Sei sulla pagina 1di 8

Finding internal electrical resistance from external IR thermography

measurements on oil-filled circuit breakers during operation


Robert P. Madding, Director
Infrared Training Center
FLIR Systems, Inc. Billerica, Massachusetts
ABSTRACT
Many electrical devices indicate internal problems as a temperature increase on their outside surfaces. For many devices
such as oil-filled circuit breakers (OCBs), the temperature rise is small as there is significant thermal insulation between the
hot contacts and the surface (indirect measurement). Under steady state conditions, heat transfer between the external surface
and the environment can be calculated without needing to know the internal properties of the device. You only need to know
the external surface temperature, area and emissivity and orientation together with environmental parameters such as
background temperature and ambient air temperature. IR thermography is an excellent tool for finding these parameters.
The internal electrical resistance is calculated by calculating the watts required to achieve thermal balance of the external
surface with the environment through convection and radiation and measuring the electrical load on the device. This paper
shows how this is possible under steady-state conditions and gives results for example case studies using visual basic
software developed by the author. The software algorithm is currently limited to steady state, no-wind conditions for fairly
simple geometrical structures. The results have been excellent for OCBs. The software will become a new tool in the
thermographers’ toolbox, allowing reporting of internal electrical resistance for certain cases using IR thermography results.
Keywords: Infrared thermography, heat transfer, convection, radiation, oil circuit breakers, watt loss, internal electrical
resistance, thermal balance, condition based maintenance, predictive maintenance.
1. INTRODUCTION
Thermographers have been evaluating OCBs for years with IR cameras as part of their normal IR surveys. They have found
that small surface temperature rises on OCBs that are due to internal conditions usually indicate major problems. Often, it is
difficult to relay these findings to electricians unfamiliar with IR thermography, as they are not trained to think in terms of
temperature. But they are taught to think in terms of resistance. They know the micro-ohm limits of their OCBs. If the
thermographer had a tool that could quickly convert surface temperature to electrical resistance, it could be a real boon to
electrical system surveys. Making this connection is also important for the thermographer. This paper will show
quantitatively the connection between OCB surface temperature and internal electrical resistance.
Using a combination of infrared thermography together with additional simple measurements and knowledge of steady state
heat transfer concepts, we can determine the internal electrical resistance on oil circuit breakers (OCBs) under operating
conditions. This is a non-intrusive, non-destructive condition based maintenance evaluation that is only slightly more effort
than a normal infrared thermography survey of OCBs. The key is understanding the thermal balance achieved between an
operating OCB tank and its surroundings. The outer surface of the OCB will lose any internal heat generated by convection
and radiation to the surroundings. How the heat gets from the internal region of the OCB to the outer surface is a very
complex heat transfer problem. The oil will circulate, the conductors will conduct and the heat will flow. Analyzing this in
detail is a difficult, expensive task requiring detailed knowledge of the internal structure of the OCB, thermal conductances
and so on. Fortunately, we don’t need to do this. We just need to know how the heat is lost from the surface of the OCB.
Sound like magic? Not really. Consider two resistors in series with a voltage drop across them to ground. Current will flow
according to Ohm’s law (see figure 1),

I=
(V − V )
1 g
(1.1)
R
where I is the current through, and V is the voltage drop across the total resistance R=R1+R2. But in steady state conditions,
the current is also equal to just the voltage drop across R2 divided by R2 as shown in equation 2.
I=
(V2 − Vg )
(1.2)
R2

If we know the value of R2 and the voltage drop across it, we don’t need to know the value
of R1 or the voltage drop across R1. Heat flow behaves in a similar manner. The heat flow, V1
Q, is measured in watts or BTU/hr and is analogous to electrical current. Temperature
difference, ∆T, is analogous to voltage difference and thermal resistance is analogous to
electrical resistance. In steady state conditions all the heat that is generated inside an OCB
flows out of the OCB. We see this as a temperature rise on the surface. In our analogy, the R1
OCB surface corresponds to the point between R1 and R2, with R1 the thermal resistance
inside the OCB and R2 the thermal resistance between the outside surface of the OCB and
the surrounding air and environment. If we know the temperature drop across R2, and the
I
value of R2, we can calculate the heat flow without knowing anything about the internal V2
temperature or thermal resistance!

Heat is lost from the OCB surface primarily by convection and radiation. For simple objects
such as boxes, cylinders and slabs, the heat transfer equations are well known. So, without
knowing the internal mechanisms for the OCB, we can calculate the heat flow out of the
R2
OCB with just knowledge of external parameters. (Some heat can flow out of the OCB by
conduction through the electrical conductors. This calculation does not capture this heat
flow. As such, we may be somewhat underestimating problem severity. At this point we Vg
believe this to be a small effect, as we have not observed a temperature rise on the OCB
conductors.) Figure 1. Resistors
in series.
The heat flowing out of an OCB in steady state conditions equals the heat generated inside
the OCB due to electrical resistive heating (there could be other sources of heat inside the
OCB, but it is unusual for an OCB to be warm for reasons other than resistive heating). In equation form,

Q = I 2R (1.3)

where the heat flow, Q, in watts equals the current, I, squared times the internal electrical resistance, R. Equation 1.3 tells us
that if we know the load current, and calculate the heat flow, we can determine the internal electrical resistance. We want the
OCB under a good load, and under steady state conditions. It shouldn’t be heating up or cooling down. Requiring steady
state conditions is the slippery slope for this work. Load changes, ambient temperature, wind, humidity or solar loading
changes all result in transient mode, or non-steady state conditions.

The good news is we don’t need to calculate the heat flow to 5% accuracy. In most cases, 20% to 25% accuracy is enough to
determine whether there is a problem. So, if conditions remain reasonably constant for two or three hours, steady state is a
good assumption.
2. HEAT LOSS CALCULATION SOFTWARE
Calculating heat transfer by convection and radiation may be well known, but it is not simple. The Infrared Training Center
has developed software that can make life much easier for the thermographer wishing to perform this calculation. This
section discusses some of the details involved. If you are not mathematically inclined, skip this section and go to the results
section.

The three heat transfer modes, conduction, convection and radiation act in concert and are always attempting to make
temperature differences zero. Only if there is a source of heat will a temperature difference be maintained over a long period
of time. If we think of how the surface of a cylindrical object loses heat to the environment, we quickly realize that from the
sides and top, heat is primarily lost by convection and radiation as air has a very low thermal conductivity. Through the base
of the cylinder, the heat transfer by conduction could be a significant contributor depending on the OCB mounting scheme.
Heat can also conduct out the current carrying conductors. In this work, we have focused only on the surface losses through
convection and radiation as the thermograms we have viewed show that is where the highest temperature differences exist.
For OCBs, convective heat transfer through the oil causes internal thermal stratification. It is hotter at the top, almost
ambient at the bottom. For each surface, we must calculate the heat loss, Q, add them together to get the total heat loss, then
divide the result by the square of the current, I, to get the internal electrical resistance, R, responsible for this whole mess. In
equation form,

Qtotal = Qside + Qtop + Qbottom


Qconvection ( i ) = hi Ai (Tsurface ( i ) − Tair ( i ) )

(
Qradiation ( i ) = ε i Fiσ Ai Tsurface
4 4
( i ) − Tbackgnd ( i ) )
(1.4)
Qside = Qconvection ( side ) + Qradiation ( side )
Qtotal
R=
I2

The letter, i, indicates one of the surfaces, such as side; h is the convective coefficient, A is the surface area, ε the emissivity
of the surface; F, the view factor (assumed 1 for these calculations), σ the Stefan Boltzmann constant, and T, the absolute
temperature. Three temperatures for each surface are required: the surface temperature, the ambient air temperature and the
background temperature. Ambient air and background temperatures are often the same for all surfaces, but they could be
different. The top surface background, especially, can be much lower for a clear sky than the sides or the bottom.

The software is written to


account for three basic shapes:
cylinder, box or slab. The
slab just calculates the heat
flow through a surface. This
is not for OCBs but could be
useful for boiler heat loss
calculations. Normally, for
OCBs, the thermographer
would use the cylinder shape.
After selecting the shape, the
user enters the appropriate
dimensions in either English
or metric units. The program
calculates surface areas on
command. Figure 2 gives an
example form where the area
of a cylinder is calculated.

When the area calculation is


satisfactory, the user selects
the “Go to Data Entry” button.

Figure 3 is an example of a
Figure 2. Example of English units area calculator input form.
data entry form. The area
calculation is automatically
entered from the previous
form. The user enters the
surface temperature and
emissivity for each
surface together with the
appropriate air and
background temperature.
The software assumes the
top and bottom disks have
the same emissivity, but
can have independent
temperatures. In this
example, we picked the
bottom surface
temperature, the air
temperature and the
background temperature
to be the same. This will
result in zero heat loss
from the bottom. Finally,
the user enters the current
in amps, then clicks on
the calculate button.

Figure 4 shows the final


result for a real-world Figure 3. Heat loss calculation form just prior to calculation.
example described below.
The convective loss for
the side is about one-half
the radiative loss for that
surface. This illustrates
the significance of
radiative heat transfer for
high emissivity surfaces.

Of the 300 watts total


almost 80% is from the
side, largely due to the
greater surface area. A
current of 370 amps
yields a resistance of
2,189 micro-ohms,
definitely too high for a
properly functioning
OCB.

A skilled thermographer
can acquire the data
needed to determine
internal resistance in a
matter of minutes, with
little additional effort
from a routine IR survey. Figure 4. Calculation form after calculation showing final result.
Except for the physical
dimensions, the thermographer should be capturing all the other data as a normal part of the IR survey.

The only real difficulty is the requirement for steady state, no wind conditions. Solar loading, if the same on all OCB cans,
can be accounted for by doing the calculation twice, one for a good OCB and the other for the anomalously warm OCB. The
difference in the grand total watts dissipated between the two cans can then be divided by the square of the current to
estimate the internal resistance.
3. SAMPLE DATA
1. Laboratory experiment—Light bulb in a box.
The science behind the above calculation is sound. But to use the software in IR applications, we must first be sure the
results withstand the pressures of our assumptions in a real world environment. To do this, we used one of the Infrared
Training Center’s sophisticated Level III student workshops: The light bulb in a box as shown in figure 5. Without removing
the box, determine
93.6°F
the light bulb AR01

wattage and
90
electrical resistance
using an IR camera,
85
clamp-on ammeter
and tape measure. 80
With the heat loss SP01

calculation software, 75
the problem is
greatly simplified. 70
One uses the IR 70.0°F

camera to input the Figure 5. Light bulb in a box workshop setup. Photo on the left shows cardboard box on
box dimensions. Styrofoam base. Inside the box is a light bulb of unknown wattage. Thermogram on the right
The IR camera is shows box surface thermal patterns due to operating light bulb.
used to determine
the average surface
temperature of the sides and
the top. The heat loss from
the bottom is negligible as it
is well insulated compared to
the cardboard surfaces.
Ambient and background
temperatures are also
measured. If the emissivity
of the cardboard is unknown,
it is readily measured using
the reference emitter
technique by applying high
quality electrical tape
appropriately to the surface.

Figure 6 shows the results.


About 21 watts are lost from
the sides and another 3.5
from the top, totaling 24
watts. Actual value of the
light bulb is 25 watts. We
measured 0.2 amps current,
giving 606 ohms resistance.

Note the average temperature Figure 6. Light bulb results. Actual nominal value is 25 watts. Calculated
of the entire side of the box resistance is 606 ohms (Displayed value in micro-ohms).
was taken, and that the temperature is non-uniform on the box side.

Does this matter? If heat transfer were linear with temperature, taking the average temperature of a non-uniform surface
shouldn’t matter. But heat transfer by radiation varies as temperature to the fourth power. However, for small temperature
differences, ∆T less than about 15 C, heat transfer by radiation can be linearly approximated with temperature difference to
within 5%. One way to test this is to divide the area up into several smaller areas where the temperatures are reasonably
uniform, find the watt loss for each area, then add them to find the total. If the results are within 5%, we should be able to
use one area average over the entire surface. We will try this in the next example.

2. Field Data—Oil-filled circuit breaker

Figure 7 is an IR image provided by Southern California Edison of an OCB in trouble. Spot 1, SPO1, gives the background
temperature (19 C) for the side of the OCB, and SPO2 the background temperature (-9.8 C) for the top. Ambient temperature
was taken as the same as SPO1. A support structure interferes with viewing the entire 48-inch height. The area box covers
about 39 inches by 24 inches. The calculation depends on the
thermal pattern being similar around the circumference of the 35.5°C

OCB. Figure 4 shows the data used for this calculation and the SP02 35

results. We used the area average temperature (31 C) for the side,
estimated the top temperature (35 C) by creating a thin area near AR02 30

the top just below the support structure. As the temperature near
SPO1*
the bottom approaches ambient, we estimated its temperature as 25

ambient. Emissivity of the paint on these breakers is about 0.95


for long wave IR cameras. 20

Watts generated by the internal problem based on the above 15

assumptions and data total 280, resulting in an internal resistance 13.1°C


of over 2,000 micro-ohms. The supplier of the thermogram
provided a post mortem measurement of over 200 micro-ohms, Figure 7. Thermogram of Mossberg 12KV OCB.
well shy of the calculated value. The large difference is probably Area shown covers 39”X24”. Credit: Mike
due to the fact the OCB was not measured until after contacts Palusso, Southern California Edison.
were opened, OCB removed from service and contacts reclosed.
Contact resistance will be significantly decreased by this type of cycling in most cases. In addition, the temperature
coefficient of resistance will exacerbate the problem as the electrical resistance of copper increases with increasing
temperature. And the internal temperatures must be quite high to generate almost 300 watts on the surface. As we do not
have full thermographic coverage of this OCB, the assumption of uniform, stratified circumferential heating may pose a
problem. This assumption is not quite true as the thermal pattern in figure 7 is not uniformly stratified, but appears hotter in
one corner. But it is the best estimate we can do with the thermographic data available. We can find the lowest estimate by
slicing the data vertically on the right hand side of the OCB. Doing this gives an average area temperature of 29 C. Using
this value for the cylinder and the top, still results in 1600 micro-ohms resistance. Data in the next section confirm our
algorithm in all cases.

Finally, there is the effect of thermal uniformity for the average temperature over an area. We divided the side into thirds
horizontally and performed the watt calculation on each, then summed them. Results were: Bottom third, 30.4 watts; middle
third, 78.5 watts; top third and top of OCB, 177 watts. This calculation totals 286 watts, compared to 280 done initially. It is
more accurate to be sure, but well within the error percentages we need, and does not change the result significantly. Taking
the entire side in one average measurement is sufficient.

3. Comparison to published data

James Bodah (1) performed analyses on problematic systems. Table 1 gives physical parameters of each problem tank and
input temperature data from his referenced publication. These are fairly large OCBs under a moderate load and
environmental conditions. Bodah does not include background temperature in the data, so we are forced to assume
background is the same as the ambient air temperature. Also, the bottom temperatures reported were equal to the side and top
temperatures, somewhat unusual for the devices studied. The bottom portion of table 1 gives the watt loss and resistance
calculation results. These results agree within a few percent of measured values.
Oil Circuit Breaker Type
Parameter
230 KV 1200 A 115 KV 1200 A
Diameter (in) 84 48
Height (in) 138 94
Current (A) 375 490
Emissivity 0.96 0.96
Top temperature C 20.5 21.1
Side temperature C 20.5 21.1
Bottom temperature C 20.5 21.1
Air temperature C 16.3 12.0
Calculated Watt Loss 950 825
Calculated Internal Resistance in micro- 6,759 3,436
ohms using itc software
Measured Internal Resistance in micro- 7,000 3,260
ohms
Per Cent Difference 3.4% -5.4%

Table 1. Data and results for two oil circuit breakers.

4. PROPOSED NEW SEVERITY CRITERION FOR INDIRECT MEASUREMENTS

The best severity criterion for oil circuit breaker internal resistance problems is, of course, the internal resistance. Next best
is watt loss. Both require significant calculation. Without the aid of computer software, this becomes a real chore. The
software shown in this paper will eventually be made available through the Infrared Training Center, probably as a part of the
Utilities CD the students receive when they attend class. We want to analyze more data to insure quality results. Until such
time, there is an alternative.

Equations (1.4) for convection and radiation have two parameters in common: Temperature and area. For small temperature
differences, less than 15 C, the radiative transfer equation can be linearly approximated as discussed above. Doing this
creates a common factor, A∆T, the surface area times the temperature difference. For IR surveys of indirectly viewed
equipment where the ∆Ts can be smaller than 15 C, and still indicate a serious problem, the A∆T parameter could be used in
lieu of straight ∆T. Table 2 gives A∆T for the above examples, together with calculated watt loss and the ratio of watt loss to
A∆T. For A∆T to be valid, the ratio should be constant. For these cases, it is within 6% of the average value.

For these real-world examples, A∆T correlates with severity based on either watt loss or internal resistance much better than
straight ∆T. In fact, severity based on ∆T alone would be exactly the opposite. This is an important result. We definitely
want to pursue this line of investigation on more OCBs as well as other indirect targets.

Item Calculated A∆T Watt Diameter Height Approximate


Watt Loss Sq. meter X C Loss/A∆T ∆T in C
Mossberg 12 KV OCB 280 36 7.8 24” 48” 12
115 KV 1200 A OCB 825 104 7.9 48” 94” 9.1
230 KV 1200 A OCB 950 129 7.4 84” 138” 4.2

Table 2. Area times ∆T calculation for example problem OCB tanks.


5. SUMMARY AND CONCLUSIONS

Under steady state operating conditions, we can calculate the internal electrical resistance of targets such as oil filled circuit
breakers with knowledge of external parameters and load. This paper uses OCBs as examples, but in principle the
calculation applies to any similar target. Load tap changers, bushings, capacitor banks and lightning arrestors are other
targets that could prove fruitful.

For ease of software development and user input, we have kept the potential shapes simple, boxes, vertical cylinders and
slabs. Also, for simplicity, the software applies to no wind conditions. The practicing thermographer will gather most of the
data during a routine survey. More thermograms of problem targets giving thermal conditions of the entire surface plus
target dimensions are the tasks in addition to normal measurements of load, emissivity, ambient and background
temperatures.

The calculations give good results compared to the limited real world data available. The author invites more data to validate
the calculation and software further.

In lieu of the complete calculation, a new severity criterion is proposed for indirect targets, or targets where ∆T is less than 15
C: Area times temperature difference, A∆T. For the limited data available, A∆T appears to be a much more applicable
criterion than ∆T alone for these indirect targets. Modern IR cameras with average temperature within a selectable area
prove invaluable for facilitating all these calculations. For A∆T, find the average ∆T, multiply by the area and you’re done.

6. ACKNOWLEDGMENTS

The author gratefully acknowledges the support of FLIR Systems, Inc for providing the time and computer resources for this
work. I also am grateful to Mr. Mike Palusso of Southern California Edison for providing the example thermogram and
ancillary measurement data. Thanks also to Mr. Gary Orlove of the Infrared Training Center for his light bulb in a box
workshop, and to Mr. Ron Lucier for reviewing and commenting on this work.

7. REFERENCES

1. Bodah, J. M.; “Estimating Off-line High Voltage Apparatus Dielectric Loss and DC Contact Resistance with On-
Line Infrared Measurements”; Conference Record of the 1996 IEEE International Symposium on Electrical
Insulation; IEEE Publication 96CH3597-2; pp 458-461

Potrebbero piacerti anche