell-form is less complex than
for core-form construction.
In comparison with core-form, shell-form experiences moderate short circuit stress because:
i. the phase assembly has more coil groups and, being arranged side by side, they oppose each
other under short circuit, keeping the phase assembly in compression,
ii. the short circuit forces are distributed over the very large area of each pancake winding, and
iii. the massive shell-form core and its integral fit with the tank structure, easily supports the
phase assemblies
The mechanical assessment of shell-form transformers is normally confined to performing electrical
testing to discern any disturbance to the geometry of the pancake windings and performing an internal
inspection to check for looseness of wedges between each phase assembly and the core and
observing mechanical damage to tapping leads and line lead exits of the phase assemblies.
Even with the mechanical assessment of core-form construction, it is helpful to know whether the
winding design has multiple layer windings or disc windings or helical windings. Each winding type has
special considerations for its short circuit withstand capability and each have particular failure modes.
It is recommended to use transformer design reviews to learn these aspects whenever a new
transformer design is procured. Later in the service life, these understandings assist in assessing the
likely mechanical condition, especially when the transformer has been subjected to short circuit events
of known severity and duration. There are other severe event conditions (such as synchronising
errors) that could also be considered.
In order to make a mechanical assessment of core-form construction, the function and importance of
clamping the transformer windings needs to be understood. Coil clamping is an important attribute of
the mechanical strength of the active part. Applying the correct coil clamping pressure allows for:
i. the complete internal assembly (core and coils) to be lifted by a crane,
ii. the bracing of the transformer windings during transportation,
iii. the reduction of the operating ‘load noise’ level of the windings and
iv. most importantly, assists the transformer to survive short-circuit events.

Importance of coil clamping for the short circuit withstand of a transformer

During a short-circuit event on a transformer, the windings are effectively in unstable equilibrium. The
coil clamping helps to maintain a near equilibrium (or minima of the short-circuit forces). Retaining the
relative position of the windings, stabilizing their length and maintaining the residual clamping force is
very important for short-circuit withstand of the transformer. Under a short-circuit fault, the windings
repel and exert large axial and radial forces that effectively try to separate the windings. The size of
the coil clamping force is designed to be commensurate with the expected magnitude of axial forces
exerted during the short-circuit event. This keeps the windings in compression and therefore stable.
However, any alteration to a winding’s axial length, or its relative axial position through dislocation or
collapse of the winding conductors, is likely to loosen the coil clamping. Furthermore, these changes
also alter the winding ampere-turn distribution. Any change in the relative position of the ampere-turns
will alter their fine positional balance and symmetry (between the winding groups) and that invariably
results in substantial increases in the short-circuit forces. For these reasons, short-circuit withstand
capability of a transformer is dependent on the status of its coil clamping and therefore coil clamping
should not be compromised [48].
Relaxation of coil clamping
Coil clamping will be compromised if the residual clamping pressure has relaxed enough to then allow
dislocation or mechanical deformation of the windings during a short-circuit event. Coil clamping may
also be compromised either by the effects of repetitive short-circuit events or a single protracted short-
circuit event. This can be due to the cumulative plastic deformation of the axial insulation components,
or to a partial winding collapse or to conductor tilting. However, the most common way that the coil
clamping relaxes is with shrinkage of transformer cellulosic insulation through ageing.
Paper shrinkage allows gradual relaxation of the residual clamping pressure. Despite some limited
elasticity or recovery to expand again, the structures can still become loose. This can occur either
during the vigorous oscillations of a short-circuit event or by eventually vibrating out after clamping
pressure is lost. Not all of the insulation structure is necessarily keyed, glued, tied or dowelled to

70
 Condition assessment of power transformers

prevent this dislocation. Furthermore, repetitive short-circuit events can cause cumulative bruising and
dislocation of components in the insulation structure. Should an axial or radial spacer or support block
be displaced due to these cumulative effects, or whether it simply vibrates loose (having lost its
clamping pressure), then the void this creates provides room for winding dislocation during short-
circuit events, resulting in even greater short-circuit forces to be exerted and further winding damage.
Transportation damage to the active part
Transport damage ought to be detected prior to transformer commissioning, however mechanical
damage can sometimes go undetected. There are two aspects to the mechanical failure of the active
part during transport. Firstly, the active part is typically locked, pinned or braced within the tank
enclosure for securing it during transport. Of necessity, these points of contact with the tank employ
solid insulation for electrical isolation, and these brittle materials are potentially subject to large
mechanical forces during transport. Should the brittle insulation material shatter or split, then the entire
active part could dislocate within the tank. Secondly, even if still secured by the fore-mentioned
locating points, the core and coil assembly itself may distort, partially collapse or eject insulation
blocks because of shocks, vibration or swaying motion during transport. Functional mechanical design
of the active part usually focuses on withstand to short circuit through events and the ability to lift the
completed assembly by crane. For that the phase assembly relies on its retention by axial
compression, the radial support back to the core, and the hoop stress withstand of the conductors.
Subjecting the active part to sudden shock, vibration or swaying may introduce very unusual
mechanical stress to the structure. The core and coil assembly design may not properly address this
transport stress, especially where the inertia of the heavy mass of the winding conductor may readily
disturb the end insulation structure that is comprised of paper based collars, washers and block work
insulation. Also, shipping transport in very heavy seas can conceivably establish inertial swaying of the
massive core limbs because of the inherent ease for flexing of the laminations. Cumulative dislocation
of the insulation, core limb and winding structure may then occur with the repetitive rolling motion
during sea transport and this may not even register as a shock loading. Refer to CIGRE Technical
Brochure 673 [49] for more information on transport damage.
Mechanical damage to windings and insulation structure
Short circuit events external to the transformer create severe overcurrent’s that will interact with the
magnetic leakage field permeating the winding structures, thus exerting electromagnetic forces on all
those winding turns that are carrying the overcurrent.
In transformers of core-form construction with two concentric windings, the ampere turns are usually
uniformly distributed and balanced, resulting in a magnetic leakage field that is predominantly axial
over the greater part of the winding height. During external short circuit events, the axial leakage field
tends to reduce the diameter of the inner winding (causing free buckling or forced buckling) by radial
compression, and it tends to increase the diameter of the outer winding (causing hoop tension)
through radial expansion. At the winding ends, the magnetic leakage field diverges (or fringes)
creating a significant radial component in the leakage field that tends to compress the windings axially.
Refer to Figure 8-1.

Figure 8-1 Electromagnetic force

71
 Condition assessment of power transformers

The electromagnetic forces are oscillatory and act instantaneously. Because of their dependence on
the square of the current, electromagnetic forces are unidirectional.
Because the windings are structurally elastic in the axial direction, they act as a tall stack of massive
discs, each restrained by axially oriented damping springs (the insulation spacers, conductor paper
insulation and the end insulation structural buffers). These springs contract when the peak axial force
is attained and relax as the force cyclically drops to zero.
These endthrusts can severely damage the windings themselves and their end insulation structures.
The greater the relative axial displacement, the greater is the magnitude of the end thrust because the
radial leakage field is further concentrated at each winding end. Maintaining the symmetrical
positioning of the winding groups is therefore imperative for minimising the end thrust forces [50].
Accordingly, the typical mechanical failure modes in windings caused by through faults can be
classified as either radial failure modes or axial failure modes.
Radial failure modes
Hoop tension
Outward radial force causes stretching of outer conductors that produces excessive hoop tension in
the conductor - this may break the conductor insulation or lead to axial collapse due to local bulging of
conductors beyond the outer radial edge of their spacer supports.
Forced buckling
Inward radial force causes collapse of the inner winding due to beam failure which can break the
conductor insulation or crush the axial duct insulation structure on the inside diameter of the winding.
The inward radial force exerts compressive stress. If that stress exceeds the elastic limit of the
conductor material, the conductors bend radially inwards between the axial strips supporting the inside
diameter of the winding, producing a shape resembling a multi-pointed star Figure 8-2.
Free buckling
This is another collapse pattern caused by inward radial force being exerted on the inner winding. The
inward radial force exerts circumferential compressive stress which causes the conductors to bulge
inwards as well as outwards at one or a few specific points around the circumference Figure 8-3. This
is essentially an unsupported buckling mode and the deformation extends along most of the winding
length. The wavelength of the buckle does not correlate with the spacing of the axial sticks supporting
the radial spacers. Free buckling occurs when the conductor strands possess very high rigidity such
that they must buckle first before any appreciable force is exerted inwards onto structures that might
support the inside diameter.

Figure 8-2 Forced buckling distortion of coil circle Figure 8-3 Free buckling distortion of coil circle

Axial failure modes

Tilting
Axial forces cause collapse of the winding due to conductor tilting. It is a failure mode mainly seen in
disc windings where the whole set of conductors within the radial depth of a disc all tilt over in the
same direction whilst the conductors in the adjacent disc (above or below) all tilt in the opposite
direction. This results in the classic zig-zag pattern observed in this failure mode Figure 8-4. This

72
 Condition assessment of power transformers

failure mode causes the conductor insulation to break, loss of clamping pressure, winding structure
instability and even greater forces as a result of mmf imbalance and positional offsets when the
conductors significantly dislocate. Thin conductors are more susceptible to tilting.

Beam failure
Axial forces exerted on each winding conductor act like a distributed load on a beam (the conductor)
that is supported between the columns of winding spacers Figure 8-5. The conductor beams deflect
and flex as in classic bending moment stress mechanics. Beam failure can break the conductor paper
insulation.

Figure 8-4 Conductor tilting Figure 8-5 Beam failure of

conductor span between winding spacers

Telescoping
A failure mode sometimes seen in narrow windings, especially winding radial groups used for layer
windings, tertiary windings or tapping windings. This failure mode occurs when axial force drives some
of the winding turns axially past adjacent turns. Telescoping of the winding is likely when narrow
windings have radial looseness or inadequate axial support.
End insulation collapse
A failure mode where the hammering of a winding radial group into the top or bottom end insulation
causes the winding end supports to collapse. The collapse occurs due to either over-stressed
materials, misalignments or instability within the support structure (e.g. columns of spacers or block
stacks), which can result in gross axial deformation of the winding, or magnified end thrust forces. The
electrostatic shields located at the coil ends of HV windings can be impacted and damaged, especially
the braid, strap or pigtail tie to the line lead. Lead exits, often with their complex insulation structures
(e.g. snouts) will also be susceptible to mechanical damage by end thrust forces, let alone the
electromagnetic forces on the line leads themselves complicating the situation. Relaxation of the
residual axial clamping force may also allow excessive vibration, even under normal service
conditions. Vibration, or alternatively the impacts of repetitive through fault events, will cause
cumulative bruising of the axial support, gradually exacerbating the degree of collapse. Accordingly, a
transformer subjected to a series of ‘soft’ repetitive through faults can ultimately fail without ever
having seen an external short circuit event with full fault current (where the transformer impedance
predominantly limits the current).
Fretting of conductor insulation
A failure mode where mechanical deterioration of conductor insulation may result from impact forces
(hammering effect) or friction resulting from the relative movement between paper covered conductors
and solid insulation or from deflection and bending of unsupported spans (causing permanent set
deflection and plastic deformation of the conductors). The conductor insulation is sometimes
reinforced by manufacturers where inter-disc crossovers or the transposition of paralleled conductors
creates localised pressure points between conductors during external short circuits or repetitive
through faults, yet these areas may remain susceptible. Misalignment of axial spacers (spacer lanes
skewed or haphazard alignment) may also exert excessive stress on conductors and directed oil flow
washers. This failure mode is extremely difficult to observe or detect and therefore usually impossible
to assess due to the windings being inaccessible.

73
 Condition assessment of power transformers

Spiralling of helical windings

A failure mode mostly seen in helical windings occurs when the end turns of the helical winding rotate
away from their correct position. This failure mode occurs when axial force is applied on the end turns
causing axial dislocation of the turns. Axial stretching of a helical turn tries to reduce its mean turn
diameter, rather than elongate the conductor. However, the resultant effect on a helical winding having
solid radial support will instead be to fore-shorten the turn, appearing as resultant rotational movement
(or unscrewing). This failure mode may sometimes result in mechanical damage to top and bottom
lead exits on helical windings due to this rotational dislocation and stretching of the lead connections.
The greater the number of helical turns displaced axially, the greater is the cumulative rotational
displacement.
Below are the main tests and transformer information to diagnose the mechanical risk of failure.
Tests and diagnostics
Frequency response analysis
Frequency response analysis (FRA) is an important test because it is sensitive to whole winding group
dislocations, telescoping of windings, as well as some localized winding dislocations or movement,
and is generally used to detect geometrical changes and electrical short-circuits in the windings
caused by external short-circuit events. Provided sufficient movement of winding conductor occurs to
alter the frequency response of the resistance, inductance and capacitance (RLC) network parameters
of a winding, then FRA may be able to discern the presence of failure modes such as severe transport
damage, buckling, hoop stress distension, telescoping, spiralling, and end insulation collapse caused
by winding movement.
Interwinding capacitance
Interwinding capacitance can be sensitive to major distortion in the radial separation distance between
a pair of coaxial cylindrical windings. Changes in the winding length or vertical position of one winding
may also change the interwinding capacitance. Dimensional changes of a few percent are required to
detect a recognizable change of interwinding capacitance, especially in three phase transformers
where the capacitance of one phase contributes only a third of the total measurement. This test may
detect the presence of severe forms of the buckling, hoop stress distension, telescoping and winding
end thrust collapse.
Leakage reactance
Single phase leakage reactance can be sensitive to major displacement of the ampere turn positions
of the windings. This is because the leakage field has been distorted and the ampere turn positional
balance altered. The test is performed on each phase to facilitate comparisons between the phase
assemblies.
Dimensional changes of a few percent are required to detect a recognizable change of the leakage
reactance. This test may detect the presence of severe forms of transport damage, buckling,
telescoping and end insulation collapse caused by winding movement.
Methods used to survey residual clamping pressure
An inspection of the internal assembly makes the following survey possible:
i. Alignment of the phase assembly heights
ii. Measurement of phase assembly height to a common datum
iii. Checks for spacers, blocks and packing shims fallen out of position
iv. Checks of alignment of spacers, block and shims
v. Observation of relative winding movement or distortion of shape
vi. Hand looseness of coil clamping devices
vii. Hand looseness of any block work in the end insulation structure
viii. Tonal differences when accessible block work is lightly tapped
ix. Ease to dislocate accessible block work when again lightly tapped
x. Measurement of residual clamping pressure (where this is practicable)
However, the extent of the coil clamping survey performed on a transformer may be limited by the
available ‘hands-on’ access to the internal assembly. Access may be restricted by having only a few
access covers, or by the limited reach of confined space access or the ability to remove the

74
 Condition assessment of power transformers

transformer’s bell cover or lid. For these reasons, such surveys are usually performed as a part of
other major work on a transformer. Note that the required skills, resources and expertise increases in
complexity down this list. This survey may detect the presence of transport damage, relaxation of
clamping pressure, dislocated end insulation structures, hoop stress distension on an outside winding
and severe forms of end insulation collapse caused by winding movement.
Measurement of the residual clamping pressure
The measurement of residual clamping pressure is not always practicable. However, any coil clamping
arrangement that is adjustable, inherently had to have some way of measuring the force, effort, torque
or applied pressure that was originally used to set it. There are two common methods.
 Torque wrench method
A common practice for pressure screw arrangements is to the record the torque wrench loading at
each pressure screw position and tabulate (or map) the load distribution on all the phase assemblies
and compare them. Note that it is usual to measure the torque required to infinitesimally increase the
existing load at each pressure screw (rather than the torque required to release it). The disadvantage
with this method is that any adjustment at one single point may slightly influence the loading at the
other pressure screw positions around the coil circle.
 Hydraulic jack method
A preferred and almost universal method to measure residual clamping pressure is to use a minimum
of four hydraulic jacks (cylinders) to bear evenly down on the top pressure ring of the phase assembly.
The main limitations are whether there is enough room to position the jacks and whether suitable
pressure points can be found to extend each jack between the top yoke clamp and the winding
pressure ring. By gradually and equally increasing hydraulic pressure on the four jacks, it is possible to
record, one by one, how much force is required to overtake the existing clamping force at each
clamping point. Hydraulic cylinders provide ease of applying force and a suitably slow rate of
increasing the force. They also provide a means of quick release of force, if required for safety.
Positioning the four hydraulic cylinders approximately 90º apart around the coil circle ensures that the
application of force is simultaneous and evenly distributed around the coil circle [48].
Dissolved Gas Analysis
In coil clamping arrangements that employ pressure screws, it was common to use metal pads
under each screw for distributing pressure onto the phase assembly. If the coil clamping relaxes
enough to let loose the metal pads, then sparking of these floating metal objects will occur. This is
due to the voltage they assume through capacitive coupling with the winding voltages below. The
sparking will produce dissolved hydrogen and acetylene in the oil.
Below is a summary of the information and test results that would be recommended for a basic fleet
assessment for the transformer active part mechanical risk of failure:
1. Frequency Response Analysis (FRA) Test. Ideally, a set of initial FRA signatures should be
recorded at transformer commissioning for making later comparisons.
2. Single phase leakage reactance
3. Inter-winding capacitance (usually already obtained from dissipation/power factor tests)
Transformers of high capital value or ones that would pose severe consequences in the event of their
failure may be considered worthwhile candidates for performing an internal inspection to ascertain a
proper mechanical assessment of their active part. This would especially be considered following any
severe external short circuit event.
Failure mode assessment
Mechanical condition can be assessed using the tests listed above, supplemented by family history,
results of inspections and inductive reasoning as appropriate and in conjunction with the scoring
matrix. The tables in Annex A can also be used as a starting point when assessing a transformer, but
users should consider if the values are suitable for their fleet of transformers, operating conditions,
maintenance practices, and the time scales used in their scoring matrix. User should refer to Chapters
1 to 7 before assessing their transformers.

75
 Condition assessment of power transformers

Scoring Matrix for Mechanical Assessment

The generic scoring matrix illustrated in 2.3 has been adapted to more easily allow the assessment of
mechanical failure.

Table 8-2 Scoring matrix for mechanical assessment

Scoring matrix for mechanical assessment

F Severe mechanical damage in main tank, likely to lead to imminent failure. Electrical tests
indicating obvious changes in leakage reactance, interwinding capacitance and FRA
signatures. Indications of a dielectric failure may also appear

E Indications of mechanical alteration in the windings because of some changes in leakage

reactance, interwinding capacitance or FRA signatures. Visual evidence of dislodged blocks,
and some winding movement may also be present. No coil clamping pressure on parts of a
phase assembly

D Not applicable to mechanical assessment

C Not applicable to mechanical assessment

B Indicators show small deviation from new, however follow normal degradation behaviour.
Normal operation is expected for long term. Through fault(s) have occurred but no evidence of
damage or change in electrical tests. Relaxation of coil clamping is evident

A All test results normal: no trace of acetylene; normal levels of other gases and no indication of
problems from electrical tests. No visible disturbance of internal assembly structures

76
 Condition assessment of power transformers

Bushings and cable boxes

Bushings and cable boxes are critical components of a power transformer, as they are the principal
interface item between the transformer winding and the rest of the substation or electrical system.
Bushings are subjected to electrical, thermal and mechanical stresses, and are a leading cause of
failure for transformers [51]. Therefore, knowledge of bushing condition is crucial in any
comprehensive assessment. Bushings assessed to be in poor condition do not necessarily result in
transformer replacement, as an economical assessment may show that bushing replacement is
sufficient.

Transformer bushings
Note: The Technical Brochure (TB) by CIGRE Working Group A2.43, titled ‘Transformer Bushing
Reliability’ [52] to be published in 2019, provides a more detailed description of some of the topics
covered by this section. Readers who require a more detailed understanding of bushings, should also
refer to the WG A2.43 TB. As the specialists who developed the two TBs have different experiences
and backgrounds, they may have slightly different views or opinions of some topics. Readers should
consider the information in both TBs when determining how to best assess bushings.
The core insulation of bushing rated above 36 kV is generally composed of graded capacitance (a
series of concentric layers made of conductive material) that allows a uniform distribution of the
electrical field. In order to facilitate the bushing diagnostic, one layer (usually the last or the last but
one layer) is accessible externally by a capacitive tap also called measuring tap or test tap. This
capacitive tap will allow to measure exclusively the capacitance and power factor of the bushing even
if the bushing is connected to the transformer winding that has usually a capacitance that is
significantly higher than the capacitance of the bushing.

Foil Layers

Oil Impregnated
Paper
Center Conductor

Figure 9-1 Core insulation of bushing

Table 9-1 notes some different bushing types, based on design and construction.
When the transformer is in service, the tap may be connected to a monitoring system for measuring of
voltage, capacitance, dissipation factor, transient over-voltages and partial discharges. If no monitor is
used, the tap is connected directly to ground. Any weakness of the connection loop to ground may
cause an overvoltage stress and damage the bushing.

77
 Condition assessment of power transformers

Some bushings don’t have grading layers to even out electrical stress. Some older RBP bushings with
grading layers don’t have a measuring tap. In this case, the outermost layer is connected directly to
the flange. Some of these bushings are mounted with an insulating plate between the bushing flange
and the transformer tank to make the measurement of capacitance and dissipation/power factor
possible. In service, both are connected by a short circuit cable or bar, but for measuring purposes
these connections can be removed and the bushing flange is used as measuring electrode.
The core insulation may be made of the following material:
Table 9-1 Types of transformer bushings

Bushing core insulation Definition

Oil impregnated paper (OIP) They are made of paper impregnated by insulating oil and additional free oil
between the core and the exterior housing.
Resin Bonded Paper (RBP) This type of bushing core is made of paper coated with a uniform and thin
layer of epoxy resin, with semi-conductive material such as graphite
interspersed to form the concentric grading layers. Most manufacturers
have discontinued the production of RBP bushings due to quality problems.
Resin Impregnated Paper This type of bushing insulation is made of paper impregnated with resin. A
(RIP) resin-impregnated paper bushing can be provided with an insulating
envelope. The intervening space can be filled with an insulating liquid or
another insulating medium. At lower voltages, the epoxy impregnated
condenser core can have the silicone sheds directly moulded or extruded on
its surface.
Resin Impregnated This new type of bushing insulation is made of synthetic textile impregnated
Synthetics (RIS) with resin. The resin can be filled with The insulating envelope is equal to
the same as for RIP.
Other Bushing Types for For direct connection to gas insulated switchgear (GIS), gas insulated
Special or Unique bushings are used. These bushings use SF6 gas as the main insulation
Applications medium in conjunction with concentric metallic screens to control the
electric field. Also used for through-wall bushings in HVDC applications
Hybrid bushings Using a combination of different insulating technologies are also used e.g. in
HVDC applications.

Bushing envelope Definition

Porcelain Insulator Historically, porcelain has been the only material available to provide both
the necessary mechanical strength and electrical insulator properties
required for the outer shell of a bushing. In cases where the installed
bushings are not suited to a heavy pollution situation, corrective measures
include applying special grease on the porcelain or regularly washing the
porcelain.
Composite Insulator Composite insulators are made either with an internal tube of glass resin
surrounded by a silicone shell, or with epoxy in the case of the full epoxy
design. The condenser body would typically be of the RIP and RIS design but
also OIP exist. Any space between the insulator and condenser core may be
filled with oil, gel, solid foam or gas or, at lower voltages, with silicone
directly moulded or extruded on the condenser body. The selection of
silicone rubber quality is of outmost importance for the service life for the
bushing

Transformer bushing failure modes

The below shows the typical defects and failure-modes of HV transformer bushings. Bushing failures
often result in fire or explosions.

78
 Condition assessment of power transformers

Contact problem between the outmost grading layer Ingress of moisture into the resin surface due to
and the conductor tube. wrong storage

Contact failure at the measuring tap due to water Contact failure between the inner measuring tap
ingress and corrosion connection and the outmost grading layer due to a
“cold” soldering joint.

Partial breakdown between grading layers or at the Partial breakdown in a OIP insulation
overlapping of a grading layer due to fast transient
over-voltages.

79
 Condition assessment of power transformers

Leakage in the expansion vessel Loss of oil pressure

Deformed connection bolt Arcing between phases due to a branch of a tree

Defective or missing sealing gasket at the cap of the Local breakdown of the paper insulation between
measuring tap the HV conductor and the conductor tube due to
fast transient over-voltages

Figure 9-2 Bushing failure modes

The various bushing failure modes generally lead to a variation in the insulation – changing the
leakage path to ground. No bushing is perfect, and there is a small leakage current to ground. As the
insulation deteriorates the leakage current may rise and may degrade in terms of power factor
(dissipation factor); partial discharge (PD) may be generated. Detecting these changes is performed
through offline testing and inspection or through on-line monitoring.

80
 Condition assessment of power transformers

Transformer bushing test and diagnostics

Table 9-2 Bushing assessment methods

Type of fault
Moisture
Ageing
Contact problems and high impedance faults between layers
Shorted grading layers
Leakage
Partial discharges
Diagnostic techniques
OIP 5 1 5 5 5 5
Visual inspection RBP 5 1 5 5 5 5
RIP/RIS 5 1 5 5 5 5
OIP 4 5 4 3 4 5
General

Thermography RBP 4 5 4 3 4 5
RIP/RIS 4 5 4 3 4 5
OIP 5 4* 1 3 4 4
Capacitance RBP 5 4* 1 3 4 4
RIP/RIS 5 4* 1 3 4 4
OIP 3 5 1 2 2 2
DF/PF RBP 3 5 1 2 2 2
RIP/RIS 3 5 1 2 2 2
OIP 3 5 1 1 1 1
Dielectric Response with FDS/PDC RBP 3 5 1 1 1 1
5
Electrical tests

RIP/RIS 3 1 1 1 1
OIP 1 5 3 1 5 4
Partial discharge RBP 1 5 3 1 5 4
RIP/RIS 1 5 3 1 5 4
Dissolved gas analysis 1 5 1 1 4 5
Moisture in oil OIP 5 5
4 5 3 1
Colour of oil 5 2 5 1 1 4
Oil

1=very detectible, 2=moderately detectible, 3=detectible, 4=poor detectability, 5=not detectible *big leakages can
be detected with hot collar test

General Inspections
Visual inspection
A first basic way of assessing status of any kind of bushings and almost the only method to assess the
so-called DIN bushings is visual inspection. Items to look out for are oil leaks, high or low oil level in
the glass reservoir or sight glass, the colour of the oil in the reservoir, burn marks, insulator damages
such as worn off flakes and micro cracks, indication of high stresses on the connection to the outer
terminal etc.
Based on any findings during visual inspection a decision can be taken to undertake any further
measurements.

81
 Condition assessment of power transformers

Thermo-vision/ Thermography
Using an IR/sensitive camera, see picture below, hotspots on the bushing surface can be detected.
Low oil level in an OIP bushing may also be detected using this method. At maximum rated current,
the bushing´s outer terminal reaches a temperature of about 35-40˚C higher than ambient air.
Significantly higher temperatures, especially at lower current loading, can indicate bad connections.

Figure 9-3 Overheating of the top terminal seen using an IR/sensitive camera.

Electrical measurements
Capacitance and dielectric dissipation factor (Tangent Delta) or Power Factor
A change in bushing capacitance or power factor (dissipation factor) can be an indicator of insulation
deterioration, failure, ageing of insulation, partial discharges or partial breakdowns of capacitive
bushing layers. It is important to understand causes of variation in test results – the Garton effect, for
example shows that power factor (dissipation factor) results at higher voltages may be elevated due to
contamination and moisture [53], but may also vary with temperature in ways unique to each bushing
design.
The most common techniques for assessing a bushing’s condition are dissipation/power factor and
capacitance measurement. Test results are then compared to nameplate values or previous tests.
Increases or decreases from reference values are usually an indication of contamination and/or
deterioration of the insulation system. Limits for the maximum permissible change tend to be
manufacturer and type specific; however, a doubling of the initial power factor value or an increase
over 1% is often taken to indicate either more frequent monitoring/testing is required or replacement of
the bushing. A 3% change in capacitance is also taken as an indication of a need for more frequent
testing/monitoring or bushing replacement. The evaluation of a bushing is based on a comparison of
tests on similar bushings on the transformer and the factory test report. Also, trend analysis and
comparison to bushings of the same type are important – the limits for particular designs and vintages
of bushings may vary.
If a change in capacitance and/or dissipation/power factor is indicated on an OIP bushing an analysis
of the insulation oil is recommended.
Frequency Domain Spectroscopy (FDS) / Dielectric Frequency Response (DFR)
The change in these electrical measurements can be an indicator of moisture in the insulation, partial
breakdowns of capacitive bushing layers, ageing and contact problems. As with all electrical tests,
care must be taken to understand factors which may confound results, including oil contaminants and
ageing by-products.
The principle of this measurement is the same as Capacitance and Dissipation/Power Factor
described previously but differs by being applied at different frequencies, typically from 0.001 Hz to
1000 Hz. Interpretation of the result is based on frequency analysis of the dielectric parameters. In its
application to high voltage bushings, the measurement of Dissipation/Power Factor at low frequencies
enables the detection of moisture.

82
 Condition assessment of power transformers

Polarisation Depolarisation Current (PDC) Method

The change in polarization/depolarization currents can be an indicator of moisture in the insulation,
insulation failures, ageing of insulation and partial breakdowns of capacitive bushing layers.
The principle of this measurement is to apply a DC voltage across the insulation during a certain
period of time and to measure the polarisation current. Then the insulation is short-circuited with the
short circuit current measured (depolarisation current). Experience shows that moisture may strongly
influence these measurements. The dielectric response of insulation can be recorded in the time
domain (PDC) or the frequency domain (FDS). The PDC results can be transformed into the
frequency domain and vice versa. They contain more or less the same information, except the
information for higher frequencies above 0.1Hz which cannot be derived from the PDC measurement.
Care must be taken.
Partial discharge (PD) measurement
Partial discharge activity typically occurs within insulation voids, on ungrounded metal objects lying in
an electric field, or as corona due to the intense electric stress on the insulation surrounding sharp
edged/shaped electrodes. The long term effect of PD is destructive for insulation systems, by
degrading their insulation qualities. A significant increase of the PD level or the rate of the PD can
provide an early indication for evolving defects. The PD pulses generate electro-magnetic waves,
acoustic signals, chemical reactions, local heating and optical signals. Different techniques can be
used to detect these phenomena and thus infer the presence of PD.
For detection, different types of coupling are used such as bushing capacitance taps, high frequency
current transformers or coupling capacitors. The test method outlined in IEC 60270 [40] is a useful
tool, because it is a sensitive method and can be calibrated. However, electromagnetic noise external
to the transformer can interfere with the measurement. For this reason, the method may be unsuitable
for some on-site measurements. Multi-channel PD measuring systems which can record the PD
simultaneously on several channels can sometimes be used to separate PD sources from each other
and from external electromagnetic noise (interference) New Diagnostic Tools for High Voltage
Bushings [54] [2] Frequency Dependent Dissipation Factor Measurements for the Assessment of High
Voltage Bushings [55].
For PD measurements on new bushings limits are given in IEC 60137 [56] and IEEE C57.19.01 [57].
In case of OIP bushings oil samples may be taken and analysed when PD levels are deemed too high.
 Note: At the time of preparation of this document, there were no guidelines published regarding
suitable or acceptable levels of PD in bushings in service. IEC 60270 does offer guidelines for
new bushings during FAT (factory acceptance testing) but these values should not be used for
bushings installed in transformers.
Oil
Dissolved Gas Analysis (DGA)
Oil sampling and connected DGA analysis is a good method of assessing OIP bushings for insulation
failures, ageing of insulation, partial discharges (e.g. bad contacts on capacitive layers of bushings),
partial breakdowns of capacitive bushing layers. However, the practicality of taking oil samples may be
difficult and there is a risk to the integrity of the bushing should a sample be taken poorly – introducing
an increased risk of failure. It should be noted that oil sampling should be done by trained and certified
labour as history has shown that failures of bushings have been have started after improper
conducted oil sampling. IEC Technical Report 61464 [58] gives a clear guidance how to take oil
samples and what actions to take based on the results from the DGA.
The colour of the oil in the glass reservoir is also an indication of the degradation of the oil, and should
be noted, together with the level of the oil in the reservoir.

Failure mode assessment

Bushing condition can be assessed using the tests listed above, supplemented by family history,
results of inspections and inductive reasoning as appropriate and in conjunction with the assessment
scoring matrix. The tables in Annex A can also be used as a starting point when assessing a
transformer, but users should consider if the values are suitable for their fleet of transformers,
operating conditions, maintenance practices, and the time scales used in their scoring matrix. User
should refer to Chapters 1 to 7 before assessing their transformers.

83
 Condition assessment of power transformers

The final assessment of the bushing condition is based on the weighing of several parameters such as
tan delta/DGA/capacitance, application (GIS, shunt reactors), loading conditions, age and known
weaknesses in the design. If possible and/or needed should the assessment be done together with the
bushing manufacturer.

Transformer cable boxes

The purpose of a cable box is to protect and insulate the connection point between the cables and the
transformer bushings. There are various types installed world-wide with a variety of insulating
mediums including oil, various compounds such as pitch or tar, or air. The assessment techniques will
vary depending on the construction type.
Cable box failure modes
Failure of the transformer cable box may require transformer cable box replacement, major work or
corrective maintenance depending on the nature of the fault.
Liquid Filled Cable Boxes

Figure 9-4 Phase to Earth Fault in liquid filled 11kV Cable Box

Figure 9-5 Degradation and Crystallisation of insulating liquid in 11kV and 33kV Cable Boxes

84
 Condition assessment of power transformers

Dry Type Cable Boxes

Figure 9-6 33kV Dry Type Cable Box Failure due to PD Activity

Failure modes include;

 Moisture ingress
 Loss of insulation medium
 Oil leakage
 Changes in oil quality
 Condensation
 Partial discharge activity
 Overheating
 Standoff supports / cable thermal expansions

Tests and information for assessment / Test and diagnostics

Proposed assessment methods for the two types
Air insulated cable boxes
1. Visual inspection, from outside and inside
Check for oil leaks from inside the box for example between the bushing flange and the
transformer cover.
Look for signs of leakage current and PD-activities
Oil-filled cable boxes

2. Visual inspection from outside for oil leaks

Oil sampling and DGA analysis from the cable box. Evaluate the result as per 12 Insulating liquid.
Additionally, capacitance and dielectric dissipation factor (Tangent Delta) and Power Factor on the
bushings in those cases where bushing test tap is accessible from outside. The assessment of the
values as per table for failure modes for bushings. Note that this is not considered part of the cable
box assessment but the assessment of the bushing component.

85
 Condition assessment of power transformers

Failure mode assessment

Cable box condition can be assessed using the tests listed above, supplemented by family history,
results of inspections and inductive reasoning as appropriate and in conjunction with the scoring
matrix. The tables in Annex A can also be used as a starting point when assessing a transformer, but
users should consider if the values are suitable for their fleet of transformers, operating conditions,
maintenance practices, and the time scales used in their scoring matrix. User should refer to Chapters
1 to 7 before assessing their transformers.

86
 Condition assessment of power transformers

On load tap-changer (OLTC) / de-energized tap-

changer (DETC)
On-load tap-changers (OLTC) can be divided into two groups according to the type of current limiting
impedance used:
1. Resistor type OLTCs are suited to high voltage power transformers but require a high speed
mechanism to limit time the resistors are inserted to typically 50 – 150 ms. These spring-operated
mechanisms can be sophisticated and require more skills for maintenance. HV applications are
more susceptible to dielectric failure. Therefore, oil quality is also a maintenance issue for most
resistor type OLTCs.
2. Reactor type OLTCs are limited to lower voltage (less than 138 kV) power transformers but they
can typically carry higher loads and have no limitation in transition time. The reactor type OLTCs
all feature the use of a reactor that provides the OLTC with twice as many voltage positions when
compared to a resistor type OLTC that uses the same number of steps in the tapping winding.
High load applications make them more susceptible to coking, but most are equipped with slower
motor-driven mechanisms that are easier to maintain. The reactor usually brings additional
transformer cost, size and, occasionally, testing issues.
The OLTC incorporates one or more of the following components:
 Diverter Switches are always used in conjunction with a tap selector that comprises two sets of
moving and stationary contacts that alternately carry the load. The operation of the Resistor Type
OLTC involves two steps. First, the target tap is preselected with the unloaded moving contact.
Then the Diverter Switch transfers the load to the pre-selected tap. This contrasts with the Reactor
Type OLTC where the Diverter Switch has to first interrupt the load on one set of moving contacts
before that set can be used to pre-select the next tap. Diverter Switches that are the Resistor Type
must have high speed mechanisms (usually spring operated) in order to limit contact wear from
arcing and to limit the temperature rise of their transition resistors.
 Selector Switches effectively carry out the two-step operation described above, but with only a
single motion. This is achieved with an arrangement of auxiliary moving contacts that can bridge
across the two taps for the necessary make-before-break transition. Selector Switches that are
Resistor Type must also have high speed mechanisms for the same reasons outlined for the
Diverter Switches. Selector Switches are usually spring-operated for high-speed action when used
on Resistor Type OLTCs, and motor-driven with slow action on Reactor Types. These devices are
usually limited to equipment voltages up to 138 kV and having less than 33 tap positions.
 Change-over Selectors (coarse or reversing) are designed to carry, but not to make or break,
through-current. They are used in conjunction with the tap selector or selector switch to enable its
contacts and the connected taps to be used more than once, thus increasing (doubling) the
number of operation positions for the same number of taps.
 Vacuum type tap-changers can use all principles described above, but in most cases through the
addition of a number steps in the sequence and by-pass or auxiliary transfer switches.
 Motor Drive Mechanism and External Drive Shafts are used for the operation and control of the
OLTC.
For more details on the different switching technologies and applications, refer to IEC 60214 Parts 1
and Part 2 [59] [60].
Unlike the active part of the transformer, OLTCs could theoretically be kept in working conditions
indefinitely by maintaining, repairing or replacing parts (including complete components). End of life
decision is therefore not typically based on a simple condition criterion but rather occurs in the
following situations [61]:
 Failure beyond economical repair
 Maintenance costs vs new technologies
 Sustainability i.e. availability of maintenance skills or parts

87
 Condition assessment of power transformers

 Inadequacy of OLTC rating and specifications

Reliability surveys show that, depending on a number of different factors such as voltage, application
and more importantly maintenance policies, OLTC can account for 10% and up to 50% of all
transformer failures [62] [63] [16]. It is clear that the OLTC (and to a lesser extent DETC) will influence
the overall reliability of the transformer and that it should be reflected by the transformer assessment
index.
Also, transformer asset managers should be aware of possible sustainability issues, e.g. cases where
the original OLTC/DETC manufacturer no longer offers parts or service. Problems related to this kind
of situation: availability or high costs of spare parts, lack of technical support and upgrades, and
scarcity of skilled service technicians will eventually impact the reliability of the transformer. Because it
is not possible to factor this case into a transformer assessment index. based on the failure modes,
asset managers should consider and handle these as special cases with an individualized action plan
(ex: retrofit with a new OLTC technology). The same applies to OLTC vintages that have some known
design related reliability issues that cannot be corrected by regular maintenance.

Failure modes
A number of failure root cause analysis are available through the literature. For example,
IEEE C57.140 Guide for Evaluation and reconditioning of Liquid Immersed Power Transformers [6]
has proposed a simplified failure mode structures for both OLTC and DETC shown in Figure
10-1below. It clearly illustrates that DETC have only a subset of all the possible failures on an OLTC,
and that these two components can therefore be treated altogether.

Figure 10-1 OLTC and DETC failure root cause analysis [6]

A more comprehensive failure model has been proposed by CIGRE in Technical Brochure 227 [64] in
which failure modes have been grouped by components (Diverter, Selector and Drive Mechanism)
and under three categories: Mechanical, Electrical (Thermal) and Dielectric. This model will be
developed further in the following sections mainly by adding a new category (Chemical) to cover
corrosive sulphur problems that have emerged more recently.
Indeed, surveys that have included maintenance issues as possible root causes show that these
account for a significant part of the total OLTC failures. For instance, Figure 10-2 Major transformer
failure due to OLTC Flashover - Root cause analysis [56]shows the OLTC failure cause breakdown
statistics by a large North-American utility [62]. It can be observed that Insufficient or Incorrect
maintenance work accounted for more than half of the major failures (flashover) originating from the
OLTC.

88
 Condition assessment of power transformers

Figure 10-2 Major transformer failure due to OLTC Flashover - Root cause analysis [56]

Thermal
For this category, two types of thermal failure will be distinguished:
 Overheating: Overheating of the OLTC compartment would typically occur when transition
resistors remain inserted due to an incomplete diverter switch operation. Depending on the
switching principle, this may lead either to a quasi-instantaneous and disastrous failure of the
resistor (load current through the resistors), or to a high heat release (up to a few kWs) over
possibly a long period of time (when resistors are inserted in parallel to load current path). The
latter leads to significant temperature increase with a degradation of the oil and insulating
materials which will eventually cause a flashover. Most likely failure causes include Incorrect or
insufficient maintenance and weakened diverter springs.
 Coking: Coking can result either from a loose connection or the degradation of a contact surface.
Reduced conductivity leads to a hot spot on which carbon particles will deposit to form a hard
coating (coke) further aggravating the contact quality and heat dissipation. This whole sequence
is unstable and eventually develops into a thermal runaway. Total heat dissipated typically range
in the few hundreds of W, but as it is concentrated on a small surface it can be sufficient to cause
severe local overheating, hence DGA techniques have been applied successfully. The overall
temperature rise of the oil compartment over the main tank temperature will be in the range of 5 to
20 ˚C so this failure could also be detected at its latest stages by a thermograph scan.
Dielectric
In this category, only purely dielectric failures will be considered, in other words a component rupture
that eventually results in a flashover will be treated as a mechanical failure. Other than external
influences (over voltage, lighting, etc.) that leaves only two possible causes for a flashover or tracking:
moisture ingress in insulation material and build-up of carbon deposits (or other contaminants/particles
in oil) due to insufficient maintenance.
Chemical
Numerous failures of transformers have been recently related to the formation of copper and silver
sulphide on metal surfaces and copper sulphide deposits in the insulating paper in the windings.
CIGRE Technical Brochure 625 [37] offers an excellent summary of the current knowledge.

89
 Condition assessment of power transformers

Concerning the possible sources elemental sulphur, there are a few possibilities according to recent
findings:
1. High concentrations of elemental sulphur can be released by oil reclamation processes using
reactivated clays due to the high temperatures used in the reactivation process. Elemental sulphur
cannot be naturally present in the original oil. Many cases have been reported “where the oil was
believed to be non-corrosive beforehand or even tested and found to be non-corrosive and,
sometime after the treatment, the silver contacts were found to be coated with silver sulphide”
[37]. In at least one reported case, it led to a dielectric failure just weeks after the treatment. This
occurred even though investigations revealed that the reclamation processes could generally be
considered appropriate. This situation would not generally cause a severe/immediate failure event
if the OLTC didn't have silver plated contacts, since copper is less susceptible to a reaction with
elemental sulphur.
2. Silver corrosion in OLTC was also observed in corrosive oils with DBDS or without DBDS, as well
as in oils tested originally as non-corrosive, and on which no reclamation process had been
performed. These are suspected to be result of oil becoming corrosive after it has reached and
advanced stage of oxidation.
3. This situation would likely not cause any problem as long as the OLTC does not have silver plated
contacts.

Figure 10-3 Examples of silver sulphide formation on OLTC contacts [37]

Mechanical
Mechanical failures can be divided into two groups: Motor drive failures and leaks.
Motor drive failures including:
1. Incomplete operation: Operation is interrupted before completion (ex: Motor trip)
2. Incorrect operation: OLTC does not reach the set position (ex: OLTC runs through)
3. Wrong tap indication: Actual position and relayed indication do not match (ex. Bad
synchronism or erroneous position signal)
4. Drive linkage failure: Mechanical failure (ex: rupture, seizure) due to inadequate maintenance.
It will normally lead to minor material consequences, but still result in unplanned outages. Statistics
about these minor outages show a strong correlation with age as shown in Figure 10-4. For instance,
a 50 years old motor drive would have a yearly probability of an unplanned outage of more than 10%.

90
 Condition assessment of power transformers

16%

14%

Minor Failure Year Rate

12%

10%

8%

6%

4%

2%

0%
<2 2-7 7-12 12-17 17-22 22-27 27-32 32-37 37-42 42-47 >47
Age (years)

Figure 10-4 Probability of OLTC minor failures [62]

The probable cause of these minor failures by components has been established by the same utility
as follows:
 40% OLTC controls (relays, contactors, cams etc.)
 14% Braking system
 12% OLTC contacts (coking)
 6% Motor
 28% Other (lubrication, wiring etc.)
The relationship with age shown above, the nature of the of failures and field observations suggested
that most problems are related more to slow degradation due to stressors (such as moisture, heat and
dust) than to wear, as no evidence could be linked to the number of operations.
Definitely more serious consequences (including internal flashover) can result from a bad synchronism
between the motor drive and the OLTC selector. Such event can occur on OLTCs that do not have a
protection (weak point acting as a mechanical fuse in a location easily accessible for repair) and will
be caused by the motor drive forcing through a mechanical stop at a limit position.
Leaks can be classified essentially as internal or external leaks. Internal leaks will mainly cause either
a misleading interpretation for DGA monitoring of the active part when the oil flows from the OLTC
compartment to the main tank or OLTC oil overflow when the oil is migrating in the opposite direction.
The oil flow is essentially determined by the difference in conservator heights and many utilities prefer
to have the oil leak from the main tank to OLTC to avoid transformer oil contamination with carbon and
dissolved gases. The issue of controlling the overflow appears comparatively more manageable,
typically a container collects the overflow of oil through the OLTC compartment drain pipe. However
some leaks may develop to an extent that topping off the transformer oil level may be required on a
regular basis.
External leaks are defined as direct leaks from the OLTC compartment to the environment, mainly
through the gaskets and seals. These are less easily manageable as they can seep and drip from all
around, therefore they can cause considerable cleaning and environmental costs.
Although leaks appear to have a marginal effect on transformer reliability, they still represent a major
reason for transformer replacement as repairs can be quite difficult and very costly.

Tests
A number of on-line diagnostic tools have emerged recently for the condition assessment of OLTCs.
Table 10-1 below gives an overview of the detection capabilities of the main diagnostic methods
available.

91
 Condition assessment of power transformers

Table 10-1 Failure detection capabilities of different OLTC tests [5]

DIAGNOSTIC METHOD OLTC TYPE

Vacuum
VIBRO-ACOUSTIC Reactor
Resistor
MOTOR TORQUE All
Vacuum
DISSOLVED GAS ANALYSIS Reactor
Resistor
In-Tank
IR THERMOGRAPHY
Compartment
DYNAMIC RESISTANCE All

PROBLEMS
Linkage/Gears G G G G
Timing/Sequence G E E E
Control/Relays E
L-Low Motor G
M-Moderate Brake E
G-Good Lubrication E
E-Excellent Contacts alignment G G G G
Arcing M G E E E E
Overheating/Coking G L M E E L M L
Contact wear M E M L
Transition E E E E

It appears that any Condition Assessment program for OLTCs would have to rely on the
implementation of more than one technique to cover the widest possible range of failures. Although
these tests can provide valuable information for the operation of an OLTC, especially those that can
be performed on-line, a comprehensive condition assessment can only be achieved by a detailed
visual inspection of the internal parts. Regular (systematic) maintenance consists as a major part in all
these visual checks in addition to cleaning and oil filtering. Also, when regular maintenance is
performed in compliance with the OEM recommendations, many failure modes can simply be
eliminated by preventive replacement of parts and timely implementation of design upgrades.
Accordingly, the maintenance policies and practices of an OLTC operator should also be considered
as a possible root failure causes.

Failure mode assessment

Tap-changer condition can be assessed using the tests listed above, supplemented by family history,
results of inspections and inductive reasoning as appropriate and in conjunction with the assessment
scoring matrix. The tables in Annex A can also be used as a starting point when assessing a
transformer but users should consider if the values are suitable for their fleet of transformers,
operating conditions, maintenance practices, and the time scales used in their scoring matrix. User
should refer to Chapters 1 to 7 before assessing their transformers.
The following data should be considered to be available and relevant for the condition assessment of
OLTCs:
1. OLTC Manufacturer, model and age
This information is key in pinpointing vintages that have identified issues such as known design
weaknesses, lower service reliability or unavailability of some parts. Also, when the OEM is still
supporting the model, all released critical modifications or upgrades can be checked
Age alone is not an adequate indicator of the general condition of the OLTC, but it can be correlated to
the reliability of some components such as the motor drive.

92
 Condition assessment of power transformers

2. Review maintenance records and policies including:

 Last date of service and observations from the inspection
 List of critical modifications and upgrades performed
 List and type of alarms/trip during service
 Qualification of the service technician
 Maintenance practices and more particularly deviations from the OEM’s recommendations (time,
operations, replacement/repair criteria)
3. Accumulated number of operations and operating range
The number of operations is normally not a relevant indicator of the expected reliability provided that
the OEM maintenance recommendations are applied and that the accumulated number stays within
the maximum lifetime number rated by the OEMs. This evaluates the overall wear of the OLTC and
gives a reference criterion above which the unit’s reliability starts to be affected and its condition is
considered to be beyond cost-effective repair.
For modern products, typical ranges of values issued by OEMs are:
 Diverter switch : 800,000 to 1,200,000 operations
 Selector switch: 1,000,000 to 4,000,000 operations
 Vacuum interrupters: 600,000 to 1,000,000 operations
 Motor drive unit: 1,000,000 to 4,000,000 operations
For older OLTC vintages, different OEM limits (typically lower) may apply. For a normal network
transformer with approximately 5,000 yearly operations, it is clear that this criterion will more likely not
impact the transformer assessment index significantly.
4. Load history
Load deviations from the rated load can affect ageing and wear mechanisms. Main components that
can be affected are insulating material degradation and contact wear. For this purpose, the average
load factor and general characteristics of overload occurrences (peak and yearly frequency) should
normally be available.

Scoring Matrix for tap Changers

The generic scoring matrix illustrated in 2.3 has been adapted to more easily allow the assessment of
tap-changers.

Scoring matrix for tap-changers

F Indications of extreme degradation or ageing, or excessive maintenance delays have occurred. A
failure is likely to occur in the short term.
E Indications of severe degradation or aging, or critical maintenance operations are overdue. Reliable
operation is impaired.
D Indications of advanced degradation or aging, important delay in required maintenance. Reliable
operation is compromised.
C Indications of significant aging or deterioration, important delay in required maintenance. Reliability
may be affected
B Indications of slight aging or deterioration but still within the manufacturer’s limit. Reliability is not
affected
A All tests results are well within normal limits and maintenance is up to date

93
Condition assessment of power transformers

94
 Condition assessment of power transformers

Cooling system, transformer tank and ancillary

components
This section focuses on the cooling system, transformer tank and ancillary components. Although
these are not part of the main power circuit it does have a major influence on the capability and
reliability of the power transformer. It is thus important to monitor the condition of these components
closely and resolve any problems as soon as they develop as it can affect the long term operational
performance of the power transformer.

Cooling System
The cooling system plays an integral part in the design and operating performance of a power
transformer. A description of each type of cooling system can be found in TB 445 [5], IEC 60076-2 [24]
and IEEE C57.12.00-2015 [65] . It is very important that the energy from the losses within the power
transformer is efficiently transferred out into the surrounding environment. Working group A2.18 have
identified that one of the defective/faulty conditions typically attributed to the general overheating or
abnormal rise of the oil temperature may be due to cooling deficiency [64].
Failure modes of cooling system
The critical wearing out of cooling system components affect the two basic functions of the cooling
system i.e. oil circulation and heat exchange. Inefficient oil circulation is affected by the convective
ability of the oil (ON) and pumping forced oil/water flow (OFAF, ODAF, OFWF, ODWF). The heat
exchange (loss dissipation) efficiency is affected by loss of oil forced flow, oil flow blockage, pump
deficiency (opposite direction of rotation, bearing wear, pump motor failure, improper oil flow rate, fan
deficiency/failure, coolers/ pipe contamination and air flow blockage due to contamination.
Fan failure
The cooling fans form an important part of the cooling system, especially for large transformers.
Usually with more than one cooling fan failure the transformer load must either be decreased or the
transformer taken out of service due to the significant drop in air flow which reduces the rate of heat
transfer from the transformer oil to the environment. This decreases the efficiency of the cooling
system with heat starting to build up within the transformer and if the transformer was running at full
load, it will result in higher oil/winding temperatures until it is alarmed / tripped.
Wrong direction of fan
When maintenance work is done on the coolers there may be a chance of the fans being connected in
the wrong direction resulting in inefficient cooling due to recirculation of warm air. This will result in
elevated oil temperature.
Pump failure
The correct flow and circulation of oil in the transformer tank, windings and piping is important to allow
for the effective heat transfer from the windings to the external environment. If a pump fails, the oil flow
rate decreases and decreases the efficiency of the cooler. This will result in elevated oil temperature.
Wrong direction of pump
When maintenance work is done on the coolers there may be a chance of the pumps being connected
in the wrong direction. This has the effect of circulating oil in the wrong direction through the coolers
and reducing the efficiency of the coolers. This will result in elevated oil temperature. This failure mode
can affect the general cooling of the transformer as it not only reduces the cooling efficiency of the
system but also hinder the natural flow of oil from bottom of tank to top of tank and the result may be
disastrous as it will work against of the basic principle of cooling by convection.
Control circuit failure
The control circuit of the cooling system plays an important role in ensuring that the correct amount of
fans and pumps from the coolers are activated for the relevant loading. If there is a problem with the
control circuitry or components this will impact the operation of the cooling system and efficiency of
cooling. This will result in elevated oil temperature.

95
 Condition assessment of power transformers

Blocked ducts/pipes
Usually with old/aged oil there is a high level of particles and sludge formation which may build up in
the cooling ducts, piping and radiators. The effect is a layer of material on the surface of these
components insulating the metal surfaces and reducing the efficiency of heat transfer. Another
mechanism is the blockage of the ducts, cooling channels and pipes due to sludge formation or
foreign debris thus impeding oil flow and reducing heat transfer. Oil-water heat exchangers can also
be blocked on the water side due to deposits or corrosion causing decreased cooling efficiency. See
Figure 11-1.

Figure 11-1: Particle build-up in water cooler

Closed radiator valves

Usually after maintenance some of the radiator valves can be left closed. This prevents oil from
circulating within the radiators and thus decreases the efficiency of cooling.
High oil viscosity
The viscosity of oil affects the oil movement in the convection process and the through the winding
cooling ducts. High oil viscosity reduces the cooling efficiency. The viscosity of oil is affected by
dissolved particles and oil ageing. The viscosity is sharply dependant on the oil temperature. The
correct oil type must be selected considering the expected minimum ambient temperature and
precautions must be taken at transformer cold-load pickup [3].
Dirty radiators and fins
Radiators, pipes and headers may experience a build-up of rust, corrosion, dust deposits and peel off
paints which could reduce the efficiency of heat transfer to the environment.
Leaks
The cooling circuit may develop oil leaks usually found at the interface points to the tank and ancillary
components. Corrosion due to galvanic action of dissimilar metals in the lower portion of the cooler at
the juncture of the tubes and the tube sheets can cause oil leaks. Thermal expansion and contraction
can cause leaks by means of deforming the relatively soft tubes especially at the joints. Leaks cause
loss of oil, which must be environmentally controlled. Moisture ingress also occurs at the point of
leaks.
Oil temperatures below zero Degrees Celsius of WF coolers can lead to ice creation in water pipes.
Ice may destroy the integrity of WF coolers and may cause water penetration to the oil side of the
cooler. Water pipes integrity can be affected also by coolers ageing, repetitive cycles of pipes thermal
extension/contraction and sealing wear. Water in oil can drastically reduce its dielectric strength and
cause accelerated insulation aging.
Elevated air ambient temperatures
Although ambient temperature is an external factor and not controllable in open air environments; it
may affect the heat transfer from the transformer to the environment. The effect is significant for
transformers located in enclosed areas (buildings). This usually occurs if the transformer was not
properly designed for the environmental conditions and will have to be managed accordingly. The
higher the ambient temperature the slower the heat transfer and the equilibrium temperature is higher.
This causes the internal temperature of the transformer to increase.
Elevated external cooling water inlet temperature
For water coolers it is important that the inlet cooling water temperature is low enough or as per
designed specifications to allow for efficient heat transfer. During summer months the cooling water
temperature is usually higher and may decrease the cooling efficiency. This usually occurs if the

96
 Condition assessment of power transformers

transformer was not properly designed for the environmental conditions and will have to be managed
accordingly.
Test and diagnosis
Working group A2.18 have described diagnostic methods for the cooling system consisting of forced
oil flow rate/pressure, pump vibration, pump motor current, metal-in oil analysis, difference in
inlet-outlet cooler temperature, temperature distribution across the transformer tank and oil pipes,
temperature distribution across the cooler, radiator air flow rate /direction, fan motor current and
bearing monitoring system indicating wear [64].
Infrared scanning
Infrared scanning provides a relative difference in surface temperature. On a transformer infrared
scanning can locate areas of elevated temperature and assist in identifying for example a radiator
valve being closed. One of the advantages of infrared scanning is that it can be performed without an
outage. Figure 11-2 indicates how IR scanning can be used to identify blocked radiators.

Figure 11-2: IR Image showing blocked radiators [66]

Temperature monitoring
Temperature monitoring is an important parameter for identifying an increase in energy within the
transformer and problems with the cooling system. The top oil temperature is usually used as an
indication of the inlet temperature of the coolers. Monitoring the bottom oil temperature either at the
bottom main tank or at the outlet of the cooler allows for a differential temperature calculation to be
made. This value can then be compared between cooler circuits to identify low differential values
which may be an indication of blocked piping, poor radiator surface condition, etc.
Working group A2.27 Recommendations for Condition Monitoring Facilities recommended that the
following temperatures should be available for condition monitoring [67]:
 Top oil – at a measurement point representative of the temperature of the oil at the top of the tank
and as close as possible to where it was measured during the factory temperature rise test.
 Bottom oil – at a measurement point representative of the oil entering the bottom of the windings
which is normally the oil temperature at the cooler outlet close to the transformer tank.
 Cooler inlet oil – measurement taken from the cooler inlet oil and may be used to monitor cooler
efficiency. In some transformer designs the top oil measurement can be used.
 Cooler outlet oil – measurement taken from the cooler outlet oil and can be used to monitor
cooler efficiency. In some transformer designs the bottom oil measurement can be used.
 Cooling medium at inlet to coolers – a measurement representative of the temperature of the
cooling medium (normally air or water) at the inlet to the coolers. In the case of an air temperature

97
 Condition assessment of power transformers

measurement, the sensor should be mounted in the shade. Air ambient temperature can be used
if this measurement is not available. For water cooling medium a sensor or thermometer pocket
should be included at both the cooler inlet and outlet.
 Cooler Performance Index:  Temp (Inlet – Outlet) - This cooler performance index is derived
by finding the difference between the Inlet and Outlet temperatures of the cooler and comparing
between coolers. This can be done easily if there is temperature monitoring on the inlet and outlet
of the cooler. If there are none available, portable temperature scans can be taken externally to
get these values on a routine basis. This delta is then compared with the designed values for the
cooling circuit. Temperature deltas more than 30% of between coolers should be investigated
further.
 Ambient temperature – Monitor the ambient temperature if the transformer is located in an
enclosed area (building). This temperature must be alarmed for immediate investigation when the
alarm value is triggered.

Oil flow monitoring

Oil flow is a very important parameter, especially for forced oil (OF, OD) cooling systems. The
designed flow rates must be maintained to achieve the associated cooling efficiency. Cooling circuits
usually have oil flow indicators which confirm that there is oil flow. It is recommended to install
analogue oil flow meters so that the flow rate can be measured and trended per cooler. Any flow
impediment can then be easily and quickly identified.
 Cooler Flow Index:  Flow (Each cooler circuit) - Particular cooling system designs have oil flow
meters installed to display oil flow in cooling system branch or in each cooler. Where there are no
such meters, periodic measurements can be performed with a portable flow meter at load or as
close as possible to transformer rating at a given cooling stage. Such measurements can be done
on the transformer online, without the interruption of the customers power supply. Where oil flow is
defined by a working pump, the measured value should be compared with about 80% of indicated
on oil pump nameplate or in pump manual (considering 20% oil flow reduction due to oil path
hydraulic resistance). Each cooler flow can also be compared with the average of all the cooling
system coolers/branches to identify any anomalies between coolers.
Visual inspection
Visual inspection of the transformer oil-cooling loop components should be performed as regularly as
prudent but should not exceed 12 month frequency.
Fans should be manually energized to ensure proper operation. Any significant noises (e.g. grinding,
rubbing, scraping) should be noted, investigated further, and corrected. Fan blade guards must meet
government safety standards. Obvious airflow obstruction through coolers via debris should be noted
and cleaned. Periodic infrared imaging of the coolers/radiators should be taken to ensure proper oil
flow and cooler/radiator heat transfer. All areas including piping, valves, and surrounding ground area
must not show evidence of oil leakage. Typical leakage points will include all pipe flange connections,
valve stems, pump electrical connections, radiator/cooler headers, etc.
Any missing, failed, or defective fans should be repaired or replaced. If the transformer’s future
expected service includes loading beyond rated power, a possible solution could be increasing the
number of fans as part of the reconditioning of the transformer. However, any cooling revisions and or
uprating of the transformers should be developed and designed by OEMs/Experts with specific
knowledge of the transformer’s cooling design, and original designed parameters of the transformer.
These may include other methods most suitable for the design.
 Cooler Condition Index: Visual Inspection - Cooler Condition Index forms part of the routine
inspections of the transformer. This requires a visual assessment of the general condition of the
cooling system. Any deviations identified must be rectified as soon as possible or at the next
available maintenance schedule.
Sludge
Sludge usually starts building up in oil and start getting deposited on tank surfaces, windings and core
and then start blocking in the cooling ducts provided in windings and in header piping’s as the
condition becomes excessive. This is a slow process and indication of this starts with the gradual
increase in temperature under normal loading and ambient conditions. Although the formation of
sludge can usually be detected by a visual inspection, IEC 60422 [42] suggests tests such as

98
 Condition assessment of power transformers

colour/appearance, acidity, dielectric dissipation factor (DDF), resistivity, acidity and interfacial tension
(IFT), which can provide indications of sludge components before visible sludging actually occurs.
Failure mode assessment
The cooling system condition can be assessed using the tests listed above, supplemented by family
history, results of inspections and inductive reasoning as appropriate and in conjunction with the
assessment scoring matrix. The tables in Annex A can also be used as a starting point when
assessing a transformer, but users should consider if the values are suitable for their fleet of
transformers, operating conditions, maintenance practices, and the time scales used in their scoring
matrix. User should refer to Chapters 1 to 7 before assessing their transformers.

Transformer tank
Power transformer tanks are generally manufactured of hot-rolled, unalloyed steel sheets. In case of
low ambient temperature at the site of installation of the transformer, fine granular steel is used to
manufacture the transformer tank. The areas of the tank where strong eddy currents can be generated
due to high currents are made of non-magnetic (austenite) steel. The transformer tank is an integral
part of transformer where the main isolating and coolant (oil) of the active part is stored. Thus, its
mechanical strength is of utmost importance to hold the pressure of the oil. The tank must also be
capable of handling vacuum pressure (during oil filling) and a sudden pressure increase which might
be experienced during ‘through faults.’ e.g. during faults in overhead lines connected to transformer
bushings.
Failure modes of transformer tank
Transformer tanks are designed depending upon the site requirements and it plays an important part
in the successful operation and utilisation of the asset throughout the life cycle of the transformer.
Generally, the tank is a very robust part of the transformer and does not hinder the operation of
transformer, but it can be a problematic area if there were any issues during transport, installation at
site, due to climatic conditions or proper maintenance not being done. The transformer tank can have
following failure modes where proper action plans for mitigation and control is required.
Oil leakage from main tank gasket joint and other welding joints
There are many joints in a transformer tank and interconnecting pipes, accessories and instruments
where any leakages must be attended to as a priority. The main tank may be of Top Cover or Bell
Tank type (Bottom Joint) and the joint may be provided with gasket, chord or O ring. In all situations,
leakage from these joints are critical to the reliability of the transformer especially due to moisture
ingress and the slow drop in the oil level from main tank can trigger the low oil alarm and can further
operate the Buchholz relay alarm and trip contacts forcing the transformer out of service. Depending
on the design of the tank, it may be possible for water to enter and collect in hollow steel components,
e.g. U-channel stiffeners. In cold climates this can freeze, effectively destroying the tank.
Corrosion and improper painting
Transformer tanks are painted with a primer and external paint coatings as per site condition
specifications, but it is important that proper maintenance is carried to check for any corrosion, rust
and paint feel off. These conditions do not have an immediate effect but the corroded/rusted areas will
eventually have an effect on the mechanical strength of the tank which can be critical when
experiencing excessive pressures developed in main tank due to abnormal conditions.
Another problem is the excessive or improper thick coatings of paint applied to the transformer over
time. This includes the effects of using the wrong type of paints (such as metallic based) which
significantly affects the heat transfer. The surface of tank forms an important heat transferring medium
and if it is coated with too much paint, or the wrong type of paint, the radiating capacity of tank can be
reduce resulting in the transformer overheating under normal load conditions.
Test and diagnosis
The transformer tank must be routinely inspected visually for defects. While doing visual inspection,
the notable problem shall be identified, and necessary additional tests may be conducted. For leakage
issues, the replacement of gaskets and repair of the tank must be planned as early as possible. For
problems associated to welding cracks/poor surface paintings/peel off the root cause must be
ascertained and corrective or preventive action must be taken.

99
 Condition assessment of power transformers

For an improperly painted transformer tank with bad surface conditions, the paint thickness must be
measured, and it shall be within permissible limits. During major maintenance, overpainted/high
thickness areas can be repainted as per OEM specifications.
Failure mode assessment
The transformer tank condition can be assessed using the tests listed above, supplemented by family
history, results of inspections and inductive reasoning as appropriate and in conjunction with the
assessment scoring matrix. The tables in Annex A can also be used as a starting point when
assessing a transformer, but users should consider if the values are suitable for their fleet of
transformers, operating conditions, maintenance practices, and the time scales used in their scoring
matrix. User should refer to Chapters 1 to 7 before assessing their transformers.
The findings from the inspection of one transformer of the fleet must be recorded and any defects
identified must be verified for possible recurrence on similar transformer designs in the fleet.

Ancillary components
A power transformer makes use of important ancillary equipment which could also affect the
performance when these components fail. The most important or generally used components are
discussed in this section.
Failure modes of components
Buchholz relay
The Buchholz relay provides an important protection for the transformer detecting gas build up and
rapid movement of oil due to a fault. Failure or malfunction of the Buchholz will either result in the
unnecessary tripping of the transformer or catastrophic failure to the transformer if an incipient fault is
not picked up with the build-up of gas. The Buchholz relay must be routinely maintained and calibrated
to ensure optimum performance.
Pressure Relief Device (PRD)
The pressure relief device is used when there is an over pressure condition within the transformer.
Failure of the PRD can either result in false operation with related draining of oil or failure to operate
causing the pressure not to be released and thus rupturing of the tank. The PRD must be routinely
inspected for leaks around joints, diaphragm cracking and tested to ensure optimum performance.
Preservation system
Liquid immersed transformers usually have a liquid preservation system as specified in IEC 60076-1
[47]. The most common types of liquid preservations systems are free breathing system or
conservator system, diaphragm or bladder type, inert gas pressure system, sealed-tank system with
gas cushion or sealed system.
Breather system
Refrigerating and dehydrating breathers with silica gel are normally used to prevent moisture from
entering the transformer tank. Failure of the breather usually results in moisture ingress into the
transformer which promotes the aging processes of the insulation.
Temperature devices
The oil and winding temperature devices provide information on the temperature of the top oil and the
simulated temperature of the windings respectively. These component settings could drift over time
and thus must be routinely calibrated.
Oil level Indicator
The oil level indicator is usually a dial type gauge which provides an indication of the oil level within
the transformer. Failure of this device usually results in the wrong oil level and if the oil level is low an
alarm is generally received. This device must also be routinely tested and calibrated.
Fire suppression system
The fire suppression system is used to extinguish or contain a fire on a power transformer. The fire
system does not affect the performance of the transformer, but it is required when a severe fault
condition is present. The failure modes are mainly related to corrosion, leakage or blockage.

100
 Condition assessment of power transformers

Lightning arresters
Usually power transformers connected to overhead lines utilise lightning arresters to protect the
transformer from insulation failures arising from lightning and other overvoltage conditions such as
switching surges. Failure of arresters is usually due to damaged, contamination (moisture, dirt and
corrosion), lightning strokes, long-duration surges due to switching and misapplication. Lightning
arresters should be inspected annually for looseness, broken parts, dirt, and other deposits.
Instrument Transformers
Instrument transformers like current and voltage transformers play an important role by linking the
primary and secondary system of the power transformer to enable for protection and measurement
circuits to be connected. The insulation system must be capable of withstanding high voltages and
currents and maintain the isolation of the systems. Other key characteristics of instrument
transformers are the accuracy, thermal and burden considerations.
Test and diagnosis
The ancillary components can be inspected, routinely tested and calibrated to ensure the reliability.
The following tests discussed are relevant to surge arresters. IEC 60099-4 [68] and 60099-5 [69]
provides a detailed guide on the testing of surge arresters. The following are basic tests that can be
utilised to either assess the condition of surge arresters or to make the replacement decisions.
Visual inspection
Visual inspection of the transformer components should be performed as regularly as possible.
Inspection frequency should not exceed 12 months. Inspection activities should include as a minimum:
 Inspect the pressure relief device on a routine basis for the general condition and check if the
yellow (oil) or blue (silicon) rod/button is visible
 For sealed or pressurised breathing systems, check the pressure gauge (against weekly data
sheets) for seasonal temperature changes. If there is no change it is an alert that the device
may be defective. Monitor the pressure and if below acceptable levels investigate possible
leaks and top up gas to correct pressure
 For sealed systems with bladders, on a medium-term routine (5 yearly), inspect visually the
conservator internally through the inspection flange
 On desiccant breathers, confirm that colour change takes place from the bottom of the
breather. If the colour change should take place from the top, check the seals, flanges, union
joints and pipes for leaks.
 On instrument transformers [IEC60044-1], verification of terminal markings must be carried out
especially after any work has been done in the circuit.
 Inspection of surge arresters to identify dirt, particle or salt deposits on the arrester housing.
These must be routinely cleaned. Inspections must also identify cracks and damage to the
porcelain housing. These must be either repaired or replaced depending on the severity of the
damage.

Calibration
Calibration testing must be performed on components that provide indications of plant condition, such
as temperature, pressure, flow or level. These can be done:
 On oil and winding temperature devices. This is usually performed by simulating the heat by
means of a hot oil bath setup.
 For oil level indicators with magnetically coupled float mechanism, remove the outside
mechanism for testing without lowering the transformer oil. A magnet can be used to confirm
function by rotating the magnet and confirming operation on the dial.
 Calibrate pressure monitoring devices on the sealed or pressurised breathing systems.
Functional, alarms and trip verification
Routine verification must be performed to check correct functioning, alarming and tripping of the
auxiliary components. These can be done;
 Manually on the Buchholz relay on a routine a basis by pumping a small amount of air into the
top chamber as per OEM testing procedure. Observe correct float operation and associated
alarm or trip contacts activating.

101
 Condition assessment of power transformers

 Use a volt-ohmmeter to check the integrity of the alarm and trip switches / contacts on the oil
and winding temperature devices, pressure relief device, Buchholz and oil level indicator.
Routine testing
All piping should be flushed to clear debris and to verify the spray pattern is still effective.
Infrared scanning

Figure 11-3: IR image showing hot spot [66]

Infrared scanning provides a relative difference in temperature which can locate “hot spots” on
arresters. One of the advantages of infrared scanning is that it can be performed without an outage.
Figure 11-3 provides an example of a localised hot spot. Thermal scans can be compared between
arresters of the same transformer or from sister units to pick up any anomalies.
Leakage current
Leakage current provides a good measure of the condition of the surge arrester, especially the metal
oxide type. The total leakage current consists of a capacitive and resistive component with the
capacitive component being the most significant. However, any change, degradation or aging in the
arrester causes a significant increase to the resistive component of the leakage current with little
change to the capacitive component. Thus, by measuring the arrester’s resistive leakage current, a
diagnosis of its condition can be obtained. An increase of the leakage current usually leads to an
increase in the element temperature which in turn leads to a further increase in leakage current. This
then results in thermal runaway and failure of the arrester.
A sudden increase in the resistive leakage current is usually an indication of water contamination
which is the major cause of arrester failures.
Insulation resistance testing
Insulation resistance test maybe conducted to assess the condition of the surge arrester. Compare
results to the previous test results.
Power factor testing
Power Factor tests are effective in detecting problems in lightning and surge arresters which are
evaluated on the basis of dielectric loss (milliwatts or watts) of comparable units or previous year’s
benchmark test data. Abnormal dielectric losses can be divided into higher than normal and lower than
normal and which are indicative of the problems as listed below:

102
 Condition assessment of power transformers

Type Higher than normal losses Lower than normal losses

Silicon Contamination by moisture, dirt or dust. Broken shunting resistors. Poor contact
Carbide Corroded gaps and open circuits between elements

Contamination by moisture, dirt or dust.

Discontinuities in internal electrical
Metal Oxide Corroded gaps in early designs (newer
configuration
designs are gapless)

Failure mode assessment

The ancillary components condition can be assessed using the tests listed above, supplemented by
family history, results of inspections and inductive reasoning as appropriate and in conjunction with the
assessment scoring matrix. The tables in Annex A can also be used as a starting point when
assessing a transformer but users should consider if the values are suitable for their fleet of
transformers, operating conditions, maintenance practices, and the time scales used in their scoring
matrix. User should refer to Chapters 1 to 7 before assessing their transformers.

103
Condition assessment of power transformers

104
 Condition assessment of power transformers

Insulating liquid
Introduction
Transformers may use one of three classes of insulating liquid for their non-solid insulation and as a
cooling medium; dielectric liquids, pressurised gas, or atmospheric air.
The various dielectric liquids, often loosely referred to as oils, are almost universally used in high
voltage power transformers and are the sole consideration for this chapter. Refer to CIGRE TB445 [5]
for discussion on condition assessment of the pressurised gas used in gas insulated transformers and
for more details regarding some of the mentioned tests.
The dielectric liquid is used for heat transfer and as an insulation medium, as well as for impregnating
paper-based, solid insulation, and for diagnostics purposes, being responsible for providing
information for assessment of normal / abnormal behaviours:
 Dielectric liquid transfers heat through its circulation. It takes up heat at transformer heat sources
(including the core, windings and stray loss heating in tank structures) and transfers heat to the
transformer heat coolers (involving either heat dissipation or heat exchange);
 Dielectric liquid acts as an insulation medium to provide insulation between the live parts and the
live parts to ground. To be a functional electrical insulator in power transformers, dielectric liquid
must attain and retain certain dielectric, chemical and physical properties. Changes in these
properties are where the condition assessment of the dielectric liquid is mainly focused;
 Dielectric liquid impregnates paper-based, solid insulation, establishing a compatible composite
insulation system. This insulation system establishes both a solid and liquid medium having
optimal resistivity and permittivity in zones subject to high electrical stress. This insulation system
is ideally suited to the complex winding geometry, connections and electrode shapes found in
power transformers. The liquid, in principle, eliminates gas voids within the paper. Otherwise gas
voids would initiate partial discharge activity;
 Dielectric liquid is responsible for transmitting information of any abnormal processes which may
be undergoing inside of the transformer. Through dielectric liquid analysis, understood hereby as
the assessment of key physical and chemical parameters of the liquid and, especially, tracking its
variation along time, and the dissolved gas analysis, it must be possible to identify potentially
harmful processes of transformer active part and internal components.
A possible division of dielectric liquids may include four groups – mineral oil, natural esters, synthetic
esters and silicone. Despite one can identify more types of dielectric liquids, this brochure will only
cover the listed ones. Mineral insulating oil is the most common and is produced from refining crude
oil. Refer to CIGRE Technical Brochure 436 [70] “Experiences in Service with New Insulating Liquids”
for a comparative study of these four transformer dielectric liquid types.
Regardless of the dielectric liquid, it deteriorates in service. It degrades and is contaminated. This is
recognised by there being international standards that address either new or used dielectric liquid. The
principal international standards currently being used for mineral oils, natural ester liquids, synthetic
ester liquids and silicones can be found in Annex D.
These standards are intended to set the essential requirements for transformer dielectric liquids. Most
power companies, network operators and transformer manufacturers have further developed their own
transformer dielectric liquid specifications, guidelines and supervision methods. Accordingly, there is a
wide variety of procedures and criteria for monitoring dielectric liquids.
Monitoring and maintaining dielectric liquid quality is essential to ensure the reliable operation and
service life of a transformer. Like most materials, dielectric liquid deteriorates. It diminishes both in
quality and its ability to withstand electrical stress and transfer heat from the core and coils to the
environment. Invariably, dielectric liquid condition deteriorates though contamination or degradation
(ageing), and, depending on the type of dielectric liquid, forming acids, and other polar compounds,
such as metallic sulphides. Dielectric liquid degradation can be premature. Rather than ‘fail’, dielectric
liquid breaks down if subjected to arcing, partial discharges or overheating. It then forms gases of low
molecular weight (the gases monitored in DGA) and sometimes carbon particles or coke. The
behaviour of each of the dielectric liquid types will differ in the quantity of each by-product formed,
especially the coking tendency.

105
 Condition assessment of power transformers

This chapter will compare the value and significance of various standardised dielectric liquid tests. It
will provide a broad understanding of degradation and contamination of dielectric liquid. It will explain
why they are detrimental to a transformer and discuss the available methods to slow, prevent or
correct degradation and contamination of dielectric liquid. It will also illustrate how to apply criteria for
the evaluation of test results and test data so that informed assessments can be made about the
condition of the dielectric liquid in a transformer.

Degradation and contamination

Dielectric liquid degradation occurs throughout its use in the transformer, the rate of degradation,
assessed by the regular dielectric liquid analysis, is key for the differentiation between a normal or an
abnormal behaviour of the transformer. Dielectric liquid condition is also degraded by the presence of
physical and chemical contaminants. Furthermore, some contaminants can increase the rate of
dielectric liquid degradation. Dielectric liquid left in poor condition leads to increased risk of
transformer failure. In extreme cases, it can lead to impairment of the essential cooling of the
transformer. This happens through deposition of decay products (in the case of mineral oil, for
example), or as a result of the increase in the viscosity of the dielectric liquid (in the case of the esters,
for example).
The physical contaminants will be considered first because they are the contaminants usually
encountered in the dielectric liquid of power transformers. They include water, particles and fibres.
Degradation and contamination of the dielectric liquid which can result in the failure of the transformer
may require corrective maintenance prior to having detrimental effect on the active part or result in a
transformer failure. The decision between treating/reclaiming or replacing the dielectric liquid make
take into considering the level of degradation and the related costs.
Physical contamination
In general, there is a correlation between breakdown voltage of the dielectric liquid and its water
content relative to saturation, which is the relative humidity of the dielectric liquid (defined by the ratio
of absolute water content to the maximum water content which can be dissolved in the dielectric liquid
at certain temperature). There is also a correlation between breakdown voltage of the dielectric liquid
and its particle content, even if non-conductive particles, as the interfaces between them and the
dielectric liquid may create preferential paths for the discharges. High moisture, together with high
particle or high fibre content (concentration) in the dielectric liquid, may significantly lower the
breakdown voltage. Usually, dielectric liquid reconditioning processes, essentially formed by filtering,
drying and degassing, are used to remove moisture, gases, fibres and other particles from the
dielectric liquid, thereby restoring the breakdown voltage of the dielectric liquid. More details can be
found in the Technical Brochure 349 [35] regarding the moisture contamination and in [71] regarding
particles.
Water
The majority of the water in a transformer is actually held within the paper insulation (> 99% for a
mineral oil filled transformer and about 95% for an ester oil filled transformer). One unavoidable
source of moisture is its formation as a by-product of paper degradation. Under the influence of heat,
insulation paper will decompose forming carbon oxides, water and furans. When the moisture remains
impregnated in the paper it has a catalytic effect for paper degradation, almost as a chain reaction
(self-induced).
Also, moisture may enter the unit as a contaminant from the atmosphere due to transformer leaks,
cracked seals and from poorly maintained breather systems. The high moisture content in the
insulation paper may increase the risk of bubble formation or cause electrical faults such as partial
discharge.
For minimizing the moisture ingress, free breathing transformers should be fitted with a desiccant
breather that has an oil cup seal (hydraulic seal) or, preferentially, fitted with a mechanical barrier
between the dielectric liquid and the ambient, such as rubber bags, diaphragms or air cells. Gaskets
and oil seals should be replaced before wet oil leaks develop. Check also dielectric liquid seals and
breathers in any other dielectric liquid compartments of the transformer (on-load tap-changer, cable
boxes, passage terminals and external wiring).
In natural ester filled transformers the moisture present in the paper (including the moisture generated
by paper degradation) may migrate to the dielectric liquid, due to its higher saturation point. Hydrolysis
of the dielectric liquid (chemical reaction between the dielectric liquid and water) then results in

106
 Condition assessment of power transformers

consumption of that water, having as by products free fat acids. This is claimed to keep both the
dielectric liquid and paper dry, increasing the neutralization index (acidity), which, in the case of
natural esters, is due to the increase of concentration of long chain organic acids, which are not
harmful for the transformer.
Particles and fibres
Transformers unavoidably retain some quantity of residual paper insulation fibres, rust, particles and
swarf. Despite metallic particles (conductive) are most critical, the non-conductive ones, especially if
associated with high moisture content, are also of concern, with variations according to the type of
dielectric liquid. Metal fragments are always created during transformer manufacture through
component fabrication, core cutting, coil winding, phase assembly, brazing, crimping and tanking
activities and, despite the several stages of decontamination and cleaning, residual content may
eventually be found. Dielectric liquid flowing through transformer cooling systems and dielectric liquid
transfer operations may introduce fibre and particle contamination. Carbonaceous particles can also
be introduced into the dielectric liquid through the pyrolysis of paper. Severely overheated dielectric
liquid or sparking in dielectric liquid usually causes localised carbon coking. Carbon particles can also
be suspended in dielectric liquid as a result of electrical arcing.
Accordingly, the particles and fibres present in dielectric liquid are an amalgamation of different types
and sizes and accumulate from different sources. Limiting the velocity of dielectric liquid flow during
the design of the cooling equipment may reduce the formation of particles. Other actions to minimize
the contamination with particles are: the use of dedicated handling equipment, always suitably clean,
and flushed before use; pre-clean dielectric liquid valves and dielectric liquid hose connection devices;
regularly maintain tap-changer diverter switch; take extreme care during internal inspections and on-
site openings of the tank.
Dielectric liquid properties affected by physical contaminants
The following properties of the dielectric liquid are tested to check whether physical contaminants
have affected the dielectric condition of the dielectric liquid:
Moisture content must be kept low. It is important to use relative moisture content of dielectric liquid,
rather than absolute moisture content, especially when dielectric liquid immersed paper insulation is
present. In assessing the significance of moisture content in dielectric liquid, it must be known that the
moisture content in dielectric liquid changes with the temperature of the transformer. Therefore, when
obtaining the samples of the fluid, the temperature of the sample taken must be recorded. This is
because there is a dynamic moisture exchange between dielectric liquid and paper that is governed by
their temperatures. Sampling the dielectric liquid at different loading conditions and temperatures is an
effective way to evaluate the total content of water inside the tank, both in solid and liquid insulation.
An accurate estimation of the moisture content in solid insulation based on the moisture content in the
dielectric liquid is possible for transformers operating at stable conditions over a period of six months
to one year, having dielectric liquid temperature above 40C, as the equilibrium between the
components would only be reached after a relatively lengthy period of stable temperature. Otherwise it
may be an inaccurate method. If the relative moisture content is high, then the electrical strength of
the dielectric liquid will be impaired and moisture will be absorbed into paper insulation. This reduces
insulation life and increases the risk of dielectric breakdown. For all dielectric liquid types, the
breakdown voltage is proportional to the relative moisture saturation value (that is moisture content
divided by moisture saturation at same temperature).
Breakdown voltage needs to be of sufficiently high value to ensure the dielectric liquid does not
breakdown under electrical stress. The uneven distribution of the moisture inside the tank may also
affect partial discharges.
Particle count must be kept low. Dielectric liquid is filtered to reduce the count of the larger particles
and fibres. Low particle and fibre content are especially important in extra high voltage transformers
and HVDC converter transformers because the presence of such contaminants, especially in the
presence of moisture, can considerably reduce the electrical breakdown strength of the dielectric
liquid.

107
 Condition assessment of power transformers

Chemical contamination
Chemical contaminants are undesirable chemicals that have been introduced to the dielectric liquid by
one of the following mechanisms:
a. intentionally introducing an additive that later proves to have a deleterious effect on the
transformer
b. cross contamination (mixing with an already contaminated dielectric liquid, for example with a
product like PCB or silicone)
c. contact with an incompatible solid material during dielectric liquid storage, or handling, or within
the transformer that introduces a chemical contaminant, or an undesirable chemical effect
(materials such as plasticisers, vulcanised rubber, some plastics, some paints, or the effect of
stray gassing)
d. inadvertently producing undesirable by-product (during dielectric liquid refining, transfers, storage,
reconditioning or reclamation)
e. accidentally adding a foreign liquid when topping up the dielectric liquid level (such as diesel)
f. accidentally adding an incompatible dielectric liquid type (even if the dielectric liquids are miscible
and compatible, concern may arise because of the unintended additive dilution or introduction of
an additive)
g. contamination due to contact with ambient air
h. Furthermore, if a chemical contaminant does not impair the actual condition of the dielectric liquid,
it may still cause concern. For example, in the following situations;
i. if it is PCB, the transformer must now be managed according to local applicable regulations;
j. if it causes stray gassing, it complicates interpretation of dissolved gas analysis;
k. if it has a deleterious effect on other components of the transformer (examples include corrosive
sulphur affecting copper and silver, accelerated ageing of paper insulation, dilution of metal
passivator, lost compatibility with some solid materials, and increased wear of the on-load tap-
changer)
Dielectric liquid properties affected by chemical contaminants
In order to assess whether chemical contaminants have affected the condition of the dielectric liquid,
the following properties of the dielectric liquid are checked. This is to ensure the overall quality of the
dielectric liquid is maintained.
 Dissipation factor (IEC “tan ”) or power factor (IEEE “cos ”) both parameters are directly related
with presence of polar components in the dielectric liquid. They have to be sufficiently low,
depending on the dielectric liquid type, and must not show an excessive increase over time.
Despite IEC using dissipation factor and IEEE using power factor, the difference in their numerical
value is small in the value range of concern. Both measurements will respond to dielectric liquid
deterioration (oxidation by-products) or to chemical contamination, being essential to the dielectric
liquid quality assessment, however, there is no direct correlation between them and the
breakdown voltage. Dissipation factor is less sensitive to high water content, unlike breakdown
voltage, whereas dissipation factor is more sensitive to track polar ageing products then
breakdown voltage.
 DC resistivity has to be sufficiently high, showing variations according to the insulation system,
and it must not show an excessive decrease over time.
 Breakdown voltage needs to be of sufficiently high value to ensure the dielectric liquid does not
breakdown under electrical stress.
 Interfacial tension (IFT) has to be sufficiently high, in the case of mineral oil, to ensure that
chemical contamination or oil oxidation have not introduced polar compounds into the oil. Despite
having no direct relation with any of the dielectric liquid functions, the IFT is a sensitive
measurement for polar contaminants. In dielectric liquids having a slightly polar characteristic,
such as esters in general, this meaning of IFT is lost, since the value shows very little to no
variation when new dielectric liquid is compared with highly oxidized dielectric liquid.
 The neutralization index / total acid number (TAN) must be sufficiently low. Since the parameter
indicates only the total content, not the types, the assessment of the reactivity of the acids
compounds is important. Short-chain acids are more corrosive than long chain ones. High
operation temperature may result in increasing the reactiveness. In general, the acids identified in
mineral oil filled transformers are short chain, very reactive in all temperatures. In synthetic esters
the acids are of intermediary chain length (C7~C9), presenting limited reactiveness at low
temperatures, but becoming a concern for temperatures above 120C~130C. In natural esters

108
 Condition assessment of power transformers

the acids are long chain (C18), presenting very low reactiveness at all temperatures (mostly
limited to 160C~180C). Reactive acids have catalytic effect to paper degradation and may be
critical for other bare surfaces and components. When associated with reduction of interfacial
tension a high value of neutralization index is strong indication of sludge generation.
 Determination of dielectric liquid corrosiveness to copper and silver; deposition of copper sulphide
on the copper conductor and insulating paper, and deposition of silver sulphide on OLTC contacts.
Sulphur corrosion may be introduced into the mineral oil due to contamination by adding additives
based on disulphides, by residual disulphides left in the oil after refining, or by introduction of
undesirable by-products into the mineral oil during oil reclamation (elemental sulphur after
improper oil reclamation); but also may be present as consequence of mineral oil degradation, as
observed in oxidized mineral oils in service (usually uninhibited oils of lower degree of refining).
Such tests only indicate that the dielectric liquid has been affected in a detrimental way. By
themselves, most tests cannot determine the type of chemical (or physical) contamination involved. To
ascertain the actual contaminant present, further investigation of the contamination event and
specialised laboratory testing of dielectric liquid samples may be required. These tests can include
sulphur speciation, liquid chromatography, specific additive content in the dielectric liquid, scanning
electron microscopy and chemical compatibility tests for materials known to be present in the
transformer. Chemical compatibility is an evaluation of the interaction of the dielectric liquid with other
materials but does not assess long term performance of the material, despite being an important
parameter. For the chemical compatibility the control parameters are the dielectric liquid properties,
which may be affected by any reaction of the dielectric liquid with, or even some extraction of
substances from, the tested material.
Dielectric liquid degradation
Despite all the usually performed treatments, dielectric liquid will always contain some dissolved
oxygen, leading to oxidation. Oxidation is the major process of thermal degradation or ageing of the
dielectric liquid under normal operating conditions. It is the rate at which the dielectric liquid oxidises,
that condition assessments are aimed to monitor. As the refining process removes most of natural
protection from dielectric liquids, highly refined liquids require the use of oxidation inhibitor. In free-
breathing equipment, where dielectric liquid remains directly in contact with air, oxidation is relatively
rapid. This is not a recommended use condition for natural ester liquids, as, despite the use of
oxidation inhibitor, they are less stable than mineral oil. Being a thermal degradation process,
oxidation is accelerated by higher temperatures. The presence of metals, especially bare copper,
organo-metallic compounds, or both, may act as a catalyst for oxidation.
For temperatures above recommended limits, oxidation process becomes intensive. In general, the
maximum temperatures indicated at the standards are:
 Mineral oil: continuous service temperature according to IEC 60076-2 [24], which considers usual
the free breathing condition, is defined as 100C. For IEEE C57.12.00, considering essentially
oxygen free applications, the limit temperature is set to 105C. For emergency conditions both
standards limit the temperature to 115C.
 Ester: both IEEE C57.154 [25] and IEC 60076-14 [26] limits the continuous temperature to 130C
and, at emergency condition, 140C.
o In the case of synthetic ester liquid, which is more stable than mineral oil regarding
oxidation, this limitation is mainly caused by the reactivity of the acids.
o For natural ester liquids, as the long chain acids are much less reactive, for an essentially
oxygen-free condition, producer of natural ester liquids accepts a continuous operation
temperature as high as 160C and a limit for emergency condition of 180C.
As thermal degradation takes place, dielectric liquid changes colour. It forms acidic compounds, and
at an advanced stage of oxidation, precipitation of sludge may occur. Precipitation of sludge may
cause the dielectric and, in extreme circumstances, the thermal properties of the transformer to be
impaired.
The ageing mechanisms of mineral oil are complicated. For reference, verify CIGRE TB 526.
Generally, oxygen reacts with certain hydrocarbons by a free radical process. This is a chain reaction
mechanism that generates hydroperoxides. Hydroperoxides are unstable and quickly decompose to
form ketones and water. Ketones oxidise to form carboxylic acids or cleave to make aldehydes. The
presence of hydroxyl groups will result in the production of alcohols and phenols. The final products of
mineral oil oxidation are acids, water and sludge. Most oxidation products will have a negative effect

109
 Condition assessment of power transformers

on the electrical properties of mineral oil. The carboxylic acids that are produced will either dissolve in
the oil or volatilise into the headspace. Dissolved acids may cause damage to the paper and copper
windings, while volatile acids may corrode the top of the unit.
The oxidation mechanism in natural esters process is very similar to mineral oil. The lower stability of
natural esters is attributed to the presence of unsaturated links (double bonds) in the hydrogen-carbon
chain. Accordingly, the use of natural esters usually excludes the use on free breathing transformers,
when the dielectric liquid is in continuous contact with ambient air. Eventual contact or contact with
confined air is acceptable, as the availability of oxygen is limited. Constructions such as corrugated
tank or expandable radiators, gas headspace (which may be nitrogen or ambient air, which will be
confined) or dielectric liquid conservator having a rubber bag or bladder are the most frequent
alternatives suiting for the application of natural ester liquids. Another difference is the behaviour of
the free radicals, which will recombine with the ester molecule, increasing the length of the
hydrocarbon chains. This process is called olygomerization. This results in the long term, in a potential
increase in the viscosity of the dielectric liquid, but, positively, prevents most other side effects of
oxidation (much less concern with the acids, almost zero formation of sludge, no degradation of
breakdown voltage), making an increase in viscosity the primary condition assessment parameter for
the natural esters liquids. The transformer operating temperature is much less critical for natural
esters. This allows the safe operation of natural ester filled transformers at much higher temperatures,
provided the availability of oxygen is limited. The operating temperature, according to international
standards, is limited to 130C, despite special designs have been developed using natural esters for
temperatures up to 160C. The good stability at low oxygen conditions and the low reactivity of the
long chain free fat acids of the natural esters are key for this application. Natural ester has negligible
sludge formation.
Synthetic esters, like the natural esters, have a dielectric liquid molecule comprising of an ester group
(two oxygen atoms in each of its four chains) with the majority of the dielectric liquid molecule being
hydrocarbon. Synthetic esters, unlike natural esters, have saturated chains (no double bonds), leading
to their oxidation stability being higher than that of mineral oil. The synthetic esters also remain stable
at higher temperatures than mineral oil. Again, good control of oxidation is critical, especially in
applications where operating temperatures exceed 130C, when the behaviour of the acids, whose
molecular weights are higher than mineral oil acids, but much smaller than natural esters, must be
carefully evaluated. Synthetic ester has negligible sludge formation.
Degradation of dielectric liquid, by oxidation, can be controlled by oxidation inhibitors. Mineral oils can
be used with and without inhibitors. In recent years, application of inhibited mineral oils is increasing.
Natural and synthetic esters are also used with inhibitors, being especially important for the natural
ester based liquids.
Oxidation inhibitors became important additives for highly refined mineral oils. Natural esters and
synthetic esters also have oxidation inhibitors in their typical formulation. In contrast, silicone liquid
does not. Advances in refining techniques of mineral oils have resulted in higher extraction of their
natural oxidation inhibitors, so synthetic inhibitors have to be added. Whether natural, or added after
refining, these inhibitors added to mineral oils and esters retard oxidation mechanisms by destroying
free radicals and initiators, or by reacting with peroxides form more benign products. Free radical
reactions will take place more quickly in the presence of catalysts such as iron, copper and aluminium.
The effect of these catalysts can be neutralised by metal passivators. Natural inhibitors such as
phenanthroline and some organosulphur compounds act as radical scavengers as well as deactivating
metal catalysts. Some of the phenols, which are produced by oxidation of aromatic compounds, also
act as antioxidants, by donating a hydrogen atom to a free radical, stabilising it and terminating the
free-radical reaction. There are certain phenols which are particularly effective as antioxidants and
these are usually the synthetic additives used in inhibited oil.
The effectiveness of the inhibitors (whether natural or additive) is measured by performing the
oxidation stability test. This test is intended to evaluate the dielectric liquid’s response to pre-defined
accelerated ageing conditions and is not intended to represent actual transformer operation. The
dielectric liquid is subjected to a set amount of excess oxygen (typically by bubbling oxygen or air), a
defined temperature and a measured amount of metal catalyst (typically copper). Although different
standards exist to measure the oxidation stability of the dielectric liquid, importantly they all measure
the secondary effects of the oxidation. The test measures the quantity of by-products formed by the
dielectric liquid degradation. These include the quantity of volatile and soluble acids and their effect on
the electrical properties of the dielectric liquid. In the case of mineral oil the quantity of sludge and
sediment precipitated is also considered, while in natural ester this is replaced by the variation of
viscosity. Different dielectric liquid types may produce different proportions of the by-products. This

110
 Condition assessment of power transformers

needs to be taken into account in making an assessment. Comparing test result values with
acceptance limits given in the standards, only provides information on the relative condition of that
dielectric liquid type. The test does not allow comparisons between dielectric liquid types. There is no
direct correlation between the results and the service life of the dielectric liquid.
There are more sophisticated tests that do allow comparisons between the different dielectric liquid
types. These require specific equipment to allow measurement of the primary effects of the oxidation
reaction and the energy released by the oxidation reaction (exothermic reaction). One such test is
ASTM WK21616, which uses Differential Scanning Calorimetry (DSC) to evaluate the Oxidation
Induction Time (OIT). This is the only test effectively allowing an accurate comparison of different
liquids, as it measures a primary effect of the oxidation, presuming that all insulating liquids are
inhibited with similar amount of inhibitor. Most of other applied tests can just be used as “approved /
reproved” statement.
The oxidation inhibitor content depletion should be monitored and maintained as required.

Dielectric liquid Testing and Dissolved Gas Analysis

The use of dielectric liquid in a transformer provides a valuable way of assessing changes in the
internal condition of the transformer. Removing an dielectric liquid sample for dielectric liquid testing
and dissolved gas analysis works similarly to a human “blood test” for the transformers. The detection
of abnormal patterns of behavior of the dielectric liquid properties offers information about the
dielectric liquid itself, as well as transformer components immersed in that dielectric liquid. These
dielectric liquid tests are essential for the condition assessment, for the predictive maintenance and for
preventing unexpected outages, but they are much more an investigation than just a test.
This chapter however, remains focussed on detecting and assessing the condition of the transformer
dielectric liquid. Other chapters will deal with the use of dielectric liquid testing and dissolved gas
analysis to detect and assess the condition of other transformer components immersed in dielectric
liquid.
In most cases of dielectric liquid testing, it is important to consider not only the measured absolute
values, but also the trending. Also relevant to an assessment is any correlation between rates of
change, or variations, with how the transformer is operated or maintained. Some dielectric liquid
properties have defined continuous service limiting values, which are strongly recommended to be
followed. However, when a relevant variation of the properties is identified, shorter sampling intervals
and even some early investigations may be needed to intercept serious consequences or make timely
interventions. More details may be found in the references IEEE C57.104 [10] and IEEE C57.155 [30],
respectively for mineral oil and ester oils.
The integrity of dielectric liquid test results and dissolved gas analysis is dependent upon appropriate
dielectric liquid sampling procedures and operator competency. Rigid criteria, extremely careful
procedures, use of quality hardware are key aspects of proper dielectric liquid sampling. Dissolved
gas analysis measures gas content in “parts per million – ppm” or even “parts per billion – ppb”, so it
may be affected by the tiniest error in the sampling procedure.
Dielectric liquid tests
The suggested grouping of tests differs from Table 1 IEC 60422 [42], reflecting the recommendations
of this working group. This technical brochure is informative, not normative. The minimum suggested
surveillance checks for degradation of transformer dielectric liquid are highlighted in Group 1 in the
following table.

111
 Condition assessment of power transformers

GROUP 1 GROUP 2 GROUP 3

Minimum set to monitor the Specialised dielectric liquid quality Suitability for type of transformer
dielectric liquid for continued service Evaluate dielectric liquid for Environmental/operational
continued use compliance
Acidity Oxidation Stability PCB Content
Colour Sediment & Sludge Corrosive Sulphur
Water Content Particle Count Aromatic Content
Breakdown Voltage Flash Point Biodegradability
Dissipation Factor Pour Point Toxicity
DC Resistivity Density
Interfacial Tension* Viscosity*
Inhibitor Content
* For natural ester oils, interfacial tension does not offer relevant information regarding oil degradation,
being typically replaced by viscosity measurement. Interfacial tension may be replaced by viscosity in
the Group 1 tests.

Group 1 Dielectric liquid Tests are commonly performed on the dielectric liquid of transformers, for
condition assessment, throughout their service life. All the tests listed in this group provide useful
information on the dielectric liquid condition and represent the minimum set of tests for condition
assessment.
Group 2 Dielectric liquid Tests are infrequently performed on in-service dielectric liquids. Inhibitor
content is used to monitor inhibited dielectric liquid or reclaimed dielectric liquid. The oxidation stability
and sediment/sludge tests are useful to check dielectric liquid following mixing, treatment or suspected
contamination. The particle count test is more often performed for EHV and HVDC converter
transformers. The verification of the oxidation inhibitor content in natural esters may be included in
Group 1.
Group 3 Dielectric liquid Tests are performed on in-service dielectric liquids by exception. They are
used to prove environmental compliance, or to detect whether the dielectric liquid’s constituent base,
or contamination, is harmful to the transformer or harmful to personnel. They are more often
performed before commissioning, or during dielectric liquid handling or intrusive maintenance, or to
evaluate the side effects of a dielectric liquid treatment. The corrosive sulphur test is an example of a
test needed to evaluate a dielectric liquid’s risk to the transformer.
For more information on these tests, see the Annex D and the bibliographical references [5] [42] [41]
[71] [72] [73] [74] [75] [76].
The results of dielectric liquid tests can also correlate for certain dielectric liquid problems. The matrix
that follows serves to illustrate how any, or all, of the dielectric liquid test groups and dissolved gas
analyses may be used to detect particular dielectric liquid problems. This matrix applies only to the
behaviour of mineral oil. It should be noted that these dielectric liquid problems may also be
consequential to a failure, degradation or maloperation of another component of the transformer.
Nevertheless, any detrimental effect on transformer dielectric liquid should be detected as early as
practicably possible.

112
 Condition assessment of power transformers

Type of Dielectric Liquid Problem

Corrosive Sulphur
Two liquids Incompatible (top up or mixed)
Over-processing (by liquid reconditioning)
Oxidation
Contamination by foreign liquid
Contamination by fine particles or fibres
Contamination by gross water ingress
Incompatible Paint, Solids or Additive
Excessive Gas Content (N2, O2)

Diagnostic Dielectric liquid Test

Acidity  
Colour and Appearance     
Water Content (and Relative Saturation)  
1 Breakdown Voltage     
Dissipation/Power Factor    
DIELECTRIC LIQUID TEST GROUP

DC Resistivity     
Interfacial Tension    
Oxidation Stability   
Particle Count 
2 Inhibitor Content  
Flash Point   
Sediment & Sludge    
Corrosive Sulphur   
3 Aromatic Content  
Passivator Content 
Dissolved Gas Analysis    

Failure mode assessment

The assimilation of dielectric liquid test data, into an overall condition assessment of the dielectric
liquid component, will differ in criteria depending on which of the four dielectric liquid types were used.
To establish a consistent approach to the assessment for each dielectric liquid type, and for that
matter any other component of the transformer, five levels of dielectric liquid condition are defined.
They are defined in such a way that each level of condition should influence the overall transformer
condition to the same degree, irrespective of which component was being considered. Together, the
five levels provide a transition or stepping down in dielectric liquid condition. The transition starts with
dielectric liquid in ‘as new condition’ and finishes with dielectric liquid degraded to ‘unserviceable, or
end of life, condition’. When dielectric liquid exhibits the first perceptible signs of degradation, it
becomes level two. As dielectric liquid degradation advances, its condition presents a flag for optimal
condition based intervention. If left untreated, its condition then affects other transformer components.
Dielectric liquid condition can be assessed using the tests listed above, supplemented by family
history, results of inspections and inductive reasoning as appropriate and in conjunction with the
assessment scoring matrix. The tables in Annex A can also be used as a starting point when
assessing a transformer but users should consider if the values are suitable for their fleet of
transformers, operating conditions, maintenance practices, and the time scales used in their scoring
matrix. User should refer to Chapters 1 to 7 before assessing their transformers.

113
 Condition assessment of power transformers

Scoring Matrix for insulating liquids

The generic scoring matrix illustrated in 2.3 has been adapted to more easily allow the assessment of
insulating liquids.

Scoring matrix for dielectric liquids

F De-energize as soon as possible. Don’t return to service until problem is repaired.

E Very poor condition – high likelihood of failure. May expected damages on the active part or other
components due to the dielectric liquid’s poor condition. May also lead to a transformer failure.
The causes for the dielectric liquid degradation should be repaired e.g. leakage.
Reconditioning/reclaiming or replacing the dielectric liquid as soon as possible is recommended.
D Poor condition. The causes for the dielectric liquid degradation should be repaired e.g. leakage.
Reconditioning/reclaiming or replacing the dielectric liquid should be considered within the short
term.
Detrimental effects on the active part may take place e.g. acceleration of degradation rate of
insulating paper due to acids or accumulation of sludge.
C Acceptable condition with significant signs of degradation. Triggering an investigation into the causes
for degradation. After repairing the issue consider reconditioning/reclaiming or replacing the dielectric
liquid.
B Good condition. Some signs of degradation are evident. Reduce the sampling interval to monitor the
degradation rate.
A Minimal signs of degradation. As new condition

114
 Condition assessment of power transformers

Annex A Transformer Condition Assessment Tables

This Annex provides tables of diagnostic information that is formatted to allow assessment using the
methods described in this technical brochure. The information is from IEC & IEEE guides, CIGRE
publications, CIGRE experts, and other industry experts. Users can use these values as a starting
point when assessing a transformer but should consider if the values are suitable for their fleet of
transformers, operating conditions, maintenance practices, and the time scales used in the scoring
matrix.

Users must note that although the tables provide six columns for assessment information (labelled
A-F) in accordance with the scoring matrix used throughout the technical brochure, information is not
provided in every column for every assessment method covered in the table. In some cases, it is not
appropriate to provide six levels of information e.g. where the test or diagnostic method normally
provides a ‘binary result’ (yes/no or good/bad). In other cases, the working group did not have
adequate information to provide suitable values or considered that appropriate assessment values
would vary with the age or design of the equipment or the operating environment and conditions.
Users should consider their experience with their own equipment to determine suitable values where
required for an assessment.

115
 Condition assessment of power transformers

Table A-1: Dielectric Condition Assessment

Applicable to
Degradation / Condition which
Component Test Unit F E D C B A Reference
Failure mode Indicator insulation
liquids?1
Insulation C57.152
Core Ground DC Resistance MOhms MO, NE, SE <1 1 - 10 11 - 25 26 - 50 51 - 500 > 500
resistance Z1
Core

C57.104
Non-
Arcing Acetylene DGA ppm MO, NE > 50 36 - 50 11 - 35 1 - 10 <1 IEC 60599
detectable
C57.155
C57.106
Insulating Liquid

Aging, Moisture, Dissipation IEC 60422

DF @25 C % MO, NE, SE See insulation liquids tables
contamination factor C57.147
IEC 62975
C57.104
Partial Discharge Hydrogen DGA ppm MO, NE, SE > 3000 1800 - 3000 700 - 1799 100 - 699 30 - 99 < 30
IEC 60599
C57.106
Karl Fisher
Moisture Content of Moisture IEC 60422
Moisture ppm MO, NE, SE See insulation liquids tables
Liquid Content C57.147
Content
IEC 62975
Elevated Moisture, CHL, CH, CL of
PF/DF % MO > 2.5 2.1 - 2.5 1.6 - 2.0 1.1 - 1.5 0.5 - 1.0 < 0.5 C57.152
Winding Insulation

aging, contamination transformer

Elevated Moisture, CHL, CH, CL of
PF/DF % NE >4 3.0 - 4.0 2.0 - 3.0 1.5 - 2.0 1.0 - 1.5 <1
aging, contamination transformer
% Variance
Displaced winding CHL, CH, CL of
Capacitance from MO, NE, SE >8 7-8 6 - 6.9 5 - 5.9 0.5 - 4.9 < 0.4 C57.152
caused by aging transformer
factory
Elevated Moisture, Estimated CIGRE TR
Major

DFR % wt / wt MO, NE, SE >5 3.1 - 5.0 2.1 - 3.0 1.5 - 2.0 1.0 - 1.5 <1
aging, contamination Moisture 414
Induced/
Apparent C57.152
Partial Discharge Applied pC MO, NE, SE > 5000 2001 - 5000 751 - 2000 501 - 750 250 - 500 < 250
Charge IEC 60076-3
Voltage Test
Insulation ratio of
degradation by Polarization Polarization resistance C57.152
MO < 1.0 1.0 - 1.1 1.11 - 1.25 1.26 - 1.5 1.51 - 2.0 >2
moisture/contaminat Index Index (PI) at 10 min Z2
ion to 1 min
Plot of
Winding turn strand Frequency Leakage Leakage
Minor

to strand short circuit response of Reactance vs reactance MO, NE, SE Slope of leakage reactance to frequency should be same for all phases
or open circuit stray loss Frequency vs
frequency

116
 Condition assessment of power transformers

Applicable to
Degradation / Condition which
Component Test Unit F E D C B A Reference
Failure mode Indicator insulation
liquids?1
Winding turn to turn Excitation Excitation C57.152
Amps MO, NE, SE Pattern - Depending on core configuration there will be a high-low-high or a low-high-
short circuit current Current Z4
low pattern.
% Variance
Winding turn to turn
Ratio TTR from MO, NE, SE > 1.5 1.26 - 1.5 1.1 - 1.25 0.51 - 1.0 0.4 - 0.5 < 0.4 C57.152
short circuit
nameplate
Thermal aging of
DP Value DP Test number MO, NE, SE < 200 200 - 300 301 - 400 401 - 500 ≥ 500
solid insulation
Z1 *put comments in this field relating to pattern/consistency
z2 Ratio of resistance at 10 minutes to 1 minute
z3 Slope of leakage reactance to frequency should be same for all phases
z4 Depending on core configuration there will be a high-low-high or a low-high-low pattern.
Z5Generally inferred from Furanic compound analysis, loading history, temperature, age, accessible paper samples. For inference from Furans, refer IEEE C57.140-2017.
Note 1 – MO = Mineral Oil, NE = Natural Ester liquid, SE = Synthetic Ester liquid

117
 Condition assessment of power transformers

Table A-2: Thermal (Solid Insulation) Condition Assessment

Applicable to
which
Failure Mode Condition Indicator Test / Criteria Unit F E D C B A Reference
insulation
liquids?1
IEEE C57.104-
Carbon Monoxide / Cellulose
CO ppm MO, NE < 5000 < 2500 571 -1400 351 - 570 < 350 2008
Covered by Paper
Hot Metal Fault

Gassing + Hot Metal Gassing

IEEE C57.155

IEEE C57.104-
Carbon Dioxide / Cellulose 4001 -
CO2 ppm MO, NE < 35000 < 17500 2500 - 4000 < 2500 2008
Gassing + Hot Metal Gassing 10000
IEEE C57.155

IEEE C57.104-
Cellulose Aging CO2/CO Ratio MO, NE <3 >3 2008
IEEE C57.155
Accelerated Cellulose Aging

Inferred from Furans

Degree of Polymerization or Methanol and MO, NE < 100 > 100 > 300 > 500 > 700 > 800 CIGRE TB 323
service history
IEEE
Winding Hot Spot
Design Calculation °C MO > 160C < 160C < 140C < 130C < 120 < 110C C57.91.2011
Temperature
CIGRE TB 659
Moisture in Oil MO, NE, SE See insulation liquids tables IEEE C57.152
Interfacial Tension MO, NE, SE See insulation liquids tables
Acidity MO, NE, SE See insulation liquids tables
Moisture in Cellulose Oil Equilibrium curves
Dew Point
Note* The limits given apply
% MO, NE, SE > 3.0% < 3.0% < 2.0% < 1.5% < 1.0% < 0.5 %
for cellulose accelerated
DFR or FDS Test
aging - not for increased
dielectric risk
Winding Insulation Power
% MO > 2% > 1% < 1% < 0.7% < 0.5% < 0.3% IEEE C57.152
Factor
Thermal aging of solid
DP Value MO, NE, SE < 200 200 - 300 301 - 400 401 - 500 ≥ 500
insulation
Hot Metal Gassing (if IEEE C57.104-
indicated by Duval Triangle Ethylene ppm MO, NE > 1000 < 1000 < 500 < 200 < 100 < 50 2008
or IEC Ratio Scheme) IEEE C57.155
% increase over
Winding DC Resistance All 10% 5% 4% 3% 2% 1% IEEE C57.152
factory tested values
Core Overheating Design Calculation °C MO, NE, SE < 170 < 150 < 130 < 100 X1

118
 Condition assessment of power transformers

Applicable to
which
Failure Mode Condition Indicator Test / Criteria Unit F E D C B A Reference
insulation
liquids?1
IEEE C57.104-
Cellulose Gassing + Hot
CH4 ppm MO, NE > 2000 > 1500 < 1200 401 – 1000 121-400 < 120 2008 IEEE
Covered by Paper
Hot Metal Fault

Metal Gassing
C57.155
Cellulose Gassing + Hot
C2H4 ppm MO, NE > 700 > 400 < 250 101 – 200 51 – 100 < 50 X2
Metal Gassing
Duval Triangle
Cellulose Gassing + Hot
C2H6 ppm MO > 800 > 500 < 200 101-150 66-100 < 65 5 Type ‘C’, ‘T2’
Metal Gassing
and ‘T3’
Calculated Bubble
Evolution
Temperature minus
Moisture in cellulose versus IEEE
Bubbling

Calculated Winding
Winding Hot Spot °C MO, NE, SE <0 < 10 C57.91.2011 +
Hot Spot
temperature curves. CIGRE TB 349
Temperature for
Defined Overload
Condition
X1 - IEEE Transactions on Power Delivery, Volume 17, Issue 4, Oct. 2002 Page(s)991 – 995
X2 - Duval, Michel; Calculation of DGA Limits, DEIS Volume. 24, No.5 September/October 2008
Note 1 – MO = Mineral Oil, NE = Natural Ester liquid, SE = Synthetic Ester liquid

119
 Condition assessment of power transformers

Table A-3: Mechanical Condition Assessment

Applicable
Failure Degradation to which
Condition Indicator Test Unit F E D C B A Reference
Mode Mechanism insulation
liquids?
Residual
Loose Good agreement in
Clamping pressure Pressure All Unequal phase pressures
Clamping pressures
Survey
%
Single 1% - 2%
variation ≤ 1% change from FAT or
Impedance phase All > 2% change from FAT change from TB 445
between SAT
reactance FAT
phases
Winding
Signature differences to
Displacement Attenuation versus IEC 60076-18
FRA All FAT or SAT signatures in Good agreement
frequency TB 342
the interpretation range
Mechanical

Between windings All Poor alignment Good alignment

Visual
Within windings All Poor alignment Good alignment
Condition of blocks, Loose or displaced
Visual/ No loose or displaced
spacers, sticks within All blocks, spaces or sticks
manual blocks, spacers, or sticks
winding assembly within winding assembly
Coil clamping devices,
Loose or displaced
Loose platforms or blocks next to
Visual/ clamping devices, No loose or displaced
Clamping top and bottom frames All
manual platforms or support blocking
that support or compress
blocks
the phase assembly
Tonal Tonal difference No tonal difference
Change in tone All
Checks identified identified

120
 Condition assessment of power transformers

Table A-4: Bushing Condition Assessment

Affected
Bushing
Failure Mode Parameter/ Criteria Unit F E D C B A Reference
type
Investigative Test
Oil level not
Oil leakage OIP Visual inspection Correct Oil level % 10% full 20% full 30% full 50% full IEEE C57.152
visible
Colour of oil in
Light Clear & without
sight glass (by Colour change of oil in Black/ Dark Dark and/ or Yellow/ Light IEC 60422,
OIP Visual inspection Brown/ visible
Visual sight glass Black turbid * Dark Yellow Yellow Table-5
Brown contamination
Inspection)
Colour of oil in Bright Pale
Colour change of oil in Dark Brown Amber (4.0 Yellow (1.0-
sight glass (by OIP Visual inspection yellow (2.5 Yellow Clear (0.0 to 0.5) IEEE C57.152
sight glass (5.5 to 8.5) to 5.5) 2.5)
lab test) to 4.0) (0.5-1.0)
Hydrogen (H2) ul gas / l oil >140 ≤140 IEC 61464
Methane (CH4) ul gas / l oil >40 ≤40 IEC 61464
Ethylene (C2H4) ul gas / l oil >30 ≤30 IEC 61464
DGA Oil OIP DGA in Bushing Oil Ethane (C2H6) ul gas / l oil >70 ≤70 IEC 61464
Acetylene (C2H2) ul gas / l oil >2 ≤2 IEC 61464
Carbon monoxide (CO) ul gas / l oil >1000 ≤1000 IEC 61464
Carbon dioxide (CO2) ul gas / l oil >3400 ≤3400 IEC 61464
Overheating in OIP/RIP/ Temperature increase
Thermal scanning ˚C >40 31 - 40 21 - 30 11 - 20 1 - 10 IEEE C57.152
bushing top RIS/RBP above ambient
Temperature difference
Overheating in OIP/RIP/ CIGRE WG
Thermal scanning of similar and Δ˚C ≥35 10 ≤ t < 35 5 ≤ t < 10 <5
bushing top RIS/RBP A2.43
comparable bushings
pD activities in
H2/CH4 (pD) >13 ≤13 IEC 61464
condenser core
Overheating in
C2H4/C2H6 (Thermal
the condenser >1 ≤1 IEC 61464
fault)
core
OIP DGA Ratio
pD activities in
C2H2/C2/H4 (pD) >1 ≤1 IEC 61464
condenser core
Overheating in
CO2/CO (Thermal fault
the condenser >20 or <1 1 – 20 IEC 61464
in paper)
core
IEEE
Contamination/ OIP/RIP/ DF/PF/Dielectric Initial value Initial value Initial value Initial test (rating
tan delta increase % C57.152,
layers RIS frequency response *3 *2 *1.5 plate value)
CIGRE TB445
shorting/excess
DF/PF/Dielectric Initial test (rating
ive moisture RBP tan delta increase % >0.6 0 - 0.6 CIGRE TB 445
frequency response plate value)

121
 Condition assessment of power transformers

Affected
Bushing
Failure Mode Parameter/ Criteria Unit F E D C B A Reference
type
Investigative Test
IEEE
OIP/RIP/ Change of Capacitance Initial test (rating C57.152,
C1 measurement % >10 5.1 - 10 2.1 – 5 0–2
RIS (C1) plate value) CIGRE TB445
Y1
Damaged sheds OIP/RIP/
Visual inspection Defined by supplier Major Minor No damage
on insulator RIS/RBP
Corrosion due
to poor OIP/RIP/ Visual inspection
Burning marks Many Few Clean
installation of RIS/RBP voltage tap
test tap cover
Y1 – Capacitance of RPB bushings is not stable in service due to oil penetration into the condenser body.
This condition measure is much less significant for RBP condition assessment compared with other bushing types.

122
 Condition assessment of power transformers

Table A-5: Cable Box Condition Assessment

Failure Affected Parameter/ Type of

Description Unit F E D C B A
Mode Investigative Test Cable Box
Traces of main oil leaking through
Correct oil level in cable
Correct oil level External Visual inspection Oil filled gland plate or covered openings on
box, and/or no oil leaks
Oil leakage box
Evidence of main oil leaking Evidence of main oil leaking
Internal Visual Inspection All types
through bushings through bushings
Changes in Please see chapter 12 and
Oil analysis Oil filled
oil quality Table A8 for oil assessment
Moisture in oil test DGA Oil filled % Saturation
Moisture in Oil Breather condition Correct colour of
Moisture Air filled Discolouration desiccant
Inspection desiccant
ingress
Dry / Corrosion Internal Visual Inspection All types
Dry / Corrosion Internal Visual Inspection All types
Internal Visual Inspection Air filled
Internal Visual Inspection Pitch filled
Condensati
Dry / Corrosion Internal Visual Inspection Gel filled
on
Compound
Internal Visual Inspection
filled
Tracking External Visual inspection All types
Internal
corona powder deposits on cable
Internal Visual Inspection Visual
termination insulation,
Inspection
Partial RFI RFI
discharge Acoustic
pD-activity
Acoustic measurement measureme
nt
Hydrogen and methane
DGA Oil filled ppm
gases present
Wall Temperature increase above
Wall Temperature Thermal scanning All types °C
ambient
Overheatin
Internal inspection (if
g Overheated components All types °C Overheated components
inspection practicable)
Heating gases present DGA Oil filled ppm
Cracks
Shrinkage Cracks Visual inspection Pitch filled Pitch normal
in Pitch
Visual inspection, sample Homogenous, natural
Miscellaneo Gel filled
for physical tests colour and composition
us
Composition Cable
Visual inspection, sample Homogenous, natural
compound
for physical tests colour and composition
filled

123
 Condition assessment of power transformers

Table A-6: OLTC Condition Assessment

Failure Sub- Degradation Condition

Test Units F E D C B A Reference
Mode component(s) Mechanism Indicator
OLTC serviced
OLTC serviced
by a technician
Training/ by a technician
Workmanship without the
Qualification with the OEM
Thermal: Overheating

Diverter
OEM
blocked in- certification
Diverter switch certification
between
or Load Selector Time or Maintenance
positions ->
switch - Resistor OEM Specs Visual operation 100% late 80% late 60% late 40% late 20% late according to
Transition
type count schedule
resistors
Transition Vibro-Acoustic Outside Within OEM
inserted ms
time or DRM OEM Specs Specs
Temperature OLTC tank
Thermography
differential temperature
Toltc >= Tmain
Thermography Relative Toltc >> Tmain lower than Main
Temperature
(bolt-on OLTC tank
differential
only) temperature
Stable gas
Significant
Contact Significant pattern - Stable gas
Thermal: Coking

deviation in
Main current degradation - Duval deviation in Slight pattern - In
Gas Pattern gassing IEEE
carrying > Carbon DGA Triangle or gassing pattern differences agreement with
Trending pattern C57.139
contacts (OLTC deposits -> other Ratios over a few with the benchmark LTC
over a few
or DETC) Thermal months benchmark population
weeks
runaway population
Operating
history - Time
Time Years > 30 > 10 >3 >1 <1
elapsed since
last operation.
Carbon Time or Maintenance
Dielectric: Tracking and

deposit and OEM Specs Visual operation 100% late 80% late 60% late 40% late 20% late according to
Insulation parts
moisture count schedule
in Diverter
Flashover

ingress in
compartment Water
insulation Karl Fischer ppm > 50 < 50 < 40 < 30 < 20 IEC 60814
content
material
Insufficient
IEC60156
Oil dielectric Breakdown Voltage kv < 20 > 20 > 25 > 30 > 40 > 50
2.5mm
properties
Fixed and Positive and
Corrosiv
e Sulfur
Chemic

mobile contacts Flashover in presence of

Corrosivity test Negative DIN51353
al:

in the main main tank silver contacts

transformer in main tank

124
 Condition assessment of power transformers

Failure Sub- Degradation Condition

Test Units F E D C B A Reference
Mode component(s) Mechanism Indicator
tank (OLTC or ASTM
Hotspot on
DETC) 1275-15
tap selector Corrosivity test Positive Negative
Copper
switch
strip
Motor drive and
Deviation from Agreement with
Mechanical:
Incomplete
operation

linkage
benchmark or benchmark or
Motor trip Torque Motor current A
Tap selector over OEM within OEM
(OLTC or DETC) specification specification

Wrong
Mechanical:

operation

direction, run
Incorrect

Age of the
Motor drive through, Years > 50 <=50 <30 < 20
Motor Drive
intermittent
stops, etc.
Bad
Mechanical: Wrong tap

synchronism Vibro- Acoustic Outside OEM Within OEM

Synchronism OEM Specs
motor drive or Manual Specs Specs
indication

to LTC
Motor drive and
Time or Maintenance
linkage Wear of
OEM Specs Visual Operation 100% late 80% late 60% late 40% late 20% late according to
linkages
Count schedule
Erroneous Age of the
Years > 50 < 50 < 30 < 20
position signal Motor Drive
Time or Maintenance
Linkage
OEM Specs Visual operation 100% late 80% late 60% late 40% late 20% late according to
Mechanical: Drive shaft

seizure
count schedule
Incorrect
failure

Motor drive and coupling ->

OLTC serviced OLTC serviced
linkage Motor drive
Training/ by a technician by a technician
forcing Workmanship
Qualification without an OEM with an OEM
through a
certification certification
mechanical
stop
Internal leaks: Diverter Diverter
Migration of compartment or compartment or
Mechanical: Leaks

gasses and barrier board barrier board

LTC Oil carbon made of made of
Material Design Type
compartment particles into cellulosic fiberglass or
main tank or material other insulating
low/high (Bakelite, composite
levels of oil in Formica, etc…) material

125
 Condition assessment of power transformers

Failure Sub- Degradation Condition

Test Units F E D C B A Reference
Mode component(s) Mechanism Indicator
the LTC Oil level Occurrences
Count >5 >3 >1 1 No occurrences
compartment alarms within 5 years
Presence of
Typical
LTC typical Typical LTC
Typical LTC Typical LTC gas LTC gas No detectable
dissolved gas pattern
DGA ppm gas pattern + pattern + C2H2 > pattern + LTC typical
gasses, in + 0 < C2H2 <
C2H2 > 1000 500 C2H2 > gasses
particular 100
300
C2H2
Major leaks
visible down
to ground - Significant leaks Minor
Visual or
Events of low visible, some leaks - Do
External leak Rate Maintenance No visible leaks
level alarm - may reach not reach
record
Topping off ground ground
the oil level
required

126
 Condition assessment of power transformers

Table A-7: Cooler/Radiator Condition Assessment

Condition
Compon Degradation Condition
Failure Mode Assessment Units F E D C B A
ent Mechanism Indicator
techniques
More than
Cooling One fan
Inadequate air one Fan
Fan or control failure / Elevated Circuit Failure / All fans
flow (elevated oil Visual inspection Failure
fan in wrong direction temperatures complete Wrong working
temperature) /wrong
failure Direction
direction
Measurement
difference of inlet
Failure of pump / pump Elevated and outlet oil
Inadequate oil ○
C ΔT < 30% ΔT < 50% ΔT < 60% ΔT < 80% ΔT < 90% ΔT ≥ 90%
in wrong direction / valve temperatures temperature.
flow (elevated oil
closed incorrectly / Compare between
temperature)
blockage coolers
Oil flow relay, no Oil flow indication / No Oil Restricted Oil Normal Oil
Coolers/Radiators

pump current measurement Flow Flow Flow

Insufficient heat
Dirty radiators and Blocked cooling Restricted air Normal air
transfer (elevated Visual inspection Poor air flow
cooling fans ducts flow flow
oil temperature
Low oil level
Oil level No oil level
(transformer Visual inspection Oil level alarm
gauge/alarm alarm
alarm/trip)
H2O > H2O <
Oil leaks
Elevated levels of 20/400ppm 20/400ppm
Moisture and
moisture and DGA etc. ppm (oil/ester) (oil/ester)
oxygen ingress
oxygen O2 > 5000 O2 < 5000
ppm ppm
Higher oil and
Elevated ambient winding Unexpected
Unexpected No WTI/OTI
air/cooling water inlet temperatures Visual inspection WTI/OTI
OTI/WTI alarms alarms
temperatures (elevated oil alarms
temperature)
Defective or
deteriorated Oil level drop,
Oil on ground or Moderate
Tank

Oil leaks gaskets and seals, Visual inspection Oil level alarm contaminatio Minor leaks No leaks
wet spots on tank leaks
incorrect n issues, etc.
tightening

127
 Condition assessment of power transformers

Condition
Compon Degradation Condition
Failure Mode Assessment Units F E D C B A
ent Mechanism Indicator
techniques
Heavy polluted
site conditions / Oil leaks and
Corrosion and improper Rust, corrosion,
improper handling Visual inspection weakening of Minor rust No rust
painting stains etc.
during installation the tank,
or maintenance
Failure of
Incorrect operation of Failed or mis- Drifting in
components / Failure / Out of
components Routine checks operating alarm or trip No defects
Drifting of alarm calibration
components settings
Components

or trip settings
Dirty with
Moderately
Damage / Elevated major particle
Cracks / high dirty, with
Lightning Arresters Contamination/ temperatures, build up /
Routine checks temperature some No defects
Lightning strikes/ power factor or moderate
difference particle
Misapplication leakage current temperature
build-up
difference

128
 Condition assessment of power transformers

Table A-8: Mineral Oil Condition Assessment

Affected
Parameter/ Voltage Class of
Failure Mode Unit F E D C B A Reference
Investigative the equipment
Test
Corrosive Less than the
All Corrosive IEC 60422
Sulphur detection limit.
Flash Point All ºC <122 130 to 135 >135
IEC-60422
Particle Count Annex B
All >18/15 16/13 to 17/14 11/8 to 15/12 9/6 to 10/7 Up to 8/5
(ISO Class)
Table B.1
Particle Count From table B1
>130 000 /
IEC 60970 All 130 000 / 100ml 32 000 / 100ml 1 000 / 100ml 250 / 100ml of IEC
100ml
above 5mm 60422:2013
Physical Contamination

<72.5 >40 30 - 40 <30 <20

Moisture in Oil 72.5-170 ppm >30 20 to 30 <20 <10 IEC 60422
>170 >20 15 to 20 <15 <10
BDV <72.5 <30 30 to 40 >40 >55
(IEC 60156 72.5-170 kV <40 40 to 50 >50 IEC 60422
>60
2.5mm) >170 <50 50 to 60 >60
Chemical Contamination

DDF (Tan δ) <72.5

>0.05 0.1 to 0.5 ≤ 0.015
@90°C 72.5-170 <0.1 IEC 60422
(IEC 60247) >170 >0.2 0.1 to 0.2 ≤ 0.010
<72.5
<0.2 3 to 0.2 >3
Resistivity 72.5-170 GΩm > 60 IEC 60422
>170 <3 3 - 10 >10
<72.5 >0.30 0.15 - 0.30 <0.15
Degradation

Acidity 72.5-170 mg KOH/g >0.20 0.10 - 0.20 ≤ 0.03

<0.10
>170 >0.15 0.10 - 0.15 IEEE C57.106,
All (Inhibited) mN/m <22 22 - 28 >28 IEC 60422
IFT All (Un- >35
mN/m <20 20 - 25 >25
inhibited)
Colour (by Lab Dark Brown (5.5 Amber (4.0 - Bright Yellow Yellow (1.0 - Pale Yellow (0.5
All Clear (0.0 - 0.5) IEEE C57.152
testing) - 8.5) 5.5) (2.5 - 4.0) 2.5) - 1.0)
Colour (by Clear & without
Black/Dark Dark and/or Light Yellow/Dark IEC 60422,
Visual All Light Yellow visible
Black turbid * Brown/Brown Yellow Table-5
Inspection) contamination *
<40% of original 40% to 60% of >60% of original >90% of original
Inhibitor All % IEC 60422
value original value value value
Sediment and
All % >0.02% by mass <0.02% by mass IEC 60422
Sludge

129
 Condition assessment of power transformers

Affected
Parameter/ Voltage Class of
Failure Mode Unit F E D C B A Reference
Investigative the equipment
Test
Environmental
PCB All mg/kg >2 <2 IEC 60422
Hazard
Note:
1. * Values referred from
2. ** Values referred from IEC 60422, Table-3

130
 Condition assessment of power transformers

Table A-9: Natural Ester Condition Assessment

Affected Voltage Class

Failure Mode Parameter/ of the Unit F E D C B A Reference
Investigative Test equipment

Slightly IEC 62535 or

Corrosive Sulphur All °C Non-corrosive
Corrosive ASTM D1275B

<72.5
(ASTM D92 or
Fire Point 72.5-170 °C <300 >300
ISO 2592
>170

IEC-60422
Particle Count (ISO
All >18/15 16/13- 17/14 11/8-15/12 9/6-10/7 Up to 8/5 Annex B Table
Class)
B.1

From table B1
Particle Count IEC >130 000 /
Physical Contamination

All 130 000 / 100ml 32 000 / 100ml 1 000 / 100ml 250 / 100ml of IEC
60970 above 5mm 100ml
60422:2013
Moisture in Oil <72.5 >550 450 - 550 385 - 450 300 - 385 <300
ASTM D1533
(ASTM D1533, 72.5-170 ppm >450 350 - 450 240 - 350 150 - 240 <150
or IEC 60814
IEC60814) >170 >400 300 - 400 200 - 300 100 - 200 <100
<72.5 <30 30 - 40 >40 >50
BDV (IEC 60156
72.5-170 kV <40 40 - 50 >50 >60
2.5mm)=
>170 <50 50 - 60 >60 >60
Chemical Contamination

DDF (Tan δ) @25°C

All % > 3% 1–3 <1 ≤0.5 IEEE C57.147
Degradation

(ASTM D924)

Acidity (ASTM D974,

All mg KOH/g >1 0.3 - 1 0.15 - 0.30 0.06 - 0.15 ≤0.06
IEC 62021-3)

131
 Condition assessment of power transformers

Affected Voltage Class

Failure Mode Parameter/ of the Unit F E D C B A Reference
Investigative Test equipment
Colour (by lab
All >1.5 ≤1.0 ≤0.5
testing)
Clear & without
Colour (by Visual Black/Dark Dark and/or Light Brown / Yellow / Dark Light Green /
All visible
Inspection) Black turbid Brown Yellow Light Yellow
contamination
Viscosity
>50% of >30% of >20% of >10% of >5% of
(ASTM D445, All <50 cSt IEEE C57.147
original value original value original value original value original value
ISO3104)
<40% of original <60% of original <90% of original >90% of original Contact the
Inhibitor All %
value value value value manufacturer
Sediment and Less than the
All % >0.02% by mass <0.02% by mass
Sludge detection limit.
Environmental
PCB All mg/kg >2 <2
Hazard

132
 Condition assessment of power transformers

Table A-10: Synthetic Ester Condition Assessment

Affected
Voltage Class
Parameter/
Failure Mode of the Unit F E D C B A Reference
Investigative
equipment
Test
Corrosive Slightly
All °C Non-Corrosive IEC 62535 or ASTM D1275B
Sulphur Corrosive
<72.5
Fire Point 72.5-170 °C <300 >300 ASTM D92 or ISO 2592
>170

Particle Count
All > 18/15 16/13 - 17/14 11/8 - 15/12 9/6 - 10/7 Up to 8/5 IEC-60422 Annex B Table B1
(ISO Class)

Particle Count
Physical Contamination

IEC 60970 All >130 000 / 100 ml 130 000 / 100 ml 32 000 / 100 ml 1 000 / 100 ml 250 / 100 ml Table B1 of IEC 60422:2013
above 5mm
Moisture in Oil <72.5 >800 600 - 800 400 - 600 <400
(ASTM D1533, 72.5-170 ppm >600 400 - 600 300 - 400 <300 <200
IEC60814) >170 >400 300 - 400 200 - 300 <200
BDV <72.5 <30 30 - 40 >40 >50
(IEC 60156 72.5-170 kV <40 40 - 50 >50 >60
2.5mm) >170 <50 50 - 60 >60
Chemical Contamination

DDF (Tan δ)
@25°C All % <0.01 ≤0.005 ASTM D924
Degradation

Acidity ASTM D974

All mg KOH/g >3.0 2.0 - 3.0 0.5 - 2.0 <0.5 ≤0.1
IEC 62021-3

Colour (by lab

All >1.5 ≤1.0 ≤0.5 ISO 2049
testing)

133
 Condition assessment of power transformers

Affected
Voltage Class
Parameter/
Failure Mode of the Unit F E D C B A Reference
Investigative
equipment
Test
Colour (by Clear &
Black/Dark Light Yellow/Dark
Visual All Dark Brown Light Yellow without visible
Black Brown/Brown Yellow
Inspection) contamination
<40% of original 40-60% of >60% of >90% of
Inhibitor All %
value original value original value original value
Less than the
Sediment and >0.02% by
All % <0.02% by mass detection
Sludge mass
limit.
Environmental
PCB All mg/kg >2 <2
Hazard

134
Annex B How to Develop a TAI
B.1. Transformer Assessment and Development of TAIs
Condition assessment is a process by which a transformer is assessed considering all aspects that
could affect future performance. This Technical Brochure has provided some suggestions on how to
go about building a tool to understand the condition of a transformer and its intrinsic components. This
section illustrates how these tools can be used in practice.
After completing a condition assessment, if the assessment or the score indicates a problem, users
could also refer to Figure 1 found in CIGRE TB 445, Guide for Transformer Maintenance, which may
assist in determining the appropriate course of action. For convenience, a small section of TB 445 is
included at the end of this Appendix. Note however, that the terminology used in TB 445 does not
directly align with the terminology in this Technical Brochure.
Section 2.3 of this Technical Brochure indicates that there are five steps required to develop a
Transformer Assessment Index (TAI) for a fleet of transformers:
Step 1 – Determine the purpose of the Transformer Assessment Score and Index
Step 2 – Identify the Failure Modes & Mechanisms to be included in the TAI
Step 3 – Determine how each failure mode will be assessed
Step 4 – Design a calibrated system for categorising failure modes (Scoring Matrix)
Step 5 – Calculate a TAI score for each transformer
The following examples illustrate how these five steps can be applied in practice using the techniques
explained in this Technical Brochure. Note that the data is fictitious and has been generated only
to illustrate the techniques.
Step 1 - Determine the purpose of the Transformer Assessment Score and Index
Three separate indices are considered in this Annex:
 Replacement Index: to identify the transformers in the fleet which should be considered for
replacement.
 Repair Index: to identify the transformers that should be considered for repairs or major
maintenance activities. Note that this index is not designed to identify transformers that are due
for routine maintenance e.g. maintenance of the On-Load Tap Changer. Routine maintenance
should generally be carried out in accordance with the organisation’s maintenance plan.
 Refurbishment Index: To identify transformer that should be considered for major refurbishment
work i.e. work that is more complex and costly than the repair work that might be carried out to the
transformers highlighted by the repair index.
Note that the Repair index has not been calculated in the examples, but TABLE B1 shows the failures
mechanisms that could be considered for a Repair index. The repair index could be constructed
similarly to the Replacement index but using the appropriate failure mechanisms.
Step 2 & Step 3 - Identify the failure modes and mechanisms to be included in the TAI and
assessment techniques
TABLE B1 indicates the failure modes and mechanisms, and the related assessment techniques that
could be used to develop the three simple TAIs, using information that many users would have
available for a mineral oil filled transformer with Oil Impregnated Paper (OIP) bushings and an oil filled
OLTC. Note that table B1 is intended only as an example to assist in understanding the methods
described in the is brochure. Users should determine the failure modes and mechanisms that are
appropriate for the index that they wish to develop. As the indices are being developed for different
purposes, not all failure modes are included in each of the indices.
The table also indicates which failure mechanisms and diagnostic technique the user has chosen to
include in each of the indices being developed for these examples. The diagnostic techniques chosen
are only for the purpose of the example. Users should refer to Appendix A for details of other
diagnostic techniques that are applicable to the failure modes and mechanisms that they wish to
include in the index.
 Condition assessment of power transformers

The TAI is also influenced by the user’s experience, level of expertise, spare parts and replacement
transformers available, the way in which the fleet is generally operated, etc. Therefore, it is expected
that users designing a TAI will consider their own circumstance and will likely choose different
failure mechanisms and test techniques to include in their indices.
Note also that users with more assessment data would likely develop more complex TAIs. The
examples have been kept simple to better illustrate the methodology. The examples are not
intended to be “Industry Best Practice” TAIs or to be CIGRE’s preferred TAIs.
For the example, the ‘user’ has determined:
 High moisture content is not a driver for replacement, as the moisture can generally be removed
as part of a refurbishment. However, a transformer will not be considered for refurbishment if the
paper is already significantly degraded (High Furans). Similarly, transformers with high levels of
Partial Discharge or Arcing will not be considered for refurbishment as it is unlikely that these
problems can be easily corrected during the refurbishment process.
 Bushings can be replaced as part of either the Repair or Refurbishment process. However, as
replacing bushings can be expensive if identical bushings are not available, defective bushings
are also one of the drivers for replacement.
 OLTCs are maintained in accordance with original equipment manufacturer’s instructions by
suitably trained technicians. It is therefore assumed that any problems with OLTCs cannot be
easily corrected by further maintenance or a simple repair. OLTC problems could be addressed
during a Refurbishment and are also considered a driver for replacement.
 Most problems with the cooling system, transformer tank and ancillary components can be
addressed by a refurbishment or repair and are not considered drivers for replacement.
 Mineral oil can be easily replaced and is not considered a driver for replacement. However, if oil
tests indicate that the oil has sulphur compounds that could potentially react with the winding
material to create copper corrosion, corrosive sulphur may have already been deposited in the
winding. The user has therefore decided not to undertake major refurbishment work on
transformers that have been filled with this type of oil. Although this type of oil is not considered a
driver for replacement, it is used in the Repair Index as filter to prevent transformers that have
been filled with corrosive oil from being refurbished.
 Diagnostic techniques have been chosen for the selected failure modes based on the techniques
currently in use in the user’s organisation, and data readily available to the user. Even though this
may not give the most accurate assessment of condition, it ensures that data is available for most
transformers being assessed.
Step 4 – Design a calibrated system for categorising failure modes (Scoring Matrix)
The generic Scoring Matrix in section 2.3.4 can been used to allow test results and transformer
assessment data related to each of the failure mechanisms to be converted into one of five levels.
(Green, Yellow, Orange, Pink or Red) in a consistent way. Note that variations of this scoring matrix
have been used elsewhere in this brochure to better allow assessment of the components of the
transformer.
Users should also consider and define the time scales used in the scoring matrix to ensure
consistency during assessments i.e. the terms ‘as soon as possible’ and ‘short term’ should be
replaced with times that suit the user. The times chosen may align with the times required to arrange
maintenance or obtain spare components for transformers, or with longer term replacement or repair
budgets.
Note that users can develop their own scoring matrix, with any number of levels; provided that a clear
description of each level is assigned to ensure that scoring of all components and sub-components of
the transformer are consistent.
Step 5 – Calculate a TAI score for each transformer
Multiple methods of calculating scores will be used in these examples to illustrate and compare
different scoring techniques. Refer to Table 2-2 Methods of calculating a Transformer Assessment
Score for other methods used to calculate a final score.

136
 Condition assessment of power transformers

B.2. Examples
B.2.1. Worst Case Component Scores
For these examples, scores have been allocated to the following sub-components using a “Worst
Case approach” scoring system. “Worst Case” scores are allocated to the following sub-components:
 Core and Windings – Repairable
 Core and Windings – Non - Repairable
 Bushings
 OLTC
 OLTC Insulating Oil
 Cooling System
 Transformer Tank
 Ancillary Components
 Corrosive Sulphur
 Main tank insulating Oil
Scores for each failure mechanism have been allocated for a transformer (Transformer 1) by the
transformer engineer using the scoring matrix and the data available for the transformer. For each
sub-component, one or more failure mechanisms has been considered and the worst score for all the
considered failure mechanisms is allocated to the sub-component. Note that the test data and other
information shows the scores allocated for the selected failure modes/degradation mechanisms and
shows how the worst case scoring method has been used to allocate scores to the sub-components.
Table B1 Worst Case Sub-Component Scores

137
Condition assessment of power transformers

138
 Condition assessment of power transformers

Refer to section 2.4.2 for details of how the above information can be used to calculate an assessment
score for the transformer

B.2.2. Assessed sub-component scores for the fleet

Sub-Component scores, which have been generated from data from 10 fictitious transformers using
the scoring matrix and the worst case scoring method (to derive sub-component scores), are provided
below, and will be used to generate TAI scores for each of the transformers. (Note that the raw data
has not been provided).

139
 Condition assessment of power transformers

Table B2 Assessed Sub-Component scores for the fleet

B.2.3. Replacement Index

The sub-components applicable to the Replacement Index and the scores for the 10 transformers are
detailed in the following table.
A number of different ways of generating a total score for the transformer are shown. The scores can
assist the Asset Manager to determine which transformers should most urgently be considered for
replacement.
Note however, that transformers that have only bushings or OLTCs scored as Red, may also be
suitable for repair or refurbishment.
Table B3 Calculation of Replacement Scores

The Asset Manager will have to further consider the appropriate action for the transformers being
considered for replacement.

140
 Condition assessment of power transformers

The Simple Score shown in the table is a worst-case score – but the number of sub-components
which have that score is also indicated. Tx 2, Tx 3, Tx 5 and Tx 8 all need urgent attention. However, it
is obvious from the score Tx 8 has 2 sub-components that need urgent attention.
The Count per Category approach provides more information about the overall condition of the
transformer. Although displaying the result in this way is not as simple as a single number, the 5
columns of numbers can be easily ordered, so that the transformers that most urgently need attention
are at the top of the list.
The Simple Numeric Score is calculated in this example, by allocating marks as follows. Green = 0,
Yellow = 1, Orange = 2, Pink = 3 and Red = 4. Although this type of scoring provides a simple
numeric score, the urgency, or the time scale for action, is lost. Tx 2 has a score of only 4, even
though one sub-component needs urgent attention! Tx 10 has a worse score, even though the worst
sub-component score is Orange, and it does not need urgent attention.
The Hybrid Score combines the simple numeric score with the Worst-Case sub-component score.
This gives an indication of the time scale for action (from the worst case score), along with an
indication of the overall condition of the transformer (from the simple numeric score). Tx 2 has a score
of 4 (Red). This indicates that the overall condition of the transformer is good, but a sub-component
needs urgent attention.
The Non-linear Score is calculated using the formula:
𝑘−1

𝑇𝐴𝐼 = ∑ 𝑥𝑛 𝑖 𝑛
𝑛=0
Where:
 i is the number base / radix, which is equal to or greater than the number of failure modes
included in the TAI
 xn is the number of failure modes per category
 k is the number of categories included in the failure mode assessment
Although it is not immediately obvious, the single numeric score per transformer indicates the
timescale for action. A score above 81 can only be achieved if at least 1 sub-component needs urgent
attention (i.e. is scored as Red). Similarly a score above 27 can only be achieved with at least 1 sub-
component scored as Pink.

B.2.4. Use of Weighting Factors

The designer of the TAI has decided that although bushings and OLTCs are drivers for transformer
replacement (it may be more economical to replace a transformer if multiple sub-components need
attention), they can also be addressed during a repair or refurbishment process. Weighting factors
have been added to the scores for bushings and OLTCs to reduce the scores of those transformers
whose bushings and OLTCs need attention. (Due to the simplicity of the example, this leaves only 1
sub-component without a weighting factor.) In the following example, bushings have been weighted by
a factor of 0.3, and OLTCs by a factor of 0.5. These weightings have been chosen only to illustrate the
effect on the resultant scores. Users wishing to use weighting factors should determine appropriate
weighting factors to suit their TAI and their needs.

141
 Condition assessment of power transformers

Table B4 Scores with Weighting Factors

The dangers of using a Simple Numeric Scoring system with weighting factors can be clearly seen
from this example! Tx 2 now has a score of only 1.2. If the transformers are ordered based on their
simple numeric score, Tx 2 would be number 8 on the list of 10. This transformer will not get the
attention of an Asset Manager, despite the fact that it has a failure mode that needs to be addressed
urgently! Note however that other scoring systems still show the timescale for required action.

B.2.5. Refurbishment Index

The refurbishment index uses the Main Tank – Non Repairable and the Corrosive Sulphur scores as
filters. If either of these scores are Pink or Red, then the transformer is not considered for
refurbishment, as it may have limited remaining life. On this basis, Tx 1, Tx 4, Tx 5 and Tx 8 have not
been scored or considered for refurbishment. Scores for 8 other sub components have been used to
determine which transformers are most in need of refurbishment.

142
 Condition assessment of power transformers

Table B5 Filtering to develop a Refurbishment Index

Refurbishment scores are show in the table below

Table B6 Refurbishment Scores

Orange

Yellow

Simple Non
Green
Pink
Red

Numeric Hybrid Score Linear

Score Score

TX 1 Score - - - - - - - -
TX 2 Score 1 0 2 3 2 11 11 (Red) 4250
TX 3 Score 3 2 1 2 0 22 22 (Red) 13392
TX 4 Score - - - - - - - -
TX 5 Score - - - - - - - -
TX 6 Score 0 0 0 5 3 5 5 (Yellow) 43
TX 7 Score 0 0 0 4 4 4 4 (Yellow) 36
TX 8 Score - - - - - - - -
TX 9 Score 0 0 1 1 4 3 3 (Orange) 76
TX 10 Score 4 0 3 1 0 23 23 (Red) 16584

As there are now 8 sub components being scored into the five categories, the non-linear score gives
much larger numbers. Red scores are multiplied by 84 as per the formula.

B.3. Extract from TB 445

If the assessment or assessment score indicates a problem, or multiple problems of different severity,
then choices need to be made with respect to actions, and determination are they MAJOR or MINOR
work.
MINOR WORK, could be classified as “Corrective Maintenance” which is an operation carried out to
restore any part of the transformer, which has “failed” or “degraded” to the point where it needs
corrective action to avoid loss of performance or a major failure. Examples might include oil
processing, cooling fan replacement, repair of leaks, possibility bushing replacement, and OLTC minor

143
 Condition assessment of power transformers

repairs. It implies that the unit, is out of service during this work for a short period of time, BUT is not
removed from the site to affect these repairs or component replacements.
MAJOR WORK, could be classified under a few definitions, all of which should be passed through a
Technical and Economic Evaluation to arrive at the decision to either Refurbish or Repair, or scrap
and replace.
Technical & Economic Evaluation
The cycle of transformer operation and routine corrective maintenance is not perpetual. When an
asset suffers severe damage or when its reliability is no longer satisfactory, a technical and economic
evaluation must be made to decide the best option between scrap and replace, repair or refurbish and,
if the work is to be performed on-site or in a workshop. CIGRE TB 248 Guide on Economics of
Transformer Management, provides an excellent normative approach to this process.
Major Repair, ranges from replacement or refurbishment of major components, such as ALL the
bushings (at one time interval), OLTC overhaul, or complete removal and replacement of the cooling
system. This still implies work which may be done at site, but over a longer outage duration.
In most cases ANY movement of a unit from the site to a repair facility is considered MAJOR work.
REFURBISHMENT, is an option, that may have been decided based on an economic evaluation that it
is worthwhile, such as complete replacement of major insulation, or complete new windings. Any work
that involved complete removal, and disposal of the dielectric fluid can be defined as major work.
Scrap and Replace, (or END of Life) The service life of a transformer should end when its condition is
such that it cannot be kept in service, nor be put back into service, primarily because a technical and
economic evaluation has determined that its return to a serviceable condition is not economical, nor
potentially unsafe.

144
 Condition assessment of power transformers

Annex C Literature overview

Introduction
Many other groups and individuals have developed and published information about Transformer
Assessment Indices; although they have generally referred to their indices as a “Health Index.” For the
benefit of readers, this Appendix aims to provide an overview of some of the current documents that
have been published about this topic, and the methodologies that these documents describe. Many
other documents have also been written about this topic, but it was not possible to identify, review and
reference all documents.
It must be noted that concepts, methodologies and terminology detailed in these publications, will in
many cases vary from what is presented in the Chapters of this Technical Brochure. The references to
these documents are however provided to allow readers to consider the concepts and methods
proposed by others. It should also be noted that terminology e.g. “Health Index,” used in this Appendix
is the terminology generally used by the authors of the reference documents and may vary from the
terms, or the definitions of these terms, used in the Chapters of this brochure.
This document does not aim to provide a detailed description or to make recommendations or
comments regarding the published methodologies. The interested reader should refer to the content of
the references for further information. As this is a relatively new topic, research is still being
undertaken and no consensus has been reached in the literature on a preferred method.

Overview
Generally, the process encompasses several steps [1], as shown in Figure 1, and this document
presents an overview of this process with references to published literature.

INPUT DATA CLASSIFICATION AGGREGATION OUTPUT

Operation
records Condition
indicators
Weighed sum,
Components
Maintenance Non-linear,
and/or Encoding
records Artificial Intelligence,
failure modes
...
Probability of failure
Monitoring
records

Other

Figure C.1 – Process for the calculation and uses of the health index

Input data
 Origin of data
The pragmatic approach, as presented in [2], is to collect the available data about the transformers’
condition that will later be fed into the aggregation algorithm. Gathering these pieces of data is a
crucial task since the value of the health index, the output of its analysis and its potential applications
directly depend on the quality of the original data. Maintenance guide and practices including typical
routinely performed diagnostic tests can be found in [3-5].
Besides results from diagnostic tests, other general information such as a transformer’s age, load
history, service conditions and previous maintenance actions are often recorded by utilities and may
also be fed into the health index [6-8]. Authors in [9] include economic parameters in their assessment
to use the health index result to prioritize maintenance activities.
Often, data is not usable due to values being aberrant or unavailable. In these cases, different choices
can be made. In [10, 11], authors chose not to include transformers for which data is missing. In [12],

145
 Condition assessment of power transformers

the authors do not take into account nonexistent data for the calculation [10]. Authors have also
assumed a good condition but with a lower confidence in the assessment. In [13], each data input is
associated with a reliability index.
When very little data is available, statistical curve-fitting techniques [14, 15] may also be used to try to
determine a trend within the few results present. These curves are then compared against known
values. This protocol may help to assess the validity of the data. When possible, experts also try to
deduce data from other parameters that are already known [15].
 Encoding of data
Input parameters generally consist of direct values taken from the aforementioned tests and service
condition records. The number of input parameters varies widely from one paper to another as it
ranges from less than 5 to more than 20 [1]. When a lot of parameters are taken into account, two
different aggregation approaches are considered. Some authors keep all their parameters as
individual inputs [1, 8, 9, 11, 13, 16-21] whereas others group some parameters into sub-categories or
sub-indices [1, 7, 14, 22-32] related to one component (e. g.: the active part or the oil). The available
data can also be classified and associated to the failure modes of the components to be assessed [33,
34].
Both approaches result in the clustering of data of different natures, for instance a number obtained by
an objective measurement and a word obtained by a subjective (visual) evaluation. To place all data
on the same plan, authors in [13, 24, 35] use standardized dimensionless numbers between 0 and 1.
Similarly, it is also possible to establish a scoring system. In this case, results from maintenance or
monitoring tests are all assigned a score according to their value. To harmonize all the values before
aggregating the data in a health index, scores can be put on the same scale (e. g. from 1 to 5) [1, 9,
11, 14, 20, 21, 23, 26, 28, 30, 31, 34, 36-39]. Scales can be determined by classifying values and
applying thresholds. These scoring scales were questioned [16, 34, 40] as most papers use linear
scoring scales and this linearity may give too much or not enough weight to some values. It could also
suggest a proportionality factor within the results that may not actually exist.

Aggregation methods
 Weighted sum
In this approach, the input values are all given weights depending on their relative importance for the
condition assessment of the transformer. Weights may also be determined considering the final use of
the health index [7].
Authors from [41] perform a common factor analysis to identify correlations between results from the
oil tests and to determine which weights should be assigned to the different values of the oil tests.
A weighted sum is then calculated, and the output value is a number. Various scales can be used for
the output: a digit ranked on scale starting from 0 or 1 [1, 10, 11, 27, 39, 42] or a scale ranging from 0
to 1, such that the number can be represented as a percentage [1, 7, 13, 14, 20, 21, 23-26, 29, 35, 38,
41, 43].
In [44], failure statistics and the score related to each failure mode are associated to calculate the
individual failure rate of the transformer.
Authors from [32] choose to calculate a quadratic mean of all the parameters.
However, according to [16, 34, 40], the use of this procedure leads to some failure modes being
hidden or averaged by others.
 Sum of non-linear scores
Authors in [37] use a non-linear scoring scale and in [34], authors use a logarithmic scale for the
standardization of the input parameters. The goal is to make the process capable of pointing out a
specific component’s poor condition regardless of other components in good condition.
 Worst case
Statistics can be another approach for the calculation of the health index. Several models are
presented in the literature. In [33, 40, 45], authors analyse the different failure modes using the Failure
Mode and Effect Analysis (FMEA).
In [33, 40, 45], the input data is combined into assessment functions that are not specifically detailed
in the papers. Authors from [33, 40] combine the input data for each failure mode to determine the

146
 Condition assessment of power transformers

transformer’s probability of failure. They also use comparative statistics to determine trends in data
and to identify sudden changes that could account for an unexpected degradation of the apparatus.
In [45], results from the FMEA provide information about the worst case scenario for a transformer’s
failure. The combination of the assessment functions and this worst case approach results in the
health index.
 Statistical regression
In [19] and [28], authors aggregate the data by performing a statistical regression. For instance,
authors in [19] use binary logistic regression. Input parameters are sorted into only two categories:
healthy or unhealthy. Weights that should be assigned to each input are determined by a maximum
likelihood criterion. The health index is calculated using a non-linear equation. A logistic regression is
then performed on the entire fleet’s results to assess their validity and, if need be, it can be used to
refine the values of the weighing coefficients.
 Artificial Intelligence
The development of computer science and data processing over the last years has resulted in the
creation of new algorithms such as pattern recognition algorithms [31], neural networks [9, 16, 18],
wavelet networks [17], fuzzy logic [30, 46], optimization methods [36, 47], all based on machine
learning techniques. These new tools are capable of analysing large datasets [9], for instance results
from monitoring records. However, the accuracy of the algorithm mostly relies on the model chosen by
the user and the amount of data that the algorithm is fed with [9]. As a consequence, an increase in
the amount of input data leads to an improvement of the algorithm’s performance.
The result of the calculation is a number, typically between 0 and 1 [9, 16-18, 30, 47]. Output values
are compared against those obtained via the weighted sum but they could be equally applied to any
non-linear approaches.
In [12], authors develop a computer package based on Boolean logic to aggregate all the input data.
The package provides the user with maintenance recommendations.
 Summary
As shown previously, the methods used to aggregate the data into a health index are various and
research is still being made on this topic. Table 1 summarizes the link between the aggregation
methods and the published papers.
Table 1 – Summary of the aggregation methods
Aggregation methods Papers
Weighted sum [1, 7, 10, 11, 13, 14, 20, 21, 23-27, 29, 32, 35, 38, 39, 41-44]
Sum of non-linear scores [34, 37]
Worst case [33, 40, 45]
Statistical regression [19, 28]
Artificial Intelligence [9, 12, 16-18, 30, 31, 36, 46, 47]

Outputs
 Condition indicators
As stated in [14], the health index represents the asset condition in an easy way (a colour, a number
or a text string) to be used by the Asset Manager. The classification of the transformers may contain
three [18, 39, 46, 47], four [8, 16, 22, 27, 35, 38] or five classes [2, 9, 20, 21, 23, 25-27, 30], the
condition of the transformer being ranged from excellent/very good to mediocre/very bad. A colour
gradient from green to red may also be used [7, 8, 15, 25, 35, 45, 48] to easily bring to light the state
of the transformer. This gradient may also provide information about the quality of the original data. In
[45], the health index is displayed on a 3D graph. One axis represents the condition of the unit with a
colour gradient from green to red; a second axis represents the quality of the original data by fading
the colour as the confidence-level on the data decreases. A third axis gives information about the
number of transformers per category. Further analyses of the input data may then provide information
for fault diagnosis. For instance, authors in [12] propose a diagnosis of the transformer by combining
input data using logic gates.

147
 Condition assessment of power transformers

 Probability of failure
As for the Health Index, there is no consensus in the literature regarding a methodology to assess the
probability of failure.
In the literature related to the health index of transformers, the probability of failure is determined using
statistical distributions [10, 13, 15, 23, 24, 26, 28, 42-44], with specific failure analysis techniques such
as FMEA or probabilistic fault trees [33, 40], directly via the health index output value [2, 9, 11, 22, 26],
or indirectly via the apparent age concept. The concept of apparent age – or effective age – translates
the condition assessment information in a corrected age that may be used to estimate the present and
future probability of failure based on statistical analyses [10, 26, 42, 43].
Authors in [25, 26] use the probability of failure as a basis for estimating the remnant life of the
transformer. In [15], the remnant life is estimated directly using the input data from maintenance and
monitoring records. A flowchart is then used to determine the need for maintenance of the
transformer.

References used in the Literature review

[1] A. Azmi, J. Jasni, N. Azis, and M. Z. A. A. Kadir, "Evolution of transformer health index in the form
of mathematical equation," Renewable and Sustainable Energy Reviews, vol. 76, pp. 687-700, 2017.
[2] D. T. Hughes, "The use of health indices to determine end of life and estimate remnant life for
distribution assets," Proceeding of 17th International Conference on Electricity Distribution, 2003.
[3] CIGRE Working Group A2.18, "Life management techniques for power transformer," Brochure
227, 2003.
[4] CIGRE Working Group A2.34, "Guide for transformer maintenance," Brochure 445, 2011.
[5] M. M. Islam, G. Lee, and S. N. Hettiwatte, "A review of condition monitoring techniques and
diagnostic tests for lifetime estimation of power transformers," Electrical Engineering, pp. 1-25, 2017.
[6] IEC White Paper, "Strategic asset management of power networks," 2016.
[7] A. Babizki, M. Foata, K. Lindl, D. Beerboom, E. Pawlowski, M. Zdrallek, et al., "Practical approach
for fleet management of transformers considering different stakeholders’ perspectives," 4th
International Colloquium "Transformer Research and Asset Management", 2017.
[8] N. Dominelli, M. Lau, D. Olan, and J. Newell, "Equipment health rating of power transformers," in
Conference Record of IEEE International Symposium on Electrical Insulation, 2004, pp. 163-168.
[9] H. Zeinoddini-Meymand and B. Vahidi, "Health index calculation for power transformers using
technical and economical parameters," IET Science, Measurement and Technology, vol. 10, pp. 823-
830, 2016.
[10] C. Rajotte and P. Picher, "The use of health index for condition assessment and replacement
prioritizing," TechCon North America, 2017.
[11] A. Singh and A. G. Swanson, "Analysis and improvement of the plant health index used by
ESKOM to assess the condition of the power transformers in its distribution fleet," Proceeding of the
19th International Symposium on High Voltage Engineering, 2015.
[12] C. Roncero-Clemente and E. Roanes-Lozano, "A multi-criteria computer package for power
transformer fault detection and diagnosis," Applied Mathematics and Computation, 2016.
[13] B. Nemeth, C. Voros, and G. Csepes, "Health index as one of the best practices for condition
assessment of transformers and substation equipments - Hungarian experience," CIGRE, 2014.
[14] CIGRE Working Group C1.16, "Transmission asset risk management," Brochure 422, 2010.
[15] M. Vermeer and J. Wetzer, "Transformer health and risk indexing, including data quality
management," CIGRE, 2014.
[16] M. M. Islam, G. Lee, and S. N. Hettiwatte, "Application of a general regression neural network for
health index calculation of power transformers," International Journal of Electrical Power and Energy
Systems, vol. 93, pp. 308-315, 2017.
[17] M. Ahmed, M. Elkhatib, M. Salama, and K. B. Shaban, "Transformer health index estimation using
orthogonal wavelet network," in 2015 IEEE Electrical Power and Energy Conference: Smarter
Resilient Power Systems, EPEC 2015, 2016, pp. 120-124.

148
 Condition assessment of power transformers

[18] A. E. B. Abu-Elanien, M. M. A. Salama, and M. Ibrahim, "Determination of transformer health

condition using artificial neural networks," in INISTA 2011 - 2011 International Symposium on
INnovations in Intelligent SysTems and Applications, 2011.
[19] W. Zuo, H. Yuan, Y. Shang, Y. Liu, and T. Chen, "Calculation of a health index of oil-paper
transformers insulation with binary logistic regression," Mathematical Problems in Engineering, vol.
2016, 2016.
[20] M. A. Martins, "Condition and risk assessment of power transformers: A general approach to
calculate a Health Index," Ciencia e Tecnologia dos Materiais, vol. 26, pp. 9-16, 2014.
[21] W. Wattakapaiboon and N. Pattanadech, "The new developed Health Index for transformer
condition assessment," in CMD 2016 - International Conference on Condition Monitoring and
Diagnosis, 2016, pp. 32-35.
[22] Y. Sukrislismono and A. Kartika, "Case study on asset health indices implementation on
distribution transformer asset at PLN Bali Indonesia," in IET Conference Publications, 2015.
[23] I. G. N. S. Hernanda, A. C. Mulyana, D. A. Asfani, I. M. Y. Negara, and D. Fahmi, "Application of
health index method for transformer condition assessment," in IEEE Region 10 Annual International
Conference, Proceedings/TENCON, 2015.
[24] I. C. Durán, L. S. Rosero, A. Pavas, and O. G. Duarte, "Comparison of maintenance scheme
effects on power transformer reliability performance," Ingenieria e Investigacion, vol. 35, pp. 73-81,
2015.
[25] A. Naderian, S. Cress, R. Piercy, F. Wang, and J. Service, "An Approach to Determine the Health
Index of Power Transformers," in Conference Record of the 2008 IEEE International Symposium on
Electrical Insulation, 2008, pp. 192-196.
[26] A. N. Jahromi, R. Piercy, S. Cress, J. R. R. Service, and W. Fan, "An approach to power
transformer asset management using health index," IEEE Electrical Insulation Magazine, vol. 25, pp.
20-34, 2009.
[27] F. Ortiz, I. Fernandez, A. Ortiz, C. J. Renedo, F. Delgado, and C. Fernandez, "Health indexes for
power transformers: A case study," IEEE Electrical Insulation Magazine, vol. 32, pp. 7-17, 2016.
[28] J. Qiu, H. Wang, D. Lin, B. He, W. Zhao, and W. Xu, "Nonparametric regression-based failure rate
model for electric power equipment using lifecycle data," IEEE Transactions on Smart Grid, vol. 6, pp.
955-964, 2015.
[29] F. Scatiggio, L. Calcara, and M. Pompili, "Risk prevention for HV transformers: Beyond the health
index," in 34th Electrical Insulation Conference, EIC 2016, 2016, pp. 182-185.
[30] L. Sun, Z. Ma, Y. Shang, Y. Liu, H. Yuan, and G. Wu, "Research on multi-attribute decision-
making in condition evaluation for power transformer using fuzzy AHP and modified weighted
averaging combination," IET Generation, Transmission and Distribution, vol. 10, pp. 3855-3864, 2016.
[31] T. K. Saha, H. Ma, C. Ekanayake, D. Martin, and D. Allan, "Pattern recognition techniques for
determining the health index of oil-paper insulation of in-service power transformers," CIGRE, 2014.
[32] I. V. Davidenko and T. S. Kuzina, "Analysis of the modern methods of the power transformers
health index calculation," in Proceedings of the 2017 IEEE Russia Section Young Researchers in
Electrical and Electronic Engineering Conference, ElConRus 2017, 2017, pp. 1491-1495.
[33] P. Lorin, L. Cheim, L. Pettersson, K. Gustafsson, and E. Tenyenhuis, "Transformer health index
and probability of failure based on failure mode effects and analysis (FMEA) of a reliability centered
maintenance program (RCM)," CIGRE 2016, 2016.
[34] R. J. Heywood and T. McGrail, "Clarifying the link between data, diagnosis and Asset Health
Indices," in IET Conference Publications, 2015.
[35] F. Scatiggio, A. Fraioli, V. Iuliani, and M. Pompili, "Health index: the TERNA's practical approach
for transformers fleet management," CIGRE, 2014.
[36] K. Ibrahim, R. M. Sharkawy, H. K. Temraz, and M. M. A. Salama, "Selection criteria for oil
transformer measurements to calculate the Health Index," IEEE Transactions on Dielectrics and
Electrical Insulation, vol. 23, pp. 3397-3404, 2016.

149
 Condition assessment of power transformers

[37] E. Duarte, D. Falla, J. Gavin, M. Lawrence, T. McGrail, D. Miller, et al., "A practical approach to
condition and risk based power transformer asset replacement," in 2010 IEEE International
Symposium on Electrical Insulation, 2010, pp. 1-4.
[38] G. Tanasescu, B. Gorgan, S. Busoi, A. Badita, and P. V. Notingher, "Monitoring and diagnosis of
electrical equipment using a web software. Health index and remaining lifetime estimation," in
Proceedings of the International Conference - 2016 Conference on Diagnostics in Electrical
Engineering, Diagnostika 2016, 2016.
[39] H. Malik, A. Azeem, and R. K. Jarial, "Application research based on modern-technology for
transformer Health Index estimation," in International Multi-Conference on Systems, Sygnals &
Devices, 2012, pp. 1-7.
[40] L. Cheim, P. Lorin, L. Lin, P. Patel, and J. Vines, "A Novel Dynamic Fleet Wide Condition
Assessment Tool of Power Transformers," presented at the CIGRE SC A2 & C4 Joint Colloquium
2013, Zurich, CH, 2013.
[41] A. D. Ashkezari, H. Ma, C. Ekanayake, and T. K. Saha, "Multivariate analysis for correlations
among different transformer oil parameters to determine transformer health index," in IEEE Power and
Energy Society General Meeting, 2012.
[42] P. Picher, J.-F. Boudreau, A. Manga, C. Rajotte, C. Tardif, G. Bizier, et al., "Use of Health Index
and Reliability Data for Transformer Condition Assessment and Fleet Ranking," presented at the
CIGRE, Paris, 2014.
[43] E. Abbasi and O. P. Malik, "Failure rate estimation of power transformers using inspection data,"
in 2016 International Conference on Probabilistic Methods Applied to Power Systems, PMAPS 2016,
2016.
[44] J. H. Jürgensen, L. Nordström, and P. Hilber, "Individual failure rates for transformers within a
population based on diagnostic measures," Electric Power Systems Research, vol. 141, pp. 354-362,
2016.
[45] M. Vermeer, J. Wetzer, P. Van Der Wielen, E. De Haan, and E. De Meulemeester, "Asset-
management decision-support modeling, using a health and risk model," in 2015 IEEE Eindhoven
PowerTech, PowerTech 2015, 2015.
[46] A. E. B. Abu-Elanien, M. M. A. Salama, and M. Ibrahim, "Calculation of a health index for oil-
immersed transformers rated under 69 kV using fuzzy logic," IEEE Transactions on Power Delivery,
vol. 27, pp. 2029-2036, 2012.
[47] K. I. Mohamadeen, R. M. Sharkawy, and M. M. Salama, "Binary cat swarm optimization versus
binary particle swarm optimization for transformer health index determination," in ICET 2014 - 2nd
International Conference on Engineering and Technology, 2015.
[48] M. Vermeer, S. J. D.:, and J. Wetzer, "The application of health- and risk indices as a decision-
support tool for utilities," presented at the The Asset Management Conference 2016, 2016.

150
 Condition assessment of power transformers

Annex D Principal INTERnational standards for

insulating liquids
The principal international standards currently being used for mineral oils, natural ester liquids,
synthetic ester liquids and silicones are listed following:
Mineral Oil:
IEC 60296:2012 [71]
NEW MINERAL OIL Fluids for electrotechnical applications – Unused mineral insulating oils for
transformers and switchgear
IEC 60422:2013 [42]

USED MINERAL OIL Mineral insulating oils in electrical equipment – Supervision and maintenance
guidance

Natural ester liquids:

IEC 62770:2013 [75]
NEW NATURAL ESTER Fluids for electrotechnical applications – Unused natural esters for transformers
and similar electrical equipment
IEC 62975 (under development)
USED NATURAL ESTER Use and maintenance guidelines of natural esters insulating liquids in electrical
equipment

Synthetic ester liquids:

NEW SYNTHETIC IEC 61099:2010 [77]

ESTER
Specifications for Unused Synthetic Organic Esters for Electrical Purposes
IEC 61203:1992 [78]
USED SYNTHETIC
ESTER Synthetic organic esters for electrical purposes – Guide for maintenance of
transformer esters in equipment

For silicone liquid (polydimethylsiloxane oil):

IEC 60836:2015
NEW SILICONE LIQUID
Specifications for unused silicone insulating liquids for electrotechnical purposes
USED SILICONE IEC 60944:1988
LIQUID
Guide for the maintenance of silicone transformer liquids

Note that these standards may use different test methods and therefore state different acceptance
criteria.

151
Condition assessment of power transformers

152
 Condition assessment of power transformers

Annex E References
This Annex provides details of documents that are relevant to the work of the CIGRE Working Group
that developed this Technical Brochure. Many of these documents are referenced in the text of the
brochure, however some are detailed only as additional sources of information.
Note that some, but not all, of the documents referenced in Annex C are also included in this
reference list.
[1] ISO, “Condition monitoring and diagnostics opf power transformers,” 2018.

[2] CIGRE, Technical Brochure 735 - Transformer Post-Mortem Analysis, 2018.

[3] IEEE Std, “C57.91-2011 - Guide for Loading Mineral-Oil-Immersed Transformers.,” 2011.

[4] IEC, 60076-7 Loading guide for mineral-oil-immersed power transformers, 2018.

[5] CIGRE, “Technical Brochure 445 - Guide for Transformer Maintenance,” 2011.

[6] IEEE Std, “C57.140 - Guide for Evaluation and reconditioning of Liquid Immersed Power
Transformers,” 2017.

[7] V. Peralta and M. Bouzeghoub, “A framework for analysis of data freshness,” in International
Workshop on Information Quality in Information Systems, 2004.

[8] V. Catterson and B. Sparling, “BIG DATA … where is all the data, and is it useful?,” in TechCon
North America, 2017.

[9] CIGRE, “Technical Brochure 630 - Guide for Transformer Intelligent Condition Monitoring
(TICM)”.

[10] IEEE Std, C57.104 Guide for the Interpretation of Gases Generated in Oil-Immersed
Transformers, 2008.

[11] M. Duval, “A Review of Faults Detectable by Gas-in-Oil Analysis in Transformers,” IEEE

Electrical Insulation Magazine - Vol 18, No. 3, May/June 2002.

[12] M. Duval, “Use of Duval Pentagons and Triangles for Interpretation of DGA in electrical
Equipment,” in Proceeding s from Techcon North America, Alburquerque, 2016.

[13] CIGRE, “Technical Brochure 228 - Guide on Economics of Transformer Management”.

[14] IEC, “61649 Weibull Analysis”.

[15] CIGRE, “Technical Brochure 706 - Guidelines for the use of Statistics and Statistical Tools on
Life Data”.

[16] CIGRE, “Technical Brochure 642 Transformer Reliability Survey,” 2015.

[17] P. Picher, J.-F. Boudreau, C. Rajotte, C. Tardif, G. Bizier, N. Di Gaetano , D. Garon, B. Girard,
J.-F. Hamel and S. Proulx, “A2-101 - Use of Health Index and Reliability Data for Transformer
Condition Assessment and Fleet Ranking.,” 2014.

[18] G. Russel, O. Krause, T. Saha, J. Marks, D. Martin and A. Alibegovic-Memisevic, “On the
development of power transformer failure models: an Australian case study," IEEE PES
conference, 2017.,” in IEEE PES conference, 2017.

[19] E. Abbasi and O. P. Malik, “Failure rate estimation of power transformers using inspection
data,” in International Conference on Probabilistic Methods Applied to Power Systems, PMAPS
, Beijing, China.

[20] A. Jahromi, R. Piercy, J. Service, S. Cress and W. FAN, “An approach to power transformer
asset management using health index,” IEEE Electrical Insulation Magazine, vol. 25, pp. 20-34,
2009.

153
 Condition assessment of power transformers

[21] C. T. B. 494, “Furanic Compounds for Diagnosis”.

[22] CIGRE, Technical Brochure 323 - Ageing of cellulose in mineral-oil insulated transformers,
2007.

[23] IEC, 60450 Measurement of the average viscometric degree of polymerization of new and aged
cellulosic electrically insulating materials, 2004.

[24] IEC, 60076-2 Power transformers - Part 2: Temperature rise for liquid-immersed transformers,
2011.

[25] IEEE Std, C57.154 Design, Testing, and Application of Liquid-Immersed Distribution, Power,
and Regulating Transformers Using High-Temperature Insulation Systems and Operating at
Elevated Temperatures, 2012.

[26] IEC, 60076-14 Liquid-immersed power transformers using high-temperature insulation

materials, 2013.

[27] IEEE Std, “Transactions on Power Delivery,” vol. 17, no. 4, p. 991 – 995, Oct. 2002.

[28] IEC, 60599 Mineral oil-filled electrical equipment in service - Guidance on the interpretation of
dissolved and free gases analysis, 2015.

[29] C. 12-206, “New Mechanisms of Moderate Hydrogen Gas Generation,” Paris Session, 1998.

[30] IEEE Std, C57.155 IEEE Guide for Interpretation of Gases Generated in Natural Ester and
Synthetic Ester-Immersed Transformers, 2014.

[31] N. Dominelli, “The analysis of transformer oil for degradation products from overheated solid
insulation, sec. 6-21,,” in 58 th Annual international Conference of Doble Clients, 1991.

[32] J. Jalbert, R. Gilbert, P. Tétreault, B. Morin and D. Lessard-Dézie, “Identification of a chemical

indicator of the rupture of 1,4--glycosidic bonds of cellulose in an oil-impregnated insulating
paper system,” Cellulose, Vol. 14, pp. 295-309,” 2007.

[33] CIGRE, “Technical Brochure 393 - Thermal Performance,” 2010.

[34] E. S. W Lampe, “The Oxygen-free transformer, Reducing ageing by continuous degassing.,”

CIGRE Paper 12.05, 1976.

[35] CIGRE, “WG A2.30 TB 349 Moisture Equilibrium and Moisture Migration within Transformer
Insulation Systems,” 2008.

[36] T. Oommen, “Moisture Equilibrium curves - use and misue,” in Doble Client Conference,
Boston, 2003.

[37] CIGRE, Technical Brochure 625 - Copper Sulphide Long Term Mitigation and Risk
Assessment, 2015.

[38] CIGRE, Technical Brochure 378 - Copper Sulphide in Transformer Insulation.

[39] CIGRE, Techncial Brochure 659 - Transformer Thermal Modelling.

[40] IEC, 60270 High-voltage test techniques - Partial discharge measurements, 2000.

[41] IEEE Std, C57.106 Guide for Acceptance and Maintenance of Insulating Oil in Equipment,
2006.

[42] IEC, 60422 Mineral insulating oils in electrical equipment – Supervision and maintenance
guidance, Edition 4.0 ed., 2013-01.

[43] IEC, 60599 Mineral oil-impregnated electrical equipment in service – Guide to the interpretation
of dissolved and free gases analysis, Edition 2.1 ed., 2007-05.

[44] IEC, “60156 Insulating liquids - Determination of the breakdown voltage at power frequency -
Test method,” 1995.

154
 Condition assessment of power transformers

[45] IEC, 60247 Insulating liquids - Measurement of relative permittivity, dielectric dissipation factor
(tan d) and d.c. resistivity, 2004.

[46] IEC, “60814 Insulating liquids - Oil-impregnated paper and pressboard - Determination of water
by automatic coulometric Karl Fischer titration,” 1997.

[47] IEC, “60076-1 Power transformers - Part 1: General,” 2011.

[48] R. Willoughby, “Do Transformer Windings Need Re-clamping?,” in TechCon Asia Pacific,
Sydney, 2015.

[49] CIGRE, “Technical Brochure 673 - Guide on Transformer Tansportation”.

[50] Alstom Grid, Power Transformers - Expertise, vol. Volume 2, p. Section 1.6.

[51] W. Bartley, “Analysis of Transformer Failures,” in 79th International Conference of Doble

Clients, Boston, MA, 2012.

[52] C. W. A2.43, “Transformer Bushing Reliability,” 2018 .

[53] C. G. Garton, “Dielectric Loss in Thin Films of Insulating Liquids,” Journal of the Institution of
Electrical Engineers - Part III: Communication Engineering, including the Proceedings of the
Wireless Section of the Institution, vol. Vol. 88, pp. pp. 23-40, March 1941.

[54] M. Krüger, A. Kraetke, M. Koch, K. Rethmeier, M. Muhr and S. Ch, “New Diagnostic Tools for
High Voltage Bushings,” in Cigre VI Workspot, Foz do Iguacu, Brazil, 2010.

[55] G. Kopp, “Frequency Dependent Dissipation Factor Measurements for the Assessment of High
Voltage Bushings,” in TLM Hannover, 2010.

[56] IEC, 60137 - Insulated bushings for alternating voltages above 1000 V.

[57] IEEE Std, “C57.19.01 - Standard Performance Characteristics and Dimensions for Outdoor
Apparatus Bushings”.

[58] IEC, “Technical Report 61464, "Guide for the interpretation of dissolved gas analysis (DGA) in
bushings where oil is the impregnating medium of the main insulation (generally paper)",” ISBN:
2- 8318- 4459-2, 1998-08.

[59] IEC, 60214-1 Tap-changer- Part 1: Performance requirements and test methods, 2014.

[60] IEC, 60214-2 - Tap-changer- Part 2: Application guide, 2018.

[61] CEATI, “T067300-3042 - End of Life Decision Support Model for Transformer Load
Tapchangers,” 2007.

[62] L. Allard, M. Foata and e. al, “Paper A2-202 - Vibro-Acoustic diagnostic contributing to an
optimized maintenance strategy,” , paper A2-202.,” in CIGRÉ 2010 Session.

[63] CIGRE, “Electra Reliability Survey,” 1988.

[64] CIGRE, “Technical Brochure 227 - Life Management Techniques for Power Transformers”.

[65] IEEE Std, C57.12.00-2015 - General Requirements for Liquid-Immersed Distribution, Power,
and Regulating Transformers, 2015.

[66] UNITED STATES DEPARTMENT OF THE INTERIOR BUREAU OF RECLAMATION,

TRANSFORMER DIAGNOSTICS, FACILITIES INSTRUCTIONS, STANDARDS, AND
TECHNIQUES VOLUME 3-31 ed., 2003.

[67] CIGRE, Technical Brochure 343 - Recommendations for Condition Monitoring and Condition
Assessment Facilities for Transformers.

[68] IEC, 60099-4 Surge arresters - Part 4: Metal-oxide surge arresters without gaps for a.c.
systems, 2014.

155
 Condition assessment of power transformers

[69] IEC, 60099-5 Surge arresters - Part 5: Selection and application recommendations, 2013.

[70] CIGRE, Technical Brochure 436 - Experiences in Service with New Insulating Liquids.

[71] IEC, 60296 Fluids for electrotechnical applications — Unused mineral insulating oils for
transformers and switchgear, Edition 4.0 ed., 2012.

[72] ASTM, D3487-09 Standard Specification for Mineral Insulating Oil Used in Electrical Apparatus.

[73] IEC, 62975 (not published yet) – “Use and maintenance guidelines of natural esters insulating
liquids in electrical equipment”, TBC.

[74] IEEE Std, C57.147 Guide for Acceptance and Maintenance of Natural Ester Fluids in
Transformers, 2008.

[75] IEC, 62770 Fluids for electrotechnical applications – Unused natural esters for transformers and
similar electrical equipment, Edition 1.0 ed., 2013-11.

[76] ASTM, D6871-03 Standard Specification for Natural (Vegetable Oil) Ester Fluids Used in
Electrical Apparatus, 2008.

[77] IEC, 61099 - Insulating liquids - Specifications for unused synthetic organic esters for electrical
purposes, 2010.

[78] IEC, 61203 Synthetic organic esters for electrical purposes - Guide for maintenance of
transformer esters in equipment, Edition 1.0 ed., 1992-12 .

[79] T. V. Oommen, “EPRI Report EL-6761 - Bubble Generation During Transformer Overload EPRI
Report EL-6761,” 1990.

[80] T. V. Oommen, “Further Experimentation on Bubble Generation During Transformer Overload -

EPRI Report EL-7391,” March 1992.

[81] T. V. Oommen, E. M. Petrie and S. R. Lindgren, “Bubble Generation in Transformer Windings

Under Overload Conditions- Paper 8E, Doble Client Conference, March 1995,” March 1995.

[82] R. Gilbert, J. Jalbert, S. Duchesne, P. Tétreault, B. Morin and Y. Denos, “Kinetics of the
production of chain-end groups and methanol from the depolymerization of cellulose during the
aging of paper/oil systems. Part 2: Thermally upgraded insulating paper,” Cellulose, Vol. 17, pp.
253-269,” 2010.

[83] J. Jalbert, E. Rodriguez-Celis, S. Duchesne, B. Morin, M. Ryadi and R. Gilbert, ““Kinetics of the
production of chain-end groups and methanol from the depolymerization of cellulose during the
ageing of paper/oil systems. Part 3: Extension of the study under temperature conditions over
120oC,” Cellulose, Vol. 22, pp. 829-845,” 2015.

[84] R. Gilbert, J. Jalbert, P. Tétreault, B. Morin and Y. Denos, ““Kinetics of the production of chain-
end groups and methanol from the depolymerization of cellulose during the aging of paper/oil
systems. Part 1: Standard wood Kraft insulation,” Cellulose, Vol. 16,pp. 327-338,” 2009.

[85] A. Schaut, A. Stijn and S. Eeckhoudt, “Applicability of methanol as a new marker for paper
degradation in power transformers,” IEEE Trans. Dielectr. Electr. Insul., Vol. 18, No. 2, pp. 533-
540,” 2011.

[86] Y. Denos, A. Tanguy, J. Jalbert, R. Gilbert, P. Gervais and P. Guuinic, “Ageing diagnosis by
chemical markers influence of core-type and shell-type technology,” CIGRE, Paris, France,,”
2010.

[87] J. Jalbert, R. Gilbert, Y. Denos and P. Gervais, “ “Methanol: a novel approach to power
transformer asset management, IEEE Trans. Power Delivery, Vol.27, No. 2, pp. 514-520,”
2012.

[88] H. Bertrand, S. Besner, R. Beauchemin, L. Bolduc, R. Busatta, M. Chaaban, P. Couture, J.-B.

Dastous, L. Dupont, M. Duval, M. Gauvin, J. Goulet, J. Jalbert, F. Léonard, A. Lessard, B.
Noirhomme, P. Picher, S. Riendeau, F. Torriano, M. Foata and C. Rajotte, “Advanced

156
 Condition assessment of power transformers

monitoring and modelling methods for power transformer asset management,” European
Journal of Electrical Engineering Subject Area: Engineering: Electrical and Electronic
Engineering Vol. 10, Issue 5-6, pp. 591-620, 2010.

[89] J. Jalbert and M.-C. Lessard, “Cellulose chemical markers in transformer oil insulation. Part 1:
Temperature correction factors,” IEEE Trans. Dielectr. Electr. Insul., vol. 20, no. 6, pp. 2287-
2291, 2013.

[90] J. Jalbert and M.-C. Lessard, “Cellulose chemical markers relationship with insulating paper
post-mortem investigations, IEEE Trans. Dielectr. Electr. Insul.,” IEEE Trans. Dielectr. Electr.
Insul., vol. 22, no. 6 December, 2015.

[91] J. Jalbert, S. Duchesne, E.-M. Rodriguez-Celis, P. Tétreault and P. Colin, “Robust and sensitive
analysis of methanol and ethanol from cellulose degradation in mineral oils,” Journal of
Chromatography A, vol. 1256, pp. 240-245, 2012.

[92] E. Rodriguez-Celis, J. Jalbert, S. Duchesne, B. Noirhomme and M.-C. Lessard, Chemical

markers use for diagnosis and life estimation of power transformers, a preliminary study of their
origins”, CIGRE Canada Conference, CIGRÉ-083, 2012.

[93] E. M. Rodriguez-Celis, S. Duchesne, J. Jalbert and M. Ryadi, “Understanding ethanol versus

methanol formation from insulating paper in power transformers,” Cellulose, vol. 22, pp. 3225-
3236, 2015.

[94] CIGRE, “Technical Brochure 576 - Guide for IT Strategies for Asset Management of
Substations - General Principles”.

[95] CIGRE, “Technical Brochure 453 - Glossary of Terms Used in the Electric Industry”.

[96] CIGRE, “Technical Brochure 420 - Generic Guidelines for Life Time Condition Assessment of
HV Assets and Related Knowledge Rules”.

[97] CIGRE, “Technical Brochure 248 - Guide on Economics of Transformer Management”.

[98] IEEE Std, “C57.19.100 - Guide for Application of Power Apparatus Bushings”.

[99] IEEE Std, “C57.19.00 - General Requirements and Test Procedure for Power Apparatus
Bushings”.

[100] IEEE Std, “IEEE C57.152 - Guide for Diagnostic Field Testing of Fluid-Filled Power
Transformers, Regulators, and Reactors”.

[101] IEEE Stc, “C57.146 - Guide for the Interpretation of Gases Generated in Silicone-Immersed
Transformers”.

[102] IEEE Std, “C57.149 - Guide for the Application and Interpretation of Frequency Response
Analysis for Oil-Immersed Transformers”.

[103] IEEE Std, “C57.143 - Guide for Application for Monitoring Equipment to Liquid-Immersed
Transformers and Components”.

[104] IEEE Std, “IEEE C62.22 - IEEE Guide for the Application of Metal-Oxide Surge Arresters for
Alternating-Current Systems.”.

[105] CIGRE, Technical Brochure 443 - DGA in Non-Mineral Oils and Load Tap Changers and
Improved DGA Diagnosis Criteria.

[106] CIGRE, Technical Brochure 413 - Insulating Oil Regeneration and Dehalogenation.

[107] IEEE Std, C57.637 Guide for the Reclamation of Insulating Oil and Criteria for its use, 2007.

[108] IEEE Std, C57.121 Guide for Acceptance and Maintenance of Less Flammable Hydrocarbon
Fluid in Transformers, 1988.

157
 Condition assessment of power transformers

[109] IEC, 60567 Oil-filled electrical equipment - Sampling of gases and analysis of free and
dissolved gases - Guidance, Edition 4.0 ed., 2011-10.

[110] IEC, 60475 Method of sampling insulating liquids, Edition 2.0 ed., 2011-12.

[111] M. Krüger, M. Koch, A. Kraetke, S. Rätzke and J. Schnitzler, “New Tools for the Assessment of
High Voltage Bushings,” High Voltage Symposium, Stuttgart University, Germany, 2012.

[112] F. Huellmandel, M. Appold, R. K. A. Kuechler and J. Titze, “Condition Assessment of High

Voltage Bushings by Means of Dielectric Diagnosis with PDC, XV International Symposium on
High Voltage Engineering, University of Ljubljana, Elektroinštitut Milan,” in XV International
Symposium on High Voltage Engineering, University of Ljubljana, Elektroinštitut MilanVidmar,
Ljubljana, Slovenia, August 27-31, 2007.

[113] ABB, “Bushings Diagnostic and Conditioning,” ABB Brochure 2750 515-142en, 2014.

[114] W.Zaengl, “Dielectric Spectroscopy in Time and Frequency Domain for HV Power Equipment,
Part I: Theoretical Considerations, DEIS,” IEEE, vol. 19, no. 5 , September/October 2003.

[115] W.Zaengl, “Applications of Dielectric Spectroscopy in Time and Frequency Domain for HV
Power Equipment", Part II, DEIS,” IEEE, vol. 19, no. 6, November/December 2003.

[116] I. Atanasova-Hoehlein; et al., “Assessment of Oil Corrosivity,” in CIGRÉ A2-C4 Joint

Colloquium, Zurich, 2013.

[117] ASTM D, “D 1275 - Test method for corrosive sulphur in electrical insulating oils,” 2006.

[118] IEC, “62535 - Insulating liquids – Test method for detection of potentially corrosive sulphur in
used and unused insulating oil,” 2008.

[119] BS EN, 60814 - Insulating liquids. Oil-impregnated paper and pressboard. Determination of
water by automatic coulometric Karl Fischer titration, 1998.

[120] J. Fabre and A. Pichon, “Deteriorating Processes and Products of Paper in Oil. Application to
Transformers,” in Application to Transformers CIGRE, 1960.

[121] CEATI, “Report No. T063700-3039 - Health and risk index tool for transformers and LTCs”.

[122] P. Jarman, R. Haooton, L. Walker, Q. Zhong and T. Ish, “Transformer life prediction using data
from units removed from service and thermal modelling," in 43rd International Conference on
Large High Voltage Electric Systems,” in CIGRE 2010, Paris, 2010.

[123] I. Gutman, A. Holmberg, C. Ahlholm, U. Åkesson and L. Jonsson, “Long-term service

experience and inspection results of HV equipment made of silicone rubber insulators,” in
CIGRE Symposium, Auckland, New Zealand, Auckland, New Zealand, 2013.

[124] W. A2.30, “TB 349 MOisture equilibrium and Moisture Migration withing Transformer Insulation
Systems,” CIGRE, 2008.

158