2008 5th International Multi-Conference on Systems, Signals and Devices
Thermal Modeling and Simulation of Distribution Transformers
A. Elmoudi
Electrical and Computer Engineering, Sultan Qaboos University, Oman
Abstract-This paper presents a thermal model to
predict the top oil and hot spot temperatures of a
transformer. The model is applied to 2500 kVA ONAN
(without external cooling) distribution transformer
equipped with thermo-couples and tested at varying load.
A non linear least square algorithm is used based on
measurements to determine the model parameters. The
model showed good results
Index Terms - Power transformers, Temperature,
Thermal factors
I. INTRODUCTION
The loadability and useful life of a transformer is
determined partially by the ability of the transformer to
dissipate the internally generated heat to its surrounding.
Therefore, the knowledge of the transformer thermal
performance could lead to improving the utilization of
transformers
There are several thermal JTIodels used to predict the top
oil and hot spot temperatures of a transformer. The
commonly used model is described in clause 7 in the
IEEE loading guide and IEC 354 [1][2]. The top oil rise
equation of clause 7 of the IEEE guide is modified to
allow for continuously varying ambient temperature [3].
An alternative method is suggested in Annex G. The
method requires the use of bottom oil rise over ambient
at rated conditions. The duct oil temperature is
introduced which may be higher than the top oil
temperature under certain conditions [1].
Another transformer themlal modeling approach in the
form of an equivalent circuit based on fundamentals of
heat transfer theory is proposed. The top oil temperature
is calculated from the oil-to-air model. The top oil
temperature becomes the ambient temperature for
winding-to-oil model. The top oil temperature model is
shown to be valid for a 250 MVA transformer in the
field. [4][5].
The hot spot equation which is analogous to the top oil
is introduced and the thermal model is shown to be valid
for 250 MVA transformer equipped with fiber optic
probes and heat run tested according to a load cycle
comprising a short time load up to 2.1 pu [6][7].
This paper, presents the thermal model equations. The
model is validated by ITleaSUrements carried out on
2500 kVA distribution transformer equipped with
978-1-4244-2206-7/08/$25.00 ©2008 IEEE
thermocouples. The model parameters are determined
using a non-linear least square algorithm. The model
showed good results
II. THERMAL MODEL EQUATIONS
The thermal model is established to determine the hot
spot temperature. The top oil temperature is calculated
from the top oil model. The top oil temperature becomes
the ambient temperature for the hot spot model.
A. Top-oil Model
The top oil temperature is often labeled as liquid
temperature and it is traditionally measured at the top of
transformer oil. The differential equation used to
calculate the top oil temperature is [4]:
2
1
IpufJ+ [fj,() ]1/n dOo [() e ]1/n
p-;l' O-R =ro"d{+ 0- A (1)
The variables and parameters are:
Input variables (Functions of time, t)
I pli is the load current per unit.
()A is the ambient temperature, 0 C.
Output variable (Function of time, t)
()o is the top oil temperature, 0 C.
Parameters (constants)
P is the ratio of no load to load losses,
conventionally R.
A()O-R is the top oil temperature rise over ambient, K.
To is the top oil time constant, min.
n is an exponent defines non-linearity.
Equation (1) considers only the top oil temperature and
the model to calculate the hot-spot temperature is
presented in the next subsection.
B. Hot-spot Model
The hot spot temperature is the maximum temperature
occurring in the winding insulation system. This
temperature represents the thermal limitation of the
transformer loading
 2008 5th International Multi-Conference on Systems, Signals and Devices

The differential equation which is analogues to the top fifth layer and the other after the tenth layer. The
oil temperature (1) based on the winding-to-oil transformer is equipped with 28 thermocouples. The
equivalent circuit is [6]: transformer is heat run tested under load cycle of 1 pu
50Hz/ 300 minutes and 2 pu 50Hz/90 minutes. The
thermocouples installed in different locations to measure
characteristic temperatures such as top oil, hot spot,
bottom oil and duct oil. In Fig.3 the transformer cross-
section showing the position of some of the
thermocouples installed to measure the top oil and the
K -- OK +OH (3) winding hot spot temperature in the transformer. The
() -- ()K +75
thermocouples installed to measure the top oil at
different locations oil pocket, tank surface and under
The variables and parameters are:
tank cover. Others installed in the winding hot spot
Input variables (Functions of time, t) temperature location based on the manufacturer
experience. The measured data for 1 pu load is shown
I ptl is the load current per unit.
in Figs. 4-5. The measured top oil in tank, pocket and
()o is the top oil temperature, 0 c.
tank surface and the measured ambient temperature is
Output variable (Function of time, t) shown respectively. A time delay of up to 12 min is
observed. Fig.6-7 shows the loading current cycle and
()H is the hot spot temperature, 0 c. the measured top oil pocket for the loading cycle. The
Parameters (constants) data was recorded at two minutes intervals.
PEC-R is the rated pu eddy current losses at the hot
spot location.
A ()H-R is the hot spot temperature rise over top oil, K.
TH is the hot spot time constant, min.
m is an exponent detines non-linearity.
Ko is the resistance correction due to temperature
change

III. SIMULATION MODEL

Equations (1) and (2) are modeled in Simulink/Matlab. Fig. 2 Block diagram of the hot spot model
At each discrete time the top oil temperature is
calculated and it becomes the ambient temperature to
calculate the hot spot temperature as shown in Fig .1. V. PARAMETER ESTIMATION
The equations are solved numerically using Runge-
Some of the thermal model parameters are known
Kutta method.
accurately from the factory tests. The measurements can
be used to determine the unknown thermal model
parameters.
Top-oil Hot-spot I
model model
m:n ~L(F(X)i-Yi) (3)
i=l

Fig. 1 Thermal model for hot-spot calculation

where
F(x); is the model calculated temperature at each
Figure 2 shows the block diagram of the hot spot model
step.
equation.
X are the parameters to be determined.
Yi is the measured temperature at each step.
IV. MEASURED DATA I is the index over the period.
The tested transformer is 2500 kVA 20.5±2x2.5%/0.71
Using the non linear least squares [8], the model
kV, ONAN, without external cooling Dyl1. The LV
equation (1), the parameter values are estimated from
winding consist of 18 Aluminum foil layers with an
the measured data and the corresponding model
axial cooling duct after the 9th layer. The HV winding
parameters are found to be:
consist of 15 layers and has 2 axial ducts one after the
 2008 5th International Multi-Conference on Systems, Signals and Devices

l1()o-R = 50°C, To=130 min, n =1 30 1--- ~r----------T----------'------~ --.-------l

29~ ~
The parameter n was found to be 1. This reveals that the
heat transfer from the oil in the tank to the outside air is 28l I

- I
only proportional to the temperature difference between U 27 1

the liquid and the air. This can be due to the fact that the ~ 26l
transformer is with out external cooling. 1
E 25 r
'1

The model parameters for the hot spot model equation

~ 24
1

(2) are found to be

~ 23l
« I

m=O.8 22[ 1
21 ~
20 - - .1..

1
------l.-

2
--'----

3
.L

4
J5
Time (hrs)

Fig.5 Measured ambient temperature

IIV

TIme (hrs)

Fig. 6 Applied current

~
:O~::~duct
insulation

:.~~
•••••• ••••••••• I

Fig.3 Transfonner under test: A thennocouple to measure top oil.

A thennocouple to measure winding insulation hot spot •

TIme (hrs)

Fig. 7 Measured ambient temperature

v. RESULTS

The top oil and hot spot temperatures for any loading
Fig. 4 Measured top oil temperature at different location can be predicted by inputting the measured pu load,
ambient temperature and the optimized parameters into
 2008 5th International Multi-Conference on Systems, Signals and Devices

equation (1) and (2). The model output and the ACKNOWLEDGEMENT
measured and predicted top oil and hot spot
The author wish to acknowledge Dr. Hasse Nordman
temperatures are as shown in Fig. 8 and 9. It is clear that
from ABB Transformers Vaasa for the measurements.
the prediction does represent fairly well the top oil and
hot spot temperatures.
REFERENCES

140Ir-~as~l [1] IEC Guide for loading for Oil Immersed Power
_ Predi~ Transformers, International Elctrotechnical
120 Commission Standard, Geneve, Suisse, 1991.
0' 100 [2] IEEE Guide for loading Mineral Oil Immersed
CL..
Transformers, IEEE Standard C57.19, 1995.
[3] B.C. Lesieutre W.H. Hagman J.L. Kirtley Jr. , "An
Improved Transformer Top Oil Temperature Model
for Use in An On-Line Monitoring and Diagnostic
System," IEEE Trans. on Power Delivery, vol. 12,
40 no. 1 Jan 1997, pp.249-256.
[4] G. Swift, T. S. Moliniski, and W. Lehn, "A
Fundamental Approach to Transformer Thermal
fime(hrs)
Modeling-Part I: Theory and Equivalent Circuit,"
IEEE Trans. Power Delivery, vol. 16, No.2 pp. 171-
Fig. 8 Predicted and measured top oil temperature
175, Apr. 2001.
[5] G. Swift, T. S. Moliniski, R. Bray, and R. Menzies,

:1:6:0[C ~:~:j -,-- ----,------~~.- .... ---1 "A Fundamental Approach to Transformer Thermal
Modeling-Part II: Field Verification," IEEE Trans.
Power Delivery, vol. 16, No.2 pp. 171-175, Apr.
j 2001.
[6] A. Elmoudi, M. Lehtonen, Hasse Nordman
E 140
"Transformer Thermal Model Based on Thermal-
~
~ 120 Electrical Equivalent Circuit," proceeding of CMD
~
~
&
100 2006 South Korea, 2-4 April 2006.
t::c 80 [7] Elmoudi, Asaad, Lehtonen, Matti and Nordman,
Hasse, " Thermal Model for Transformers dynamic
60
loading", IEEE International Symposium on
Electrical Insulation Toronto-Canada 11-14 June
LU , -----L -'-- .L._~----L

3 4
._----'----
5
.. -1_
6
2006.
rime (hrs)
[8] Simulink parameter estimation, non linear least
square.
Fig. 9 Predicted and measured hot spot temperature in
the HV winding

VII. CONCLUSION
In this paper, a thermal model to predict the top oil and
hot spot temperatures of a transformer is applied to a
2500 kVA, ONAN distribution transformer. The
transformer has been heat run tested according to a load
cycle of 1 pu and 2 pu. The model parameters are
determined using a non-linear least square algorithm
and measured data. In general the thermal model has
shown good results compared with measurements.