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

net/publication/334574164

Theoretical and Empirical-Based Thermal Modelling of Power Transformers

Conference Paper · September 2018

DOI: 10.1109/ICHVE.2018.8642180

CITATIONS READS
13 433

4 authors, including:

Ahmed Gamil Ali Al-Abadi

Friedrich-Alexander-University of Erlangen-Nürnberg HITACHI ENERGY
23 PUBLICATIONS 53 CITATIONS 40 PUBLICATIONS 100 CITATIONS

SEE PROFILE SEE PROFILE

E. Schlücker
Friedrich-Alexander-University of Erlangen-Nürnberg
204 PUBLICATIONS 1,965 CITATIONS

SEE PROFILE

All content following this page was uploaded by Ali Al-Abadi on 19 July 2019.

The user has requested enhancement of the downloaded file.

 Theoretical and Empirical-Based Thermal Modelling
of Power Transformers
A. Gamil1*, A. Al-Abadi1*, F. Schatzl1, E. Schlücker2
1
SGB Power Transformers (SGB-SMIT Group), Regensburg, Germany
2
Institute of Process Machinery and Systems Engineering (iPAT), Friedrich-Alexander-University, Erlangen, Germany
( ahmed.gamil@sgb-smit.group , ali.al-abadi@sgb-smit.group )

Abstract: In the current study, we present an accurate double e-functions introduced in IEC standards, the following
approach for matching the calculated thermal performance of information can be gained in addition:
power transformers with measurements. It is done by adapting
some well-known and used assumptions for building a theoretical • Winding temperature gradient to oil [K]
thermal model based on special measurements for the objective of • Average winding-oil temperature rise [K]
minimum deviation (±3 K). This can guarantee to meet specified • Thermal winding time constant [min]
temperature rises while providing an optimized high economical • Oil time constant [min]
cooling system. The theoretical model calculates oil temperature The winding time constant should be considered to
distribution in medium power transformers. The model handles
different winding designs and arrangements and solves for determine at which minute the pure resistive (thermal effect) of
multilayer top-oil temperatures in a matrix form. It also the winding takes over during cooling. The average winding-
calculates the temperature distribution in radiators, which can oil temperature rise corresponds to the oil flowing through the
easily be validated during temperature rise test. First, the winding winding, which is directly influenced by the winding cooling
temperature is calculated, then the measured data from cooling- and presented in the cooling curve per winding. The overall
curves of a set of different transformer designs is inserted in the
model. This implies correcting the calculated gradient and average oil temperature rise of the transformer is a function of
recalculating the top-oil temperatures. The model however is the average oil temperature rise per winding (per cooling curve)
under development and in validation with a set of new and the corresponding oil flow rate inside the winding as well
manufactured transformers in parallel. as in the main gaps (residual oil). That means, using the
information available from the cooling curves provides direct
I. INTRODUCTION access to the winding thermal status, which is a great support
Calculation of the thermal characteristics of a power for adapting the thermal model.
transformer with minimum deviation (±3 K) from The advantage of detecting the temperature gradient to
measurements is crucial and one of the key parameters winding-oil volume for each winding is that the assumed
between transformer manufacturers. Previous researches relied linearity (based on standards) between Top, Bottom and
mainly on the simplified winding’s temperature distribution average oil temperature rise per winding and for the whole
approach presented in the standards (IEC, ANSI, etc.) to transformer is checked. The analyzed results show, that this
improve thermal models without considering that this relationship depends on winding geometry, oil flow and
simplicity would result in a mean oil temperature rise, which is winding losses distribution.
not presenting at all the real status of the oil flow inside the In the current study, we develop a theoretic thermal model
windings. that can calculate the oil temperature distribution in power
Accuracy must be put on both theoretical modelling and transformers. The model handles a set of contributing windings
measurements. Using the information from the measurement and solves for multilayer top-oil temperatures in a matrix form.
to feed up the theoretical model helps in improving the model Different winding designs, and corresponding cooling layouts
towards minimum deviation against measurements. This can can be modelled as well. In addition, the main-gap, by-pass
guarantee to meet specified temperature rises while providing flow and core losses are included in the model. In addition, the
an optimized high economical cooling system. Using fiber- radiator temperature distribution is calculated, which can easily
optics directly at the winding’s inlet and the outlet of oil does be validated during temperature rise test.
not insure the linearity of temperature distribution along the The theoretical model is based on thermal hydraulic
winding. Besides, placing fiber-optic sensors at the mid-height principles, [1, 2]. The multilayer model is developed to satisfy
of winding is not always a simple work. One source of the physical conditions with its corresponding constraints.
information, which used to be handled superficially, is the Analogy between thermal and electrical circuits are used.
winding cooling curve. It is used to give an indication of the Losses and covering factor per conductor are used for the
winding average resistance, which is converted to average detailed temperature calculation. With the covering factor it is
winding temperature rise. Using the mathematical formula with possible to distinguish between different winding designs. It
defines the local exposure to oil and the cooling layout.
 First, the winding temperature is calculated, and then the
data from cooling-curves of a set of different transformer Additional assumptions in the model are:
designs is inserted in the model. This implies correcting the
calculated gradient and recalculating the top-oil temperatures. ▪ Mixed top and bottom oil temperatures are equal to top
The model however is under development and in validation and bottom radiator temperatures, respectively.
with set of new produced transformers in parallel. ▪ For multi winding model, a constant bottom temperature
is considered which is more reliable than the average oil
II. THEORETICAL MODELLING temperature, since the average oil passes through
complicated geometries (winding configuration),
The thermal hydraulic model is based on applying the heat accordingly, it is subject to temperature variation.
and mass conservation principles with pressure balancing in a The equivalent circuit of the thermal hydraulic model for
closed loop as shown in Figure (1). It iterates for the unknowns two windings (LV and HV) and the radiator is shown in
(axial temperature distribution and mass flow rate) until the Figure (2).
stopping criteria is meet as the driving pressure must exceed
the total pressure drop to calculate the temperature lines of oil
and windings. The basic balancing model is as follows, [2]:

▪ Thermal driving force = produced pressure – head losses

(pressure loss)
▪ Power loss in winding = power gained by oil
▪ Heat in oil = heat exchange in radiator + heat exchange in
tank walls

Figure 2. Equivalent pressure, resistance and flow rate circuit

of the thermal model for two windings (LV, HV) with radiators

The thermal model iterates for the temperatures and the

corresponding flow rates of two (or more) windings and
radiators in a matrix as follows, Eq. (1)
θtop, LV θtop, HV θtop, Rad

θave, LV θave, HV θave, Rad (1)

θbot, LV θbot, HV θbot, Rad

The iteration continues until fulfilling the stopping criteria
and results in axial temperature distribution along the windings
and the radiator. Figure (3), shows the results of applying the
Figure 1. Close loop of oil flow in typical power transformer model on a case study transformer of 40 MVA (ONAN)/ 50
showing the temperature distribution within transformer MVA (ONAF). Results show that the top oil temperatures of
components the transformer are different, while the bottom oil temperature
is the same.
The winding’s pressure drop is calculated by applying
Simpson’s integration for the bottom, top and interval points
along the winding height. The total pressure drop in the tank is
calculated by summing up the pressure drops in windings with
exit and entrance of the collecting pipes. The same approach is
used to calculate the pressure drop in radiators. The
characteristic length used to calculate the pressure drop in the
windings is the average thickness of cooling ducts. In the
radiators the characteristic length is the thickness of the fins,
and in the connection pipes the pipe diameter. The assumption
here is constant flow rate through all radiators and radiator
plates as well. This is done by taking a loop over the radiator Figure 3. Axial temperature distribution of 40 MVA two
and equally dividing the flow for each plate. windings transformer.
 III. MEASUREMENT
The cooling curves of transformer windings are recorded
Top-oil HV
during the temperature rise tests. They provide an indication
Top-oil LV
for the increase in average winding resistance, which is
converted to average winding temperature. By extrapolation of Top-oil main-gap

the temperature curve backwards to time 0, the average Bot-oil HV

winding temperature rise is determined at shut down. IEC
Bot-oil LV
60076-7 [3] proposes different mathematical methods to
analyze the cooling curve as accurate as possible. The one with Bot-oil main-gap

most detailed information about the cooling process is as Time [hr]

follows, (Eq. 2), [3]: Figure 5. Oil temperatures at top and bottom of LV and HV
winding and main gap.
(2) Applying the cooling curves analysis on the same
transformer for each individual winding and its corresponding
Where, oil as shown in Figure (6), results in an axial temperature
θw: Average winding temperature rise [K] distribution for each winding as shown in Figure (7) (a, b).
w: Thermal time constant of winding [min]
o: Oil time constant [min]

The parameters a and b represent the winding temperature

gradient and the average oil temperature rise, respectively. The
physical interpretation of this formula is that the winding and
the oil in a certain small volume around the winding cool down
instantaneously with different time constants. Depending on
the winding time constant, the average winding temperature
rise meets the average oil temperature rise and starts to cool
down further until approaching ambient temperature. A
demonstration of a typical cooling curve analysis on a real
transformer measurement is shown in Figure (4). The analysis
of the cooling curves for the HV and LV winding shows Figure 6. Application of cooling curves analysis on the case
study transformer for each winding
different average winding temperatures and gradients. The
gradients are based on two different average oil temperature
rises for each winding individually.

Cooling HV Cooling LV
Curve fitting of average winding
Average oil (a)
Measured average winding

Time [min] Time [min]

Figure 4. Application of cooling curve analysis on two (b)

windings and their corresponding oil individually

For the case study, the oil temperature was measured with
fiber-optic sensors at the top and bottom of LV and HV
Figure 7. Cooling curves analysis results of axial oil
winding and top and bottom of the main gap between the
windings. The results are shown in Figure (5). temperature distributions of the two windings (a)LV, (b)HV
for ONAN and ONAF ratings.
 Results show that the axial temperature distribution of the oil
is linear as suggested in the standards. Comparing the cooling
curve analysis and fiber-optics measurement shown in Figure
(7) with the measurement of the thermometer pockets
presented in table I, show that there are high deviations in all
top and bottom oil temperatures. The mixed top-oil
temperature of HV and LV windings measured with fiber-
optics is around 10 K higher than the top-oil of the
thermometer-pocket. If the mix considers the residual oil, the
deviation will stay at 6 K to the top oil of the pocket, while the
temperature mix should include efficient cooling oil volume. It
seems from table 1, that the sensors on tank cover pockets
measure the temperature of residual oil through the main gap
between the winding. Figure 8. Oil circulation loop through LV, HV and radiator.
The more interesting is the deviation between oil
temperatures of the bottom pocket of the radiators against these IV. CONCLUSIONS
measured at the bottom of the windings using fiber optics. It is
surprising that the fiber optics at the bottom of the winding The current presented analysis allows to get more detailed
show higher temperature than the bottom of the radiators, and realistic results from the conventional heat run tests.
which means that the oil temperature increases on the way Using the input values of the cooling curves for the
between the outlet pipe of the radiators to the bottom of the developed thermal model is an accurate and practical tool to
windings. Especially in ONAF operation, the bottom study and optimize transformers based on real values from the
temperature is increasing by 10 K. winding itself. On the other hand, the linearity shown between
fiber optics values and cooling curves distributed over the
height of the winding confirms that the cooling curves
TABLE I delivered the real average oil temperature per winding.
MEASURED TEMPERATURES AT THE END OF HEAT-RUN The positioning of temperature pockets on the tank cover
Bot-Oil [K] Top-Oil [K] Main-gap Oil Top [K] and the bottom of the radiators with respect to the position and
Rating
(Pocket) (Pocket) (Fiber) the geometry of the winding can have a big influence on the
40 MVA 28 46,4 45,8 measured values and therefore increase the deviation between
50 MVA 15,6 46,5 43,3
measured and real temperatures.
The linear temperature distribution of the oil in the windings
is still under investigation. There are some indications gained
The deviations of the bottom temperature based on Figure (7) from this case study and other cases, where this linearity exists
show that there is an effective residual oil flow between the under some conditions of winding geometry, oil flow and
active part and the tank walls, which is not affected by the cooling condition. This will be discussed in future works.
cooling through the radiators. Once the oil flow back from the
radiators meets this residual oil with higher temperature, the REFERENCES
bottom oil temperature of the windings increases. This level is [1] Zoran R. Radakovic and Marko S. Sorgic, “Basics of Detailed Thermal-
determined by the position of the radiators and the distance Hydraulic Model for Thermal Design of Oil Power Transformers”, IEEE
TRANSACTIONS ON POWER DELIVERY, VOL. 25, NO. 2, APRIL 2010
between the outlet pipe and the bottom of the windings [2] “Transformer Thermal Modelling”, Cigre Working Group A2.38, 2016.
Figure (8). [3] IEC 60076-7, “Loading guide for oil-immersed power transformers”,
2005-12.

View publication stats