See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/335737093

Earth Fault Protection in Sub Transmission Systems

Conference Paper · November 2015

CITATIONS READS

0 5,706

1 author:

Chris O'Halloran
Tesla Consultants
2 PUBLICATIONS 0 CITATIONS

SEE PROFILE

All content following this page was uploaded by Chris O'Halloran on 11 September 2019.

The user has requested enhancement of the downloaded file.

Earth Fault Protection in Sub Transmission Systems

CHRIS O’HALLORAN
Tesla Consultants
chris.ohalloran@tesla.co.nz
EESA NSW Annual Conference & Trade Exhibition
25 November 2015

Abstract
The sub­transmission networks in many New Zealand distribution lines companies are
changing. Transpower is introducing neutral earthing resistors (NERs) at many Grid Exit
Points (GXPs), unearthed wind farms with significant cable networks are being embedded,
and distribution companies are increasingly looking to mesh their existing and planned
networks to enhance reliability, switching flexibility, and capability.
With these changes , conventional definite time earth fault protection is not often suit­
able as the primary earth fault protection. The advent of numerical relaying with its event
recording capability has helped to highlight phenomena not normally considered in the
protection literature. Increasingly, distance and differential protection are required for the
system to coordinate effectively.
This paper covers mutual coupling between overhead circuits and how it is seen by
protection devices. It also covers the interaction of NERs with cable networks (especially
wind­farm networks), and makes comments regarding the effect of solidly earthed embed­
ded generation transformers in these now ineffectively earthed sub­transmission systems.

1
1 Introduction
Earth fault protection in a solidly earthed electrical radial distribution system with delta con­
nected loads is fairly simple. The connection of the phase CTs in a Holmgreen1 connection
requires only the simplest of protection devices to detect, time and trip. Although some dis­
tribution line companies in New Zealand apply IDMT (or Inverse Time) to their earth fault
protection timers, the use of definite time timers is common and makes coordination easier to
arrange and document.
The connection of delta connected loads means any current imbalance 2 in the phases must
be a ‘leakage’ to earth and therefore a fault has occurred and the circuit must be isolated. While
there are complicating factors such as very dry soils discouraging the promotion of detectable
levels of current (and the possibility of back feed earth fault via delta connections when single
phase line fusing and auto­reclosing are employed), the conventional earth fault protection based
on the Holmgreen connection of the phase CTs remains the standard choice of protection for most
line companies within their networks. In recent years, the introduction of Petersen Coil systems
has challenged the standard way of doing things and may eventually become widespread.
Radial distribution systems have worked well for many rural and suburban networks. But
right from the early days of electrical networks, planners of our urban areas have looked to create
redundancy of supply through looped distribution networks and the use of the ‘transformer­
feeder’ sub­transmission arrangement, often with line differential over buried pilots.
While electricity becomes even more valuable to commerce and industry, the public also be­
comes more resistant to additional overhead infrastructure. With this in mind, network planners
often look to mesh existing networks using the same corridors to enhance capacity or reliability,
sometimes leveraging off an over capacity in one part of the network to support the emerging
load in another. For the purposes of this paper, the term ‘meshing’ refers to the creation of ringed
or looped networks at the same voltage, that is, without transformers, and that the network is
run as a closed loop, not relying on automated or manual closure to restore supply ­ much like
the existing transmission system.
In comparison to radial systems, earth fault protection, when applied to meshed systems, is
vulnerable to tripping for the following non fault events. Some of these scenarios are not present
within the transmission environment.

1. Open circuits caused by mechanical stresses to overhead jumpers. (Reclose may be suc­
cessful but the open circuit now threatens earth fault protection.)

2. Single phase switching when paralleling feeders (Magnefix).

3. High resistance contacts within Air Break Switches.

4. Mutual Coupling from other over head networks, often under­built or adjacent circuits.

5. Tap Changers out of step on single phase transformer banks.

6. Open Delta regulators in off­nominal tap when paralleling between substations.

7. Network Capacitance when the system is ineffectively earthed.

8. Non­homogeneous earthing of Transpower supply points and embedded generators.

1
See fig. 2 on page 5
2
vectorial imbalance­ not an open circuit

2
 This paper discusses items 4, 7 and 8.

Because overhead sub­transmission networks in New Zealand are almost never run with
aerial earth wires 3 , the fault resistance is high for all types of earth faults, including insulator
flash­over. Therefore sensitivity remains important in meshed networks.
Table 1 shows the expected earth fault currents for downed conductors on a variety of sur­
faces (the data for this table comes from a paper that itself acknowledges a variety of sources of
networks operating at different voltages). However, the table does highlight the small magni­
tude of the earth fault currents and the difficulty of detecting and isolating downed conductors
quickly, even in solidly earthed systems.
The point being made here is that, despite the risk of tripping from non fault conditions,
sensitivity must be retained to isolate downed conductors as far as it is practical.

Surface Fault Current (A)

Dry Asphalt 0
Concrete(non­reinforced) 0
Dry Sand 0
Wet Sand 15
Dry sod 20
Dry grass 25
Wed Sod 40
Wet Grass 50
Concrete(reinforced) 75

Table 1: Typical fault currents on various surfaces[7, Page 9]

1.1 Sequence Diagrams

Understanding this papers relies on the reader having some familiarity with sequence diagrams.
The theory for this is presented in many text books and papers. The author can recommend
J Lewis Blackburn’s ‘Symmetrical Components for Power Systems Engineering’[1] and the
Schneider Cahier Technique ECT018 ‘Analysis of three­phase networks in disturbed operating
conditions using symmetrical components’[3].
As a brief recap, the voltage (or currents) of an unbalanced power system can be represented
as the summation of three phasors, known as positive, negative and zero sequence. Conversion
between actual values and ‘sequence’ values is described mathematically by eqs. (1) to (6) on
page 5. Va0 , Va1 & Va2 are the zero, positive and negative ‘sequence’ voltages. The ‘a’ in Va
shows the phasor angles are with respect to the A phase.
For a single line to ground fault, I1 = I2 = I0 and the current flowing into the ground
is 3.I0 . All conventional earth fault protection detects 3.I0 . As we shall see, 3.I0 can also
be induced into an overhead circuit making the distinction between induced currents and fault
current problematic.
Electrically speaking, 3.I0 is simply the summation of the phase currents from the CTs.
Hence, the Holmgreen connection (see fig. 2 on page 5) is the simple and conventional way
to detect earth faults within a solidly earthed network. However 3.I2 , can only be derived,
economically at least, mathematically by numerical relays.
3
except for short distances as lightning protection

3
(a) Positive Sequence ­ the normal power system con­ (b) Negative Sequence ­ present during 2ph and 1ph
dition faults

(c) Zero Sequence ­ present during 1ph faults and in­

duced current

Figure 1: Positive, Negative and Zero Sequence Network

4
 1
Va0 = [Va + Vb + Vc ] (1)
3
1
Va1 = [Va + a.Vb + a2 .Vc ] (2)
3
1
Va2 = [Va + a2 .Vb + a.Vc ] (3)
3

Va = Va1 + Va2 + Va0 (4)

Vb = a2 .Va1 + a.Va2 + Va0 (5)

Vc = a.Va1 + a2 .Va2 + Va0 (6)

where a = 1∠120◦ in complex notation.

1.2 Conventions in this paper

This paper contains a number of diagrams showing the zero sequence current (3.I0 ), that flows
through the ground and the zero sequence current caused by mutual induction. To assist the
reader, the current that flows in the ground and returned to the system neutrals are coloured green.
The induced zero sequence currents are coloured blue, while the induction ‘Zm’ is labelled in
red. The negative sequence currents are in magenta. In some diagrams, all the currents are
shown in blue but generally speaking there is no contradiction in the convention.

Figure 2: The earth fault sensisitivity of a Holmgreen connection is limited to about 6% of the
phase CT under ideal conditions. [8] [9]. The New Zealand convention is 10­12.5% of ratio.
Otherwise, a core balance CT is required to reliably detect current below this.

5
2 Mutual Coupling
When a 3ph overhead circuit carries a single phase (or 3I0 ) current, the magnetic field set up by
each of the three phases no longer cancel. This magnetic field, at the fundamental frequency,
is capable of inducing voltage (which can become currents) into neighbouring metallic circuits
including telecommunication circuits or adjacent overhead power lines. This effect can occur
over hundreds of metres. Figure 3 shows picture of a typical 33kV line with an under­built 11kV
circuit. Figure 4 on the next page shows a picture of two 33kV circuits running adjacent to each
other ­ the left most poles containing a dual circuit. Even with the road separation, the effect of
induction is significant.

Figure 3: Underbuilt 11kV construction can induce currents into 33kV circuit above.

This effect has been considered in the transmission literature for some time (in the way it
affects distance relaying for earth faults on twin circuit towers) but it was not until the advent of
microprocessor relays with their fault recording ability that the effect became well recognised
within the sub­transmission and distribution environment. Within the transmission realm, this
effect is commonly known as ‘mutual coupling’. Although not an especially precise term for
the kind of inter­circuit induction described here, the term has become common usage within
the New Zealand industry.
As previously radial sub­transmission networks have become meshed, with the intention of
enhancing or improving reliability, this inter­circuit induction (or mutual coupling) has ironically
caused many additional and unwanted outages. The problem has perhaps become more apparent
in the New Zealand context because we tend to share multiple circuits on the same poles and we
use solid or low resistance grounding instead of resonant earthing as is common in Europe.
Mutual coupling can occur between any two adjacent overhead power lines. The circuits
can be part of the same network or part of a different voltage network. For mutual coupling to

6
 Figure 4: Even circuits across the road can induced currents into adjacent circuits.

become a problem, loops including two or more substations within the same voltage network
need to be formed. If a transformer forms part a loop, generally speaking, mutual coupling is
no longer an issue since there is no path for the induced current to flow. This partially explains
the historical success of the ‘transformer­ feeder’ configuration and why the effect of mutual
coupling only becomes apparent when sub­transmission buses are formed.
As a brief introduction into the nature and scope of mutual coupling, the following formula
eq. (7) was derived by a colleague during one of the earlier investigations in mutual coupling.
The formula is based on a reading of Chapter 3 of Westinghouse [4]. It is an approximation but
gives some insight to influence of the various parameters. As a comparison, the same scenario
and parameters were modelled using DigSilent’s PowerFactory. The currents plotted are dif­
ferent but the conclusion remains the same ­ mutual coupling can occur over surprisingly large
distances.
( )
De
0.001257.L.f.3I0 . ln
GM D
I’0 = (7)
ZoL
Where:
• I’0 = induced zero sequence current magnitude (amps)

• L = length of common coupling (km)

• f = power system frequency (hertz)

• 3I0 = residual current flowing in faulted circuit (amps)

• De = depth of ground return current (metres)

7
 • GM D (Geometric mean distance) = equivalent conductor spacing (metres)

• ZoL = zero sequence impedance of the closed ring (ohms)

√
• De = 658.368 × R/f (metres)
Where:
• R = soil electrical resistivity (ohm . metres)

• f = power system frequency (hertz)

• For 50 hertz and 100 ohm m resistivity, De = 931 metres.

• GMD is calculated for six conductors (2 sets of 3 phase circuits) as the product of the nine
unique individual conductor separation distances all raised to the power of 1/9. If each
set of 3 phase circuits was in touching trefoil, with a large circuit separation distance, the
GMD is simply the circuit separation distance.
This equation should only be used for situations where the separation distance is preferably
significantly less than 800 metres [4, Chapter 3].
What is apparent is that separation of conductors does decrease the induced currents but the
effect does not diminish as fast as one might have expected. See fig. 5 on the next page. The
induced current is also a linear function of the loop impedance and fault current.
For the initial case investigated, the following results (see table 2) were obtained and derived:

Result induced residual current (3I0 )

SEL­351A relay fault data 65
DigSilent model 69
Hand calculation formula 68.8

Table 2: Comparison of measured and derived result

Unfortunately, the residual current flowing in the faulted circuit was not available directly
from the fault data (because the 11kV feeder was protected by an old electromechanical relay).
The DigSilent model assumed a 4 ohm ground fault resistance.
Since this investigation, Tesla Consultants have assisted in a number of investigations that
show mutual coupling is common and in some cases the induced currents have been above the
phase over­current protection levels. And in one case, the ‘induced’ currents induced current
into a third circuit with the third circuit’s earth fault protection also operating!
Figures 6 to 7 on page 10 show the induction effect in a more schematic way. Figure 6
on page 10 shows the induction from an adjacent conductor, similar to what might occur in the
circuits emanating from a Transpower Grid Exit Point (GXP). While fig. 7 on page 10 shows the
induction that might occur from an underbuilt lower voltage circuit. The breakers highlighted
in red may be vulnerable to tripping.
Figure 8 on page 11 shows the sequence diagram for figure 6. From here it becomes apparent
the connected load forms an important path for the negative sequence current. If the connected
load is rotating plant, the Z2 impedance can be much lower, increasing the currents and affecting
the relative distribution of currents.
In fig. 6 on page 10 , the zero sequence or conventional earth fault protection makes its
directional decisions based on 3V0 voltage and the 3I0 current. The 3V0 voltage is determined

8
 Induced Current
3I0=IA+IB+IC

d
Note: yaxis is IO

Faulted Cct
3I0=4000A

3.I0

Figure 5: The figure shows the decrease in induced current as the top circuit is moved away to
the right. Even at 800m, the effect is not insignficant.

9
Figure 6: Note the negative sequence algorithm point towards the source. And the 3V0 voltage
is determined by the fault current and not the induced current.

Figure 7: 3V0 , where it occurs, is determined by the induced voltage. In this instance, the
negative sequence current is toward the substation, the opposite of figure 6.

by the fault while the induced 3I0 is determined by the geometric arrangement of the mutually
coupled circuits. In the figure 7, the 3V0 voltage is determined by the induced current and the
Z0L impedance.
The reader might also note in figure 8 that the zero sequence directional algorithm is (largely)
unaffected by load but as we shall see in later sections, the zero sequence capacitance becomes
significant with the introduction of neutral earthing resistors.

2.1 Strategies for coping with mutual coupling

There are two strategies for coping with mutual coupling.

• Accept it and accept breakers will trip. Where possible, configure the system and the
protection to disconnect none or the least number of customers.

10
Figure 8: Sequence Diagram of fault on an adjacent circuit parallel to a looped networks

11
 • Install and configure protection immune to its effect. This generally requires the use of
line differential protection and other unit protections.

Once the problem is understood and modelled, there are a variety of strategies that can be
considered.

2.1.1 Limit the earth fault current

The introduction of resonant earthing or a Neutral Earthing Resistor (NER) will affect the
amount of current induced in adjacent circuits. While a resonant system will be completely
effective, an NER in the ‘Low Resistance’ grounding range may not be sufficient. System stud­
ies will be required.

2.1.2 Implement Fast Autoreclosing

If the breaker prone to tripping can be reclosed quickly, a system check of the voltages either
side of opened breaker will reveal if the protection at the other end has cleared. If not, then the
earth fault tripping may have been for a mutually coupled event and reclosure will be successful.
The author has not seen this done though probably effective in the simplest cases.

2.1.3 Employ 3.V0 voltage checks when the system is earthed via an NER
While not effective for faults on the same voltage network, induced currents do not develop the
same level of 3.V0 as genuine earth faults. Earth fault protection can be supervised such that the
3.V0 must exceed a nominated level before tripping is allowed. This can be a useful method to
avoid sub­transmission trippings/outages caused by faults on nearby transmission lines or under­
built distribution circuits. It is unlikely to be a practical solution on solidly earthed systems.

2.1.4 Use Inverse Time Earth Fault Protection

Because the actual fault currents are typically much greater than the induced currents, theoret­
ically, the earth fault protection could be arranged such that there is time grading between the
actual and induced currents. This is probably feasible in some simple networks and where the
designer has complete flexibility in nominating the types of earth fault protection everywhere.
Often, however, the network interface will involve at least two asset owners and legacy equip­
ment is often not capable. The author has not seen this done.

2.1.5 Install unit protection including line differential throughout the network
Where mutual coupling may cause an ‘N­2’ event (see section 2.1.7 on the following page),
the commonly applied technical solution is to employ line differential protection on the meshed
network and to ensure any faults that could induce apparent earth faults are cleared before any
unqualified definite time earth fault protection may operate. Because earth fault timers are nec­
essarily increased or inhibited, in some instances, bus bar protection must be installed to avoid
sequential clearance scenarios and much wider outages.

2.1.6 Use Current polarised directional decisions

Some relays have the facility to compare the line 3.I0 current with a local transformer neutral.
This provides an opportunity to discriminate between real faults and apparent faults caused by

12
mutual coupling. Care is required if there is more than one solidly earthed transformer at two
or more sites. The Z0 impedance of the earthed star windings may not be sufficient to prevent
induced currents flowing via the system neutrals.

2.1.7 N­1 and N­2 reliability criteria

Useful information for planners
A common strategy to provide an alternative supply to an existing or proposed substation
is to run a circuit from an existing substation that has already has two supplies. In some cases
this may mean converting a ‘transformer­feeder’ to a meshed arrangement, installing an HV
bus with a closed HV bus section. See fig. 9 on the next page. If conventional directional earth
fault protection is installed or retained, it will be vulnerable to tripping for induced currents.
Figure 9 on the following page shows that both breakers looking towards the supply (circled in
magenta) can trip for an induced current from adjacent or under­built circuits. Depending on
the load at the two substations, it may not be possible to supply the original substation from the
second substation, either for voltage or thermal reasons. In effect, the ‘N­1’ status of the original
substation is compromised. The solution is likely to be installing line differential protection on
both the original circuits in fig. 9 on the next page.

13
 (a) Transformer Feeder network ­ immune to mutual coupling

(b) Second substation added. An ‘N­2’ scenario is now possible for a

mutual coupling scenario ­ both the breakers circled in purple will trip.
See fig. 27 on page 38.

Figure 9: Where a ‘Transformer­feeder’ network gets converted into part of a meshed network
through the installation a sub­transmission bus and a tie line to a new or existing substation, it is
important for network planners to be aware that they are now exposing the network to an ‘N­2’
contingent scenario. If this supply scenario overloads the remaining circuits or causes excessive
voltage drop, supply may be lost.
14
3 Network Earthing in New Zealand
The following discussion makes reference to the common sub­transmission voltage of 33kV. If
earthed via an NER, the following discussion is also relevant to 66 & 110kV sub­transmission
systems.
The common types of earthing the author is aware of in the New Zealand distribution net­
works are:

• Solidly Earthed by Dyn11,Dyn3, etc

• Reactance Earthed by Yyn0 ­ the most common 33/11kV configuration4

• Neutral Earthing Resistor (NER), with an earthing transformer if required.

• Resonant Earthing ­ with or without active compensation. E.g. Petersen Coil or Swedish
Neutral.

The technical definition of ‘effectively earthed’ is

0 < X0 /X1 < 3 (8)

0 < R0 /X1 < 1 (9)

From the NZCCPTS 5 guide, the resistance of the NER resistor needs to be within eq. (10)
to limit transient overvoltages. Transpower follow the same criteria in this regard (‘Ro > 2Xo
to dampen capacitive transients’).

2.X0 < R0 < Xc0 (10)

2.X0 < (3 × RN ER ) < Xc0 (11)

So, in general terms, the more network capacitance, the lower the recommended resistance
of the neutral earthing resistor to dampen capacitive transients. For a 33kV network with a
significant cable network, say a windfarm, with say 15km of XLPE at 0.33 µF per km, is the
equivalent of Xc0 = 643Ω.
Two criteria sometimes chosen, not withstanding the above are:

• sizing the NER resistance to match the load of the transformer, that is the prospective
earth fault current is the same as the phase current at full load; or,

• that the prospective earth fault current will be 10 times the sensitivity of the protection
device designed to see earth faults.

In the Transpower realm, the distribution of resistance values within the 33kV North Island
GXPs is presented in fig. 10 on the next page. This data was taken from Transpower’s DigSilent
model. The modern and commonly applied NER value to the 33kV GXPs is 38Ω or that which
will yield a 500A earth fault current 6 .
4
Z0 is typically 0.5 to 1pu. Z1 is typically 0.1pu at ONAN rating
5
New Zealand Committee for the Co­ordination of Power and Telecommunication Systems[6]. See also [9] [1]
6
Modern designs that use an earthing transformer often have sufficient reactance to only just meet the 2.X0 <
(3 × RN ER ) criteria

15
Figure 10: The range of neutral earth resistor value used on 33kV supply buses within the North
Island of New Zealand

Figure 11 on the following page shows the range of prospective faults earth currents (of
course influenced by the method of earthing) and the type of protections required. Transpower’s
NER practice in New Zealand is well within the ‘Low resistance’ grounding realm and as such,
the earth fault protection used remains similar to that used for solidly earthed systems. This
distinction of ‘Low resistance’ is helpful when reading literature from other counties including
Australia. In the New Zealand context, the resistances chosen have not sought to reduce the fire
risk, or to maintain supply during an earth fault.
As aside, some readers might be surprised to know that the Yyn0 with a solidly earthed
11kV neutral configuration, so common in many distribution company networks does not yield
an effectively earthed network. The Z0 impedance of a Yyn0 transformer is often 0.5 pu to 1.0
pu whereas the Z1 is 0.1­0.15 pu. With reference to eq. (8) on the previous page, the X0 /X1 is
approximately 5­10. Figure 12 on page 18 shows the prospective healthy phase voltage rise on
these networks7 .
In recent years, and for a variety of reasons, Transpower have elected to introduce NERs at
the GXPs that supply the various distribution companies. Sometimes they have been introduced
as part of a transformer replacement and Transpower have elected to invert the previous delta­
star configuration to put the star point on the HV side. Two earthing transformers, each with
an NER have been applied to the 33kV network, since the 33kV delta winding has no neutral.
At other sites, and often in conjunction with the replacement of the outdoor bus with an indoor
switchboard (ODID projects), an NER has been introduced into each of the existing transformer
neutrals as part of the overall work.
The benefit to the 33kV networks are:

• Some reduction of step and touch potential

7
When a phase to earth fault occurs on an Effectively Earthed System the phase voltages on the un­faulted
phases should not exceed 80% of the normal system phase voltage.

16
 Figure 11: System Grounding and prospective fault current

• Reduced thermal requirements of cable screen ratings

• Reduced electromechanical and thermal stress for power transformers.

• Reduced Earth Potential Rise (EPR) and induction into telecommunication equipment
8
• Conventional earth fault relaying equipment can be retained.

• Improved power quality with reduced voltage depression during earth faults.

The disadvantage to the 33kV networks are:

9
• Increased temporary over voltages may promote cross country faults if insulation is
already damaged.

• Surge arrestors in the network may need to be replaced.

• Sensitivity, using conventional protection, to high resistance faults is slightly reduced ­

though directional security is enhanced.

• Mho based distance protection for earth faults may become vulnerable to tripping for out
of zone faults.

• Cable earth fault protection may require reconsideration.

17
Figure 12: Ground fault factor as a function ofXo /X1 for R1 /X1 = 0 and R = 0 (graph according
to IEC 60071­2). Note that Yyn transformers are not effectively earthed.

When Transpower introduces an NER into a network that has solidly earthed embedded gen­
eration, the embedded generator transformer becomes the main earthing connection to remote
earth unless NERs are also installed at the embedded generators interconnecting transformer/s.
Unless this is done, the benefits of improved step and touch potential and reduced thermal stress
are moot, in addition to complicating the directional earth fault relaying.
In the authors experience, applying NERs to the generation transformers is generally over­
looked since three independent owners are involved and the generation company is often un­
aware of the changes happening at the Transpower GXP. It is only when protection issues start to
manifest that the issue is reconsidered more seriously. However, if designers have not accounted
for the solidly earthed generation in their earth grid designs or cable screen ratings have been se­
lected based much lower earth fault levels, then the issue becomes more serious as electrocution
and the risk of damaging cable screens increase.

8
Though we shall see in some instances, this is not quite so straight­forward.
9
cross country faults are two earth faults on different phases at different locations

18
4 Cable Capacitance, Windfarms and NERs
A common arrangement for windfarms in New Zealand is to embed them directly into a distri­
bution line company sub transmission network (33kV) without an interconnection transformer.
This economical arrangement is often well suited to the sizing and scale of the existing network
but it does introduce a significant amount of system capacitance. 10
Every cable has charging current. When the V⃗a , V⃗b and V⃗c are balanced, the charging current
Ia , Ib and I⃗c , when vectorially added, cancel. In a solidly earthed system, when Va is depressed
⃗ ⃗
because of an earth fault on an adjacent circuit, the 3.I0 current is the vectorial addition of Ib
and Ic (or −Ia ).
When the system is ineffectively earthed and an earth fault occurs on the A phase, the sys­
tem neutral is shifted. Effectively V⃗a is subtracted from both V⃗b and V⃗c . V⃗b and V⃗c rise and their
angular separation shifts from 120◦ to 60◦ . Now the 3.I0 current is three times as compared to
the solidly earthed system. In most solidly earthed distribution systems, cable capacitance is
almost never considered. Within ineffectively earthed systems, it becomes important to appre­
ciate its possible effect. Equations (12) to (13) on the current page demonstrate the calculated
3.I0 current.
1 1
Xc = = (12)
j.ω.C 2.π.f.C
√
3.Vph−ph
IG = 3.I0 = (13)
Xc0
To calculate the cable capacitance in the absence of manufacturer data, eq. (14) can be used.
2πε0 εR
C= ( ) F /m (14)
ln rrdo
di

where:
ε0 = Relative permittivity of free space is 8.854 × 10−12 F /m
εR = Relative permittivity of the insulation material. About 4 for XLPE.
rdo = Inside radius of the sheath or outside radius of the insulation, if shielding tape is used
(m).
rdi = Radius of the conductor (m).
In many instances this 3.I0 is greater then the conventional earth fault setting. In some cases,
the author has seen this current at over 120A on a 33kV system. This is where care is required
when an NER is introduced into a system with significant cable capacitance. Effectively, every
time a low resistance earth fault occurs, the earth fault protection in the path connecting the
windfarm or significant cable circuit will pickup.
The issue becomes, ‘how to retain sensitivity to genuine high resistance earth fault (40­60A)
but avoid tripping the windfarm whenever an earth fault occurs anywhere else in the system?’
In some instances, time coordination is sufficient. This is covered in more detail in sec­
tion 5.2 on page 44.
Figure 14 on page 22 shows the sequence diagram of the sympathetic tripping scenario.
While the Z2 impedance does not change the sensitivity, the V2 voltage is reduced by the NER
for a given fault resistance. Some consideration of the standing V2 and V0 voltage is required
before determining the setting.
10
This new system capacitance can affect the tuning of load control plant.

19
Figure 13: Addition of voltages in an NER network[3, Figure 34 in reference]

20
 The following discusses the implications on the direction algorithms. Again the discussion
is in SEL11 terms but applicable to conventional directional relaying.

4.1 Directional Decisions

In any meshed network, directional algorithms (or relays) are used to polarise both distance
and directional over­current and earth fault protection. For solidly earthed and low­resistance
grounded networks the relays use the sequence components to determined fault direction. His­
torically, the zero sequence components have been easy to create from the CT and VT connec­
tions but with the advent of microprocessor relays, manufacturers often provide methods based
on the negative sequence components too. Negative sequence has the advantage of being im­
mune to mutual coupling but its use when mutual coupling is present is sometimes problematic.

4.1.1 NERs and Z0MTA

The following discussion describes the SEL algorithm, but the same conclusions can be drawn
for other manufacturer’s algorithm.
The following diagram (see fig. 17 on page 26) show the effect of increasing the NER resis­
tance in a network with significant cable capacitance ­ where 3RG = XC0S is the extreme end
of the NER range. Figure 17c shows where the low resistance earthing (used by Transpower)
sits on this R axis. Figure 17b shows for a conventional Z0MTA angle how the capacitance
current can appear ’forward’ for faults behind the relay. Figure 17d shows how, with a negative
Z0MTA angle, the relay can still correctly discriminate for 3.I0 contributions from capacitive
sources.
For the SEL algorithm, they recommend eq. (15) [5]:
( ( ) )
3.RG
Z0M T A = arc cot + SF A − 90◦ (15)
X0T
where SF A = 30◦ 12
And the result is often a negative Z0MTA setting. Many relays installed within the New
Zealand power system are not capable of this setting. Two configurations where the lack of a
negative Z0MTA angle may affect the system are:

• Fast Bus Blocking Schemes.

• Three Terminal Line with Blocking Schemes.

Many Fast Bus Blocking scheme have been installed within the New Zealand grid. In fact, it
was the maloperation and subsequent investigation of one these schemes that lead to the recog­
nition of inter­circuit induction or mutual coupling.

4.1.2 Fast Bus Blocking Scheme (FBB)

If an NER is installed on a bus with a fast bus blocking scheme and there is sufficient cable
impedance, unless the Z0MTA angles are reconsidered, the capacitive circuit will always block
the FBB scheme by always declaring a forward fault, regardless of the earth fault location.
11
Schweitzer Engineering Laboratories
12
Security Factor Angle

21
Figure 14: Sequence Diagram of fault on an adjacent circuit

22
 Also note that with some directional earth fault algorithms, the relay may require additional
processing time before making a directional decision. This needs to be factored in any FBB
scheme design operating near its limit of sensitivity. Given any capacitive current is likely to be
close to the typical EF sensitivity, this processing time may be critical. Another solution may be
to desensitise the FBB scheme above the maximum capacitive current experienced. Given that
a genuine bus fault is unlikely to have much fault impedance, this may be acceptable.

4.1.3 Three Terminal Line Blocking Scheme

Similarly, a three terminal line directional comparison scheme may require special consideration
since earth faults within the three terminal line or in the windfarm will be seen as reverse to
the relay connected on the windfarm side. See fig. 15. Since three terminal lines often use
blocking schemes, careful consideration of the sensitivity and/or directionalising parameters
may be required.

Figure 15: If a windfarm is part of a three terminal network within a low­resistance earthed
network and a conventional 3.V0 or a ORDER=V directional method is used, the earth fault
protection closest to the windfarm will see the windfarm capacitance as a reverse fault and block
regardless of the fault location (unless there is sufficient fault impedance to limit sympathetic
earth fault current).

4.1.4 NER and Negative Sequence Directional Control

Whenever an NER is used within a network, the prospective earth fault current is markedly
reduced. With reference to fig. 14 on the preceding page it becomes apparent that the negative
sequence voltage is very much reduced at the relaying point ­ possibly below the standing 3.V2
voltage on the system. If, at the limit of the desired earth fault sensitivity, there is insufficient
V2 voltage, it is desirable to use the V0 based method for directional control, since there is much
more 3.V0 voltage across the NER. Section 4.1.4 on the next page shows a high resistance fault in
an ineffectively earthed network and the 3.V2 voltage being insufficient to polarise the directional
relay.

23
Figure 16: During a high resistance fault, there is insufficient V2 to polarise the directional earth
fault protection. The V0 is easily adequate.

For relays that use both the V2 and V0 methods, it may be sufficient to switch algorithms
when the negative sequence current × the source impedance is below the standing V2 . This
could be arranged by amending the current sensitivity of the V2 algorithm.

4.1.5 Solidly earthed embedded generators within an NER network

There are many embedded generators installed within the sub­transmissions system of New
Zealand. Often times these were the original hydro generators that spawned the regional lines
companies until they were connected into the larger national grid (formerly State Hydro, NZED,
ECNZ and now Transpower). Following British convention, these generators were solidly
earthed.
As Transpower has introduced NERs at their GXP substations, these embedded generators
have often remained solidly earthed. This has had the effect of mitigating much of the benefit of
installing the NERs. In our example case, we show that with the Transpower transformer solidly
earthed, the fault contribution from the embedded generator is 3,858A and the Transpower trans­
former 11,624A per transformer. With NERs installed at Transpower, the embedded generators
contribution leaps to 10,373A while the Transpower transformers now contribute only 440A per
transformer. Transpower might have expected the EF level to have dropped to 1,220A with the
introduction of the NERs. The thermal stress on the Transpower transformer is partially reduced
but not to the level hoped for (the 3.I2 current remain high for earth faults) and the generation
transformers see a threefold increase!
Furthermore, these embedded generators are sometimes switched between networks. Fig­
ure 18 on page 27 shows how it is important to select the correct torque angle. For a resistively
earthed network, an Z0MTA angle of between 5 − 10◦ is desirable13 . Whereas Z0MTA for a
solidly earthed network would be normally be around 50 − 80◦ . If the Transpower supply is
13
for some earthing transformer + resistor designs, the transformer reactance is maximised (but still meeting
eq. (11) on page 15) such that a conventional 63◦ angle is still appropriate. Watched though for mixed vector
groups though!

24
earthed via an NER and the solidly earthed embedded generators are switched in and out, there
is no guarantee that a Z0MTA angle, that will suit both configurations, can be found. This may
lead to unwanted outages as backup protection is forced to operate or there is complete loss of
sensitivity.
The solution here is for the embedded generator to have an NER or NERs installed. If the
generation stations are linked via an overhead line (as they often are in run­of­the­river schemes),
it should be possible to run a fully insulated aerial neutral wire to common the stations neutrals
before connection via a single NER to earth.

25
(a) The measured zero­sequence impedance – Z0S (b) The conventional 32V element fails to discrimi­
moves from the third to the second quadrant of the nate forward from reverse faults in low­impedance­
impedance plane as the grounding resistance varies grounded systems[5].
from zero to 3RG = XC0S [5].

(c) The same as figure 17 b but showing the Z0 lo­ (d) Rotating the 32V element characteristic pro­
cus of forward faults in a typical New Zealand based vides directional ground fault discrimination in low­
33kV subtransmission system with a 34Ω resistor[5]. impedance­grounded systems[5].
The R0 = 3 ∗ N ER.

Figure 17: Z0 impedance locus for earth faults in an ineffectively earthed network

26
(a) If the Z0MTA is set to line angle in a resistively (b) When the Z0MTA is reduced below 16 degrees, ca­
earthed network, neither forward or reverse faults may pacitive faults will present as indeterminate. This does
be detected not preclude the Q algorithm making a decision

(c) When the Z0MTA is reduced to say 5 degrees, the (d) If however, the system becomes solidly earthed by
reverse capacitive current will not be seen in reverse. another star point, say an embedded generator, neither
the forward or reverse faults may be detected if the
Z0MTA angle remains at 5 degrees.

Figure 18: The correct Z0MTA angle is important to discriminate forward and reverse faults
correctly. Many older SEL relays have a minimum Z0MTA angle of 40◦ and may not be suitable
(as may other relays with limited MTA angle) when an NER is installed.

27
5 Hypothetical Network ­ Walk­through using DigSilent
The following walk­through demonstrates the development of a hypothetical network.
The intention is show how mutual coupling puts a meshed network at risk and how the
addition of an NER at the Transpower GXP will influence the existing solidly earthed generator
and the windfarm with significant cable capacitance.
Figures 19 to 24 on pages 29–34 shows the stages. These are worked through again in more
detail in section 5.1 on page 35 in figs. 25 to 32 on pages 36–43

28
Figure 19: Initial Substation in a ‘Transformer­Feeder Arrangement’.

29
Figure 20: Adding the second substation including the mutual coupling with the embedded
generator connection circuit.

30
Figure 21: Adding the second substation and adding the embedded windfarm.

31
Figure 22: Adding the second substation but also including the embedded windfarm and the line
to the embedded generator.

32
Figure 23: The Transpower GXP transformer have been changed to YNd with an earthing trans­
former and a 31Ω neutral earthing transformer .

33
Figure 24: The YNd step up generation transformer now has a 100Ω earthing transformer.

34
5.1 Detailed worked examples
Following figure captions to follow what is being presented.

35
Figure 25: A transformer feeder network ­ immune to mutual coupling. An 11kV earth fault is
placed at the end of the 5km circuit. Note there is no 3.I0 current in the 33kV network.

36
Figure 26: The second substation is built. This same 11kV circuit is now overbuilt by the second
33kV line supply to Sub B. The subtransmission system is now meshed.

37
Figure 27: When faulted, the 11kV EF can induce 631A in each of the 33kV breakers looking
toward the source. Conventional directional earth protection will trip, leaving the SubA supplied
by SubB only. If not adequately rated, the circuit to Sub B could overload or the voltage become
inadequate for Sub A.

38
Figure 28: Transpower is solidly earthed. Earth fault level is 25,918A. An earth fault will see
38A of 3.I0 in the windfarm feeder ­ below normal tripping level

39
Figure 29: Transpower and the embedded generator’s are solidly earthed . Earth fault level is
27,070A. The embedded generators contribution is 3,858A.

40
Figure 30: The embedded generator’s fault contribution is sufficent, in this example, to induce
200A into the meshed sub­transmission network.

41
Figure 31: The Transpower transformers are changed to a YNd vector group with the same
impedance and the 33kV bus is earthed via NERs. The embedded generator contribution in­
creases 3 fold to 10,373A. Despite the Transpower NERs, the Transpower transformers still
experience 3,569A per transformer for what should have been 600A. The induction with the
meshed network increases markedly with current level reaching phase overcurrent levels! The
capacitive sympathetic earth fault current has increased to 93A ­ above the nominal earth fault
pickup level of 60A.

42
Figure 32: The embedded generator now has an 100 ohm NER installed. The earth fault level
is now 1,414A, much closer to the anticipated design level. The generator contribution to earth
faults remains detectable and the induction into the adjacent circuit is now trivial. The step and
touch potential will be markedly improved and future cables will no longer require fully rated
screens. Of note, the windfarm sympathetic earth level is now at its maximum at 129A. Earth
faults on the system will cause minimal voltage dip now.

43
5.2 Strategies for Sympathetic Tripping
When directional solutions are not available, time coordination can be an option. See fig. 33 on
the next page shows how a two stage definite time earth fault protection can be configured to
ensure sensitivity to genuine faults but avoid the risk of ‘sympathetic’ tripping for faults on the
adjacent circuit. See Figure 14 on page 22 for the sequence diagram. A similar effect can be
achieved using inverse time over current IDMT protection. System studies would be required
to ensure coordination with all other protection devices if IDMT is used.

44
 (a) Current vs trip time

(b) Earth currents in both the faulted and adjacent circuit responding to 0 Ω
and a high resistance fault at the limit of the earth fault sensitivity within a low
impedance but ineffectively earthed network.

Figure 33: Example of using two step definite time coordination for sympathetic tripping

45
6 Neutral Voltage Displacement (NVD) and Back fed earth
faults
This section is included because NVD is the only effective means of detecting downed conduc­
tors when a conductor is livened from voltage source without a neutral connection connection to
ground. This is a very real possibility in all ‘transformer­feeder’ networks. Some line companies
within the industry have proposed that negative sequence current protection and line differential
protection can detect these faults and are basing future network rationalisations on the removal
of end­of­life NVD protection. This approach is not recommended especially within urban net­
works.
Figure 34 shows a scenario that can occur in an any overhead network. That is, the overhead
circuit is broken and the conductor that comes into contact with the earth is back livened from
the down stream transformer.
While earth fault protection at the source (left) end may operate, the returning earth fault
current is very dependent on the connected load. In this example, 2MVA of spinning load is
required to allow 120 amps of earth fault current to flow. If the load was less, or resistive, or
the fault had some resistance, the earth fault protection at the source end would not see this
fault (neither would line differential, nor negative sequence over­current). Only NVD at the
line company end would see this fault (at no load) and trip the transformer HV breaker.
Figures 35 to 36 on the following page shows the more typical ‘transformer­feeder’ scenario
operating in parallel. There is significantly more fault current from the source (left end) but note
that the current through the line company transformer’s HV breaker does not measure any 3.I0
current and hence the transformer protection will not trip before the source end breaker opens.
Unless there is inter­tripping or unit protection , the supply scenario will end up as in fig. 36
on the next page, resultingin a hazardous live conductor on the ground. Only neutral voltage
displacement will detect this scenario.

Figure 34: Current based earth fault protection is dependent on the connected load. Neutral
Voltage Displacement protection is the only kind of protection capable of detecting this condition
without load

46
Figure 35: Neutral Voltage Displacement protection is the only kind of protection capable of
detecting this condition without load

Figure 36: After the source end opens, unless there is intertripping, Neutral Voltage Displace­
ment is the only kind of protection capable of detecting this condition. Note there is no apppre­
ciable earth fault current in the line company’s circuit.

47
6.1 Directional algorithm selection
The following is more of an anecdote but illustrates the dynamic nature of directional algorithms
common in many modern relays and how it can operate within a non­homogeneous network. 14
To paraphrase an old quote;

“ Wisdom comes from experience and experience often comes from bad judge­
ment ” – variously attributed .

The following example is from a relay event report kindly passed onto me from a former
colleague and is for a scheme I had designed many years ago . The scheme was a POTT scheme
using mho distance with directional comparison for high resistance earth faults and implemented
using SEL311B distance relays. There were two lines in parallel though they ran on entirely sep­
arate routes. One bus was connected to a set of solidly earthed embedded generators. The other
bus was connected to the Transpower GXP via another set of parallel circuits. The Transpower
GXP was ineffectively earthed using two NERs. Figure 37 on the following page is a simplifi­
cation of this network.
At the non­embedded generator end, I decided to remove zero sequence polarising algorithm
and rely entirely on Q, the negative sequence algorithm. There had been quite a few trippings
in previous years due to mutually coupled under­built circuits (though the legacy relays didn’t
have the capability to prove this and mutual coupling was still a mystery to me at the time) and
so when I did the settings, I thought using only Q might resolve this problem ­ Q being the new
shiny thing.
Years later, an event report was sent to me showing how both circuits of the parallel line had
tripped. Both ends of the Directional Comparison Earth Fault scheme had declared forward and
both lines tripped. Because the location of the fault was known, it was much simpler to focus
on why the healthy circuit had tripped. Figure 37 on the next page shows the sequence network.
Because the system was not homogeneously earthed, the distribution of the negative and
zero sequence currents was uneven. In fact what had happened was that the fault location was
in the sweet spot of the V2 voltage being roughly equal in magnitude and phase angle. There
was little I2 current flowing in the healthy circuit.
With Q being the first algorithm to consider, both relays did their a2 = II12 > 0.1 and
k2 = II20 > 0.2 checks. With I0 being much larger the k2 ratio wasn’t satisfied and so the relay
with ORDER=QV switched to V. The relay set with ORDER=Q couldn’t switch and, by default,
discarded the k2 criteria and made a decision on Q regardless. See fig. 38 on page 50.
Because the NERs limited the Transpower (left end) contribution, the I0 from the solidly
earthed embedded generators went ‘around the corner’ and so, the relay with ORDER=QV made
a forward EF decision on the healthy circuit, whereas the facing relay (where ORDER=Q only)
made a forward EF decision too, based on the small amount of I2 in the circuit. If this relay had
V available, it would have made a reverse decision and the DCEF/POTT logic would not have
tripped the healthy circuit. See fig. 37 on the next page.
The lesson from the story is: if setting a Directional Comparison Earth Fault scheme with an
SEL relay or another relay capable of automatic selection, ensure both ends are equipped with
the same parameters. And in regard my original concern about avoiding nuisance tripping by
having V in the ORDER=QV equation, the very fact of having Q first in the order may well
have resolved those concerns since any faults on the under­built 11kV circuits (supplied from
the left bus via 33/11kV transformers) would have had the Q based directional algorithm seeing
reverse.
14
A system is non­homogeneous when the source and line impedance angles are not the same.

48
Figure 37: SLG fault sequence diagram for midline fault within parallel lines. Transpower is
earthed via an NER, while the right most generator is a solidly earthed. The supply between the
Transpower source and the left most bus is somewhat simplified

49
Figure 38: An excerpt from SEL’s automatic directional algorithm selection logic

50
7 Conclusions
The following advice is not exhaustive but may be useful to other engineers. Experts in the field
may form other technical opinions or preferences but where there is lack of understanding, I
hope this paper informs or at least provides a starting point for the technical issues raised.

7.1 Checklist when Transpower is installing NERs

• Revisit the earthing arrangement of any embedded generators and consider installing
NERs within their transformer neutrals. 15

• Determine the charging current of all cable circuits; and,

• reconsider the earth fault sensitivity; and,

• reconsider the earth fault relaying angles where directional protection is applied to any
circuit.

• Revisit all surge arrestor ratings, and replace if necessary.

• While negative sequence polarising is desirable; for very strong sources with an NER, the
negative sequence voltage generated at the limit of sensitivity may be too low. Polarising
with 3.V0 will be required if other measures are not taken.

• If no NERs are to be installed at the embedded generation sites, reconsider all earth fault
protection associated with the network.

• Often bus zone protection sensitivity is set at load (or above) to mitigate the risk of mal­
operation for inadvertently shorted CTs. In an ineffectively earthed system, this could
mean the bus zone protection will no longer operate for phase to earth faults. The remedy
is to set below the lowest credible earth fault level (500A for Transpower supplied 33kV
system ) × a safety margin (0.5 (?)) and possibly add voltage supervision of the bus zone
trip decision and especially so if the CTs are shared with other protection. The default
slopes of the operate/restraint region are unlikely to be adequate.

• Unless the line VT are rated to handle 1.9pu, they may saturate during earth faults on
the system. Saturation increases their excitation current and may possibly blow primary
fuses. Any protection reliant on this may fail to operate as intended.

7.2 Advice for Embedded Generators

• Make plans to install NERs at all sites. If the Transpower GXP changes to low impedance
earthing, the generation transformer/s will become the main contributor to earth faults on
the network, significantly increasing the thermal and electromechanical stress to genera­
tion plant.
15
Recognising a number of neutrals can be tied together before earthing at one NER.

51
7.3 Advice for Lines Companies
• When NERs are installed at Transpower, all embedded generators should receive NERs
too.

• Any overhead sub­transmission circuit capable of back­feeding a fault via transformer

without a neutral connection to ground should have Neutral Voltage Displacement, espe­
cially in urban areas.

• Any windfarm or significant cable capacitance needs to be re­assessed if an NER is to be

installed.

• Changing a network from solidly earthed to low impedance grounding may not require
new CTs or VTs but any directional protection will need to be revisited.

• If mutual coupling is more than a nuisance, line differential protection may be required.

• If the lines company is already considering the possibility of using resonant earthing in
the radial under­built distribution systems, the installation of a resonant earthing system
may resolve some mutual coupling problems.

7.4 Advice for System Planners

• When re­visting a ‘transformer­feeder’ set­up, consider the possibility of mutual coupling
between existing and proposed circuits

• When ‘N­2’ is a possibility because of mutual coupling, secure these circuits with line
differential protection first.

• Consider fast auto reclose if line differential is too expense and/or the network can accom­
modate temporary line outages with ease ­ i.e. maintaining voltage or temporary overloads
is not a problem.

7.5 Other comments

The use of modern power system software such as DigSilent’s PowerFactory that can simulate
mutual coupling is invaluable to verifying the currents and voltages the power system might
reasonably expect under fault conditions. Present day sub­transmission systems often have a
complex historical development. Circuits are frequently spliced and reconfigured and it can
take some effort to properly establish the potential for mutual coupling. Once modelled, the
protection designer can have some confidence in the proposed settings.
Interestingly, during my reading for this paper, I discovered a number of van Warrington
arc resistance formulas are quoted incorrectly in a number of publications, each different [6],
[9], [2]. The GEC Protection and Relay Application (PRAG) guide has correctly converted the
imperial units to metric from van Warrington’s original. And since it is being kept up to date,
the NZCCPTS guide may need revisiting.
Lastly, thank you to my colleagues at Tesla for sharing their knowledge over the years and
sponsoring my time to prepare this paper.

52
 References
[1] BLACKBURN, J. L. Symmetrical Components for Power Systems Engineering. CRC Press,
1993.

[2] COUNCIL, E. Power System Protection: Systems and Methods, second ed. No. 1. Peter
Peregrinus Ltd, 1981.

[3] DE METZ­NOBLAT, B. Analysis of three­phase networks in disturbed operating conditions

using symmetrical components. In Cahiers Techniques, no. 18 in 1. Schneider Electric,
October 2005, p. 25.

[4] EVANS, R. D. Electrical Transmission and Distribution Reference Book, fourth ed. No. 1 in
2. Westinghouse Electric Corporation, 1964.

[5] LAVORIN, R., HOU, D., ALTUVE, H. J., FISCHER, N., AND CALERO, F. Selecting Directional
Elements for Impedance­Grounded Distribution Systems. SEL (2007).

[6] O’BRIEN, M. Neutral Earthing Resistors or Reactors, 3 ed. The New Zealand Committee
for the Co­ordination of Power and Telecommunication Systems Inc., 2010.

[7] RUSSELL, D. B. D. Detection of downed conductors on utility distribution systems. IEEE

(1989).

[8] SIEMENS. Siemens PTD EA Applications for SIPROTEC Protection Relays, 2005.

[9] WILKS, J., Ed. Electrical Protection: Limits to Reliable Operation (2003).

53

View publication stats