Important measurements that support IR surveys in substations

By
Robert Madding
Infrared Training Center
FLIR Systems,
Ken Leonard
Carolina Power & Light
And
Gary L. Orlove
Infrared Training Center
FLIR Systems
ABSTRACT
You have found a hot spot in a substation with your IR camera. You know it is a problem, and you must now
determine just how bad it is. Does it need to be fixed immediately, or can it wait a few days, a few weeks? What
other useful information could you measure while in the substation? Two very important measurements are load
and wind speed. Both strongly affect the apparent severity of a problem. In a ring-bus, or parallel bus system,
knowing the current in each side is crucial to making the proper call. This paper discusses making these
measurements safely using inexpensive instrumentation that is currently available.
Keywords: Power loss, load correction, IR surveys, infrared, thermography, substation thermography, wind
correction.
1. INTRODUCTION
Substation IR surveys are an efficient means of identifying problems under energized operating conditions. There
are hundreds to thousands of connections in a substation depending on its size. Making individual measurements on
each one would be prohibitively time-consuming. Modern infrared cameras do a great job of spotting problems
when in the hands of the well-trained thermographer, one who understands direct and indirect measurement,
emissivity, distance, load, environmental and other parameters that impact what the thermographer sees through the
eyes of the IR camera. But not every problem needs fixing right away. Some can wait; others cannot. After
acquiring good IR data and finding problems, their severity is the greatest challenge facing the thermographer.
Experience goes a long way in the problem severity arena. But measurements that support the IR data can be very
useful in assessing how bad the problem is. Load is crucial. A 45 F temperature rise, direct reading, under 50%
load is a much more severe problem than the same temperature rise under 90% load. This temperature rise
measured in a 10 mph wind is much more severe than in no wind. Both wind and load effects are discussed in detail
in other publications (1,2, 3). Other environmental parameters such as solar load, air temperature and humidity play
lesser, but still important roles.
2. SAFETY CONSIDERATIONS
People who walk in the woods have no idea of the power stored in the mass of a tree. A lumberjack felling such a
tree gets a real hint when the tree hits the ground, the earth shakes and a large, long dent is left behind. One who
cuts this tree into smaller sections for firewood and carries them out by hand gets a real feeling for the tremendous
mass of even a fairly small tree. They look harmless, just standing there. But if one falls on you, you’re dead.
Electric power also appears harmless. Its not snarling, roaring or hissing, all prehistoric danger signals we humans
have learned to recognize for our survival. Electricity is just there, not obviously dangerous. We must learn through
training how to stay alive in this seemingly benign, but potentially deadly environment. High voltage is
unforgiving. You don’t get second chances.

InfraMation 2002 ITC 035 A 2002-08-01

For substation work, never enter a substation without being accompanied by qualified personnel. Always have a
tailgate meeting before starting your IR survey. Learn and follow their rules. Discuss their problem severity
criteria. Wear proper safety equipment. Hard hat and safety glasses are universal. Some utilities require long sleeve
cotton shirts and leather shoes. Others may only require short sleeves. Find out clothing requirements before you go
there.
Use your senses to get a first reading on the health of the substation. Sniff for oil, listen for arcing, look for obvious
problems, such as a wire on the fence. Use your IR camera to get an initial overview of the substation. Do a quick
look for major problems. If you find a critical problem, you may need to “freeze and leave”. That is, freeze and
save the image, then leave the area to a safe location. The substation electrician or qualified host will want to
contact the proper personnel to take action on such a critical problem. What you do not want to do is ask everyone
to “come and look at this thing that is about to blow up!”
Usually, you don’t want to extend anything above your head. We often do this to measure wind speed as the wind
speed can be different at the target. Get permission from the substation qualified person as to where you may
perform such activities. Let the qualified person open control cabinets and operate any controls or buttons. (Often,
loads can be read out digitally if you know how to access the data.)
Learn how to do IR surveys in substations. This paper is an adjunct to basic substation IR survey training. You
should be able to identify the components, have a basic understanding of their function and purpose, know what to
look for, know where the indirect measurements are, have a fundamental understanding of equipment criticality, and
know a dangerous situation when you see it.
3. MEASURING AND CORRECTING FOR WIND SPEED
The Kestrel 3000 wind meter, Figure 1, is used by many
thermographers to get wind speed, air temperature, and
relative humidity. It also gives dew point temperature, wind
chill and heat index. It is a pocket size instrument and sells
for about $150. Wind speed can dramatically affect hot spot
temperature and associated temperature rise. A 3 mph wind
can cut the temperature rise in half from no wind conditions!
A simple wind speed correction factor doesn’t exist as the
correction depends on the watt loss of the hot spot.
Correcting for wind speed on “direct” targets can be done
using Figure 2, taken from reference 1. To do this accurately,
you would multiply the resistance of the directly read hot
spot by load in amps squared to get the watts of power (I2R)
being dissipated as heat. Figure 2 gives three power values,
7.9, 18 and 27 watts. You could roughly interpolate between
these values. For example, for a resistance of 250 micro-
Figure 1. Kestrel 3000 pocket weather meter.
7

6
ohms, 230 amp load, the power loss is
13.2 watts. For a 10 mph wind, the
5 correction is about 3.5. You could
Correction Factor

4
multiply the measured temperature rise
A p p lie d P o w e r by this value to get the corrected value.
3 7 .9 W a tts
1 8 W a tts The following considerations are why it
2 2 7 W a tts
is difficult to get accurate temperature
rise corrections under windy conditions:
1

¾ Considering the uncertainties in

0
0 5 10 15 20 25 30 emissivity and background,
W in d S p e e d in m p h your initial temperature
estimates can have fairly wide
Figure 2. Wind speed correction factor vs. wind speed for error bars. Does it make sense
various power levels. From reference 1.

InfraMation 2002 ITC 035 A 2002-08-01

 to go to a lot of work to accurately correct an already uncertain measurement?
¾ Load and wind speed are related. The correction depends on load. Do you correct for existing load, or do
you correct for maximum load?
¾ Knowing the wind speed at the component is also problematic. The Kestrel does a great job of measuring
at its location. But the wind can vary significantly from that point to the part in question.
¾ What wind speed value do you correct to? You need to correct to the minimum wind speed expected under
load conditions, not necessarily zero wind speed. If the wind is always blowing, and there are areas where
this is the general case, what is the lowest value?
¾ Getting the connection resistance is problematic. And once you know the connection resistance, you
already have enough information to make a call on severity.
Our heads were spinning when we realized all the drawbacks to wind speed correction. Trying to get accurate wind
correction on these measurements doesn’t necessarily make a lot of sense. But multiplication factors of 3 or 4 are
not small corrections and need to be recognized at some level. Familiarize yourself with the chart. Get some idea of
the minimum wind speed in your area. Recognize that wind can play a major role in cooling a component. Try to
get measurements under minimum wind conditions. Avoid doing thermography in high wind conditions.
Should you estimate a wind speed correction every time? No. If the temperature rise is already fairly high, and the
load is nominal, you know that at full load and no wind, the temperature rise will be significantly higher. Or, if the
temperature rise is already at the critical level, the correction cannot push it to a higher level.
Wind speed correction is useful for those nominal temperature rises, like 30º F. This doesn’t sound too bad for a
direct measurement. But if measured in a 10 mph wind, the corrected value ranges from about 90º to 120º F, if you
are correcting to zero wind speed. It is often very important to make a wind speed correction estimate.
4. LOAD MEASUREMENT
Load (current) is very important for the thermographer. Heat
is generated in a bad connection by electric current flowing
through a resistance. The power loss can be calculated as
shown above. It is directly proportional to the square of the
load. But temperature rise is not directly proportional to the
square of the load. It follows a more complex pattern as
discussed below.
But if you don’t know the load, you can be completely in the
dark about problem severity. Again, if you find a very high
temperature rise, one that exceeds the critical or upper limit
of your severity criteria, you don’t need to know the load.
For problems that uncorrected are not critical, knowledge of
the load is crucial. You need to know both the actual load
and the maximum load to do a proper evaluation. Figure 3. LCD readout of amperage on an OCB
For a ring bus problem, measuring the load can save control panel.
someone’s life. Unfortunately (or fortunately for the utility),
we did not find a ring bus problem to give as an example in this paper. Earlier work (4) prior to our ability to
measure the load directly in the switchyard, gave an example where 1200 amps entered the ring bus, and by
calculation we determined 1050 amps flowed through one leg and only 150 amps flowed through the other. If the
operators did not know this, they could well open the “good” side and attempt to force all 1200 amps down the
“bad” side. The measured resistance of the bad connection that shunted most of the current down the other side was
96,000 micro-ohms. Opening the good side would cause 1200X1200X0.096=138,240 watts to be dissipated in the
bad connection. The result would have been catastrophic. We recommend never opening one leg of a ring bus
when you know you have problems, without first measuring load in both sides of the bus.

InfraMation 2002 ITC 035 A 2002-08-01

 How do you get load in a substation? Often,
control cabinets have a readout that gives the
current for each phase. Figure 3 gives an
example. Or, you may be able to contact the
control room and get a load reading from them.
Basic electricity says that all the current that
flows into a wire flows out of the wire, so this
value is valid as long as the conductor doesn’t
split, as in the ring bus example above. When
this happens you cannot assume a 50/50 split in
the current flow. It depends on the resistances
in each side of the parallel circuit.
For these cases and when no readout is
available, you can make your own measurement
with a portable ammeter on the end of a hot
stick. We used the Ampstik shown in Figure 4.
The ammeter attaches to the end of a
telescoping hot stick. The reading is stored and Figure 4. Close-up of the Ampstik and in use to measure one side
read out when retrieved. Newer models will of a switch.
store several readings so you don’t have to
retrieve the meter each time. Figure 4 shows a
typical measurement. Proper training is required for measuring high voltage live line current. Only qualified
personnel should do this!
Figure 5 is the IR image of the B-phase and C-phase of a 23 KV line side disconnect. The switches have an inverted
Y configuration, so there is a parallel path between the hinge and the jaws. In both cases the left side hinge is hot
while the right side is cool. We measured the current through each side of each phase and found almost all the
current was flowing through the left side of both the B- and C-phase switches. We interpret this as both sides of the
hinge are bad on both switches. The right side is much worse than the left side, forcing most of the current through

137.4°F

C-Phase Max Temp*: 141.4°F

B-Phase Max Temp*: 157.5°F 120

100

Normal Temp: 92.0°F

80

72.7°F

Figure 5. Thermogram/photo pair showing problem line side 23 KV OCB switch disconnects on B- and C-phases.

the left side. We also found the current beginning to become imbalanced in the A-phase switch which may indicate
an impending problem. We recommended all three switches be repaired or replaced.

InfraMation 2002 ITC 035 A 2002-08-01

 Once you have determined the load, you need to find what percentage of maximum load it represents. To do this,
divide the measured load by the maximum load and multiply by 100 to get per cent.

23 KV Oil Circuit Breaker Line Side Switches

Phase Left Side Right Side Total % Load % Load
Amps Amps Amps Normal Max Two Feeder Max
275 Amps 525 Amps
A 69.8 129.5 199.3 72% 38%
B 195 1.8 196.8 72% 37%
C 193 2.0 195 71% 37%
Table 1. Measured current on inverted Y switches, each leg and total.

In this case, we must consider two maximum loads. One is a “normal” maximum that can occur daily during hot
weather. The other is a “two feeder” maximum where due to other problems two OCBs are fed through these
switches. In the following section, we will use this as an example for load correction of temperature rise.
5. LOAD CORRECTION
There have been many attempts to derive a simple load correction factor where the temperature rise for a known
load is multiplied by a number to give the full load temperature rise. In this way, the severity of the problem can be
evaluated for full load conditions. Or, one could calculate the load that would give the maximum temperature rise
for safe operation.
The power dissipated as heat equals the square of the load times the electrical resistance (P=I2R). One might then
expect the temperature rise to vary as the square of the load. If the load doubled, the power dissipated increased by
a factor of 4, and the temperature rise increased likewise. For heat transfer by conduction alone, this is valid. But
heat transfer by radiation depends on temperature to the fourth power (T4 Stefan Boltzmann Law). Convection as
well is dependent directly on temperature rise only in limited regions. Both play an important role in electrical
problem heat dissipation. The square assumption predicts a much too high rise, and should not be used (2, 3).
One might assume the temperature rise (∆T) increases as some power of the current and so attempt to fit ∆T=InR,
where the current is raised to the nth power, n being determined by fitting data or modeling. Perch-Nielsen et al (3)
did some experiments and found n varied between 0.6 and 2 depending on conditions.
We used the power loss software (5) to calculate the effects of changing load for various emissivity targets, air
temperatures and background temperatures. We fit the resulting data to the above equation to find n for various
conditions. Our modeling shows there is no single value for n for all cases. There is no simple load correction
factor. We found a range of exponents (n-values) from a high of 1.6 to a low of 1.46. Our modeling considered a
low emissivity target (minimum of 0.6) to a high emissivity target (maximum of 0.95) with an air temperature of
about 70 F and background temperature ranging from 70 F to -13 F. The simulated target was 4” wide by 1” thick
by 6” tall. The value of n decreases as heat transfer by radiation increases due to its non-linear, T4, nature.
Strongest radiation occurs for high emissivity, low background targets.
Figure 6 shows our modeled estimates for limiting conditions that will give the thermographer a reasonable idea of
the multiplication factor to use for full load correction (values of n are given in the legend). These are “middle-of-
the-road” results compared to others referenced above. But they do represent the bounds of what we found with our
modeling. The factors are not small, and like the wind correction can make a huge difference in your severity
estimates. Figure 6 also shows that performing IR surveys under low load conditions leads to greater uncertainties
in predicting the temperature rise at full load conditions.
As the graph is in log-log format, it is a bit tricky to estimate the values between gridlines. For those mathematically
inclined, you can calculate the value as follows: Multiplier = (1/%load)n. For example, at 50% load, the maximum
correction (n=1.6) would be (1/0.5)1.6= 21.6=3.03. Many calculators have the yx function to do this type of
calculation. But as these are estimates, it is simpler to look at the chart and estimate the value.

InfraMation 2002 ITC 035 A 2002-08-01

Let’s use the load
correction chart to
estimate the
temperature rise of 10 10
the hot switch 9
components 8
7
shown in Figure 5.
6
The B-phase

Multiplication Factor
shows a 65.5 F 5
rise, the C-phase a 4
39.4 F rise. Wind
speed was 0 mph.
3
For the normal full Most Cases (n=1.6)
load condition, the High e, low Bkg (n=1.46)
load is about 70%. 2 Wind speed 0 mph
The correction
factor read from
Figure 7 is about
1¾, so the normal
full load corrected 1 1
temperature rises 100% 80% 70% 60% 50% 40% 30% 20% 10%
would be about % Load
115 F and 69 F,
respectively. For Figure 6. Load Correction. Multiply temperature rise by factor
the two-feeder full given at measured % load to correct to full load conditions.
load condition,
load is about 40%
and correction factors range from about 3¾ to 4¼. Corrected temperature rises are then about 245 to 278 F for the
B-phase and 148 F to 167 F for the C-phase. Table 2 summarizes our estimates.
% Load B-Phase Initial C-Phase Initial B-Phase Corrected C-Phase Corrected
Rise in F Rise in F Rise in F Rise in F
70% (Normal) 65.5 39.4 115 69
40% (Two Feeder) 65.5 39.4 245-278 148-167
Table 2. Raw and full-load corrected temperature rises.
These corrected temperature rises give the thermographer excellent information about what to expect when the load
goes up. They can be used to help evaluate risk of going to higher loads.
So, even though we still don’t have the complete story on load correction, we strongly recommend thermographers
consider the range of values for temperature rise that could be realized as the load increases. The load correction
curves are for zero wind conditions. Trying to correct both for wind and load can be problematic as they are related.
We recommend not performing IR surveys in low load, windy conditions. You will miss some problems, and
correcting for those you do find can be difficult. If you must go out in those conditions, plan to return as soon as the
wind goes down and/or the load goes up.
6. SUMMARY
Wind speed and load are two important measurements in addition to your IR camera data when working on
electrical systems. Though written with substations in mind, the work is equally valid for other areas such as
distribution and transmission lines.
Measuring wind speed and load in a substation is a lot easier than correcting temperature rise data for them. This
paper shows the correction factors in both cases are large, so even though we do not have all the answers at this
point, the wise thermographer will do well to consider their significance. Further research and experimentation
needs to be done to help quantify these factors and provide thermographers with even better correction tools. Be
very wary of simple wind or load correction factors. They don’t exist at this time.

InfraMation 2002 ITC 035 A 2002-08-01

For a ring bus, or parallel circuits, problems found with IR should be backed up with load measurements of each
parallel leg. It can occur that a major problem is shunting so much current it appears cool whilst a smaller problem
in the other leg shows hot due to the high load it is carrying from the shunted current. Safe practice says measure
the load and be sure. The switch example illustrates this, though safety was not an issue in this regard as both legs
of the switch would open at the same time.

7. REFERENCES
1. Madding, Robert P. and Lyon Bernard R.; “Wind effects on electrical hot spots--some experimental IR
data”; pp 80-84; Proc. Thermosense XXII; SPIE Vol. 4020; April, 2000
2. Lyon, Bernard R. Jr.; Orlove Gary L. and Peters Donna L.; “The relationship between current load and
temperature for quasi-steady state and transient conditions”; pp. 62-70; Proc. Thermosense XXII; SPIE
Vol. 4020; April, 2000
3. Perch-Nielsen T., Sorensen, J.C.; “Guidelines to thermographic inspection of electrical installations”; pp 2-
13; Proc. Thermosense XVI; SPIE Vol. 2245; April, 1994
4. Madding, Robert P.; “High voltage switchyard thermography case study”; pp. 94-99; Proc. Thermosense
XX; SPIE Vol. 3361; April, 1998
5. Madding, Robert P.; “Finding Internal Electrical Resistance from External IR Thermography
Measurements on Oil-filled Circuit Breakers During Operation”; Proc. InfraMation; Vol. 2; October, 2001;
pp37-44

8. ACKNOWLEDGEMENTS
The authors wish to thank the Infrared Training Center at FLIR Systems and Carolina Power and Light for providing
the resources to make this work possible. We are also grateful for the support of CP&L operations and maintenance
personnel who provided additional data and information for this paper.

InfraMation 2002 ITC 035 A 2002-08-01