Thermal Performance of the Radiator

for Electrical Transformers

Vasanth Vailoor, Tranter Radiator Products, Inc.

Introduction
Electrical transformers are a widely used piece of equipment in the utility
industry. There is a need to cool the transformers due to rapid electrical performance
degradation especially at elevated temperatures in large transformers. So the
transformers are cooled in an oil bath, which in turn is cooled using various
equipment and mechanisms that go through different names radiators, tube coolers,
flexoplates, fins, plate coolers, etc. Thus the transformer cooler is a critical link in the
overall optimum usage of the transformer. The performance of this transformer
radiator is a term used to denote the heat dissipation characteristics of the cooling
mechanism.

Background
Radiator performance is usually associated to heat transfer area. This is an easy
association the more the heat transfer area, obviously greater is the heat dissipated.
However, the basic heat transfer equation that governs the heat dissipated from the
transformer radiator is given as
Q = U*A* T -

(1)

Where Q is the heat dissipated,

U is the overall heat transfer coefficient
A is the heat transfer area and
T is the effective temperature difference between the oil and the air.
So it is easy to see the dependence of Q on A.
Different manufactures make radiators using different designs that affect the
performance of the radiator. These differences may be apparent or not so apparent
and may or may not be of consequence. From the equation of heat transfer, the heat
transfer area is an apparent and easily quantified parameter and so a simple

correlation is drawn between Q and A assuming that U and

T are constant for

radiators that fall into the same design type. This might be a relatively justifiable
assumption but when one starts drawing the same type of correlation between
products from different manufactures and sometimes even different designs, the
relationship is too simplistic. However, U and T are not easily measured and these
are the non-apparent factors that can significantly influence the heat dissipation rate.
We have been testing radiators of various designs from several manufacturers and
have been documenting the variations in U and T that influence the heat dissipation.
Presently our testing has been limited to natural convection heat transfer mode of
operation with heat transfer areas ranging from 5.5 to 550sqft. We have found U
ranging from 0.86 in large radiators to 2.66W.ft-2.C-1 in small radiators at Tmax of
65C. In addition, the heat dissipation is governed by the effective temperature
difference between the oil and the air or the log-mean-temperature difference that
ranges from 23C to 41C as shown in equation-1 and not the Tmax, which is held
constant at 65C. This shows the range of variation of U and

T that exist in the

market. However, quite often, due to the constrained resources, available and
allocated for radiator testing, manufacturers test a limited number of radiators and
assign the thermal performance of their entire range of products based on the heat
transfer area that can be measured quickly and easily, assuming that the other factors
are constant.
Thus, wide ranges exist in thermal performance due to design variations. Design
variations usually evolve out of leaner manufacturing initiatives when thermal
performances may be overlooked or even compromised. This adds up and compounds
with time and in the end, this information is drowned in safety factors, rules of the
thumb and factors of experience. Finally, when these safety factors occasionally or
consistently fail, new safety factors are applied uniformly to an entire range of
products and the financial impact would be passed on to the uninformed customer at
the next opportune moment.

Test Set-up
A schematic diagram of the test set up is shown below:

Radiator

Therm owell

Valves

Heating
Cham ber

Figure-1 The Schematic diagram

It consists of the following:
1) Tank or heating chamber for the oil
The heating chamber for the oil is a compact 20x20 square cross-section area
with a height of about 10 to house 7 oil heaters. The square section transitions at the
top and the bottom into 5 diameter pipes. The heating chamber and the transitions
are made of plate steel welded together into three separate pieces and then bolted
together.
2) Heaters for oil
Seven electric oil heaters, are mounted, three from one side and four from the
other side of the heating chamber. The heaters are rated at 6kW each.

3) Transformer oil
Transformer oil that meets the following specs is used.
a) ANSI/ASTM D 3487
b) NEMA TR-P8-1975
c) US Government Mil Spec VV-I-530A, supersedes Navy spec OS-1023
d) British Standards BS 148:1972.
4) Watt-Hour meters
GE model WH-4V1A3WD
5) Variac
Variable transformers are used to control input voltage to the heaters.
6) Data Acquisition System
LabView software along with hardware has been obtained from National
Instruments. It includes thermocouples (TCs) with 64-channels for temperature
monitoring. Some critical points where the temperature is monitored include:
a) Thermal-wells at the oil inlet to the radiator.
b) Surface mounted TC at oil inlet to the radiator.
c) Thermal-wells at the oil outlet from the radiator.
d) Surface mounted TC at oil outlet from the radiator.
e) Ambient air.
f) Hot air exit from the radiator.

Test Procedure
1) Set a variac voltage for necessary heat to reach a desired top oil temperature
and note electrical power input into the assembly.
2) Calculate the time constant of the system.
3) Monitor steady state (SS) readings after allowing sufficient time to reach SS.
4) Analyze and verify the data recorded.
5) Tabulate and present results.

Time Constant of assembly

1) Heat assembly to desired top oil temperature.
2) Switch off heaters and allow the assembly to cool.
3) Record the following:
a) Top oil temperature (at inlet to radiator or exit from heating tank
assembly), To,i.
b) Bottom oil temperature (at exit from radiator or inlet to heating tank
assembly), To,o.
c) Inlet air temperature to the radiator or the ambient temperature, Ta,i.
d) Exit air temperature from radiator, Ta,o.
4) Calculate the log-mean-temperature-difference, TLM from
TLM =

(To,i Ta,o ) (To,o Ta,i )

ln((To,i Ta,i )/(To,o Ta,i ))

(2)

5) With readings taken at two different times t1 and t2, the time constant,

is

calculated as follows:

= (t 2 t 1 ) * ln(

TLM2
)
TLM1

(3)

Steady State Readings

1) Set a desired electric heater power input, Pe in heater by adjusting the variac
and verify that all parameters are at SS after a period of five times the time constant
of the entire system.
2) Since the ambient temperature is not controlled, the test must be done when
ambient temperature does not change appreciably (preferably less than 1C during
this period of time).
LabView software is used to note the temperature values and is noted two times
per seconds for a period of 15 minutes to ensure steady state operation. Five readings
are taken at about 10C intervals for a temperature differential between top oil
temperature and ambient air temperature range of 30C to 70C.

Modeling and Analysis

1) Calculate the total electrical power dissipated by the assembly (Pe).
2) Calculate heat dissipated by Radiator QR, modeled as

Q R = Pe *

AR
AT

(4)

where AR is the area of the radiator and

AT is the total area of the system
3) The heat transfer coefficient, U, is calculated from equation 1. The U is not a
strong function of temperature and may be assumed constant for each radiator.
4) This data is tabulated and presented for use by the sales force and for easy use
by the customers. Under field conditions, the exit air temperature is a very
difficult parameter to measure and so U is based not on the log-meantemperature-difference as is dictated by equation-1 but the mean oil
temperature. When such is the case, due corrections must be applied to predict
the heat dissipation characteristics of the radiator for a maximum top oil
temperature of 65C.
Another factor of concern is the differential oil temperature. This can be predicted
from the resistance offered by the radiator to oil flow. Since the oil flows through the
transformer as well as the radiator, the sum or total flow resistance is of concern. So a
radiator must be matched to a transformer appropriately to accurately predict the heat
dissipation capability. In the absence of this essential data, oversized radiators will be
smothered by excessive safety factors and undersized radiators will be corrected by
mounting air-fans with loss of reliability when natural convection radiators have no
such loss of reliability.

Verification
Radiators that are used to dissipate heat from transformers have been tested in our
heat transfer laboratory to find out their thermal performance characteristics. The
parameter of interest is the product of the heat transfer coefficient and area or U*A.

This is easily modeled from the simple energy conservation equation. At steady state
condition,

Q ss = U * A * TLM

(5)

(5)

Q ss
or U * A =
TLM

Steady state condition is attained after allowing the entire system to remain at a
constant heater setting for several times the thermal time constant of the system.
The time constant of the system was determined by raising the temperature of the
system to a little over its highest operating condition and then the heater was shut off
while the temperature readings were noted. The exponential decay of the log mean
temperature difference ( TLM) was plotted and curve fitted to a function of the form
T2, LM
T1, LM

= e *(t 2 - t1 )

(6)

where is the time constant and t is the time. In addition, since

=

m * Cp
U*A

or U * A =

(7)

m * Cp

(7)

and m*Cp, the thermal mass can be easily determined from a transient ramp test,
again using the conservation principle in the form

Q trans = (m * Cp) * (

TSystem
t

) -

(8)

(8)

(9)

Q trans
or m * Cp =
TSystem
t

Combining equations 7 and 8, we have

Q
U * A = ( trans )/ TSystem

t
Thus we now have two independent methods to verify the U*A of the radiator.

For the radiator models tested, the heat transfer area A is easily defined by
physical geometry and does not change with the operating conditions. However, the
heat transfer coefficient U and the differential temperature

T do change with

operating conditions. We are primarily concerned with U when the maximum oil
temperature is 65C above ambient air temperature.
One more level of complexity must be added to fully evaluate the heat dissipation
characteristic of the radiator. That is the temperature differential or the log-meantemperature-difference, TLM. The TLM depends on the overall resistance to the oil
flow in the radiator as well as the transformer or the effective hydraulic diameter of
the entire flow path.

Results and Discussions

A summary of results of some of the radiators tested is tabulated below.
Sl No.

Description

Qrad at Tmax
of 65C

Area
2
[ft ]

U
-2
-1
[W*ft *C ]

TLM
[C]

15" x 1300mm x 15-plate, Painted

7507

160

1.613

29.1

15" x 78" x 15-plate, Painted

8388

244

1.233

27.9

520mm x 1300mm x 15-plate, Painted

7080

218

1.262

25.7

520mm x 1300mm x 15-plate, Painted

7235

218

1.272

26.1

15" x 126" x 15-plate, Painted

12883

394

1.157

28.3

520mm x 1300mm x 15-plate, HDG

7002

218

1.207

26.6

520mm x 1980mm x 25-plate, HDG

14588

554

0.864

30.5

15" x 1300mm x 15-plate, Painted

5745

160

1.397

25.7

8" x 24" x 4-plate, Painted

992

10.7

2.475

37.6

10

8" x 24" x 2-plate, Painted

576

5.33

2.661

40.6

Summarizing from the information above, the various factors that affect the
thermal performance of the radiator are as listed below:
1) Width of plate
Affects heat transfer area for the same center-to-center distance between
headers but the plan footprint changes and must remain acceptable.
Changes intake of ambient air but follows the law of diminishing returns.

2) Length/Height of radiator or center-to-center distance between headers

Performance improves with increasing Length/Height, again governed by the
law of diminishing returns.
Changes the time of contact with the ambient air and the contact area.
3) Number of Plates
Performance improves with increasing number of plates, again governed by
the law of diminishing returns.
Intake of ambient air increases with number of plates.
The heat transfer area increases linearly with number of plates.
4) Header Size
Performance improves with increasing Header Size.
Reduces pressure drop for fluid flow.
For the same top oil temperature, a larger pressure drop is available across the
radiator ensuring higher oil flow velocity and heat transfer coefficient.
5) Oil flow Capacity
Performance improves with increasing Oil flow Capacity.
Increases LMTD due to reduced Temperature drop in Radiator fluid.
A good indicator is the oil holding capacity of the radiator plates. A larger
capacity means a larger hydraulic diameter and so reduced friction for flow. This
reduces the drop in oil temperature ensuring a higher TLM.
6) Empty weight
Performance deteriorates with increasing empty weight.
Increasing the metal thickness reduces the overall heat transfer coefficient.
However, the thermal resistance offered by the metal is at least one order of
magnitude lower than the thermal resistance offered by the oil and the air and its
effect is negligible.

7) Heat transfer area

Performance improves with increasing Heat transfer area.
Enhanced surface due to ripples increases heat transfer area. However, ripples
could reduce the flow rate due to tortuous oil flow path and increase pressure drop.
8) Distance between plates
Performance improves with increasing distance between plates.
It reduces the interference between the boundary layers for air.
This outlines the various factors affecting the heat dissipation from the radiator.
Note - Though the heat transfer coefficient U is tabulated in the table above, quite often various
modified or effective heat transfer coefficient is used, especially in sales brochures. The modified U
may be used conveniently and accurately within the scope of assumptions made and due corrections
need to be factored in for good thermal performance prediction.

Conclusion
Thus the radiator manufacturer must be able to suggest the best or optimal
geometry for a given or specified design heat dissipation and other geometric and
process constraints. It must also be understood that the heat dissipation performance
of the radiator-transformer system can be accurately predicted only as a single whole
system. This is a typical situation where system optimization cannot be achieved by
optimization of the sub-systems.
It is our desire and intent here to make the industry aware of the various factors,
other than the heat transfer area alone that influence the heat dissipation characteristic
Q of the radiator.

10