INVESTIGATION OF OIL FLOW AND HEAT

TRANSFER IN TRANSFORMER RADIATOR

A Thesis Submitted to
The Graduate School of Engineering and Sciences of
Izmir Institute of Technology
In Partial Fulfillment of the Requirements for the Degree of

MASTER OF SCIENCE

in Energy Engineering

by
Özben KAYMAZ

December 2015
İZMİR
We approve the thesis of Özben KAYMAZ

Examining Committee Members:

__________________________________
Assoc. Prof. Dr. Tahsin BAŞARAN
Department of Architecture, Izmir Institute of Technology

__________________________________
Prof. Dr. Gülden GÖKÇEN AKKURT
Department of Mechanical Engineering, Izmir Institute of Technology

__________________________________
Assist. Prof. Dr. Mehmet Akif EZAN
Department of Mechanical Engineering, Dokuz Eylul University

23 December 2015

________________________________ _______________________________
Assoc. Prof. Dr. Tahsin BAŞARAN Prof. Dr. Aytunç EREK
Supervisor, Department of Architecure, Co-Supervisor, Mechanical Engineering
Izmir Institute of Technology Dokuz Eylul University

_______________________________ ______________________________
Prof. Dr. Gülden GÖKÇEN AKKURT Prof. Dr. Bilge KARAÇALI
Head of the Department of Energy Dean of the Graduate School of
Engineering Engineering and Sciences
 ACKNOWLEDGEMENTS

This study would not have been possible without the support of my family, professors
and mentors, Balikesir Electromechanical Industrial Plants Co. (BEST Co.), my colleagues, and
my friends.
I express my warmest gratitude to my family for encourage me in all of my pursuit and
inspiring me to follow my dreams. I am especially grateful to my family, who helped me very
much, especially when I was working in Balikesir and studying in Izmir.
I would like to express my deep gratitude to Energy Engineering Department of Izmir
Institute of Technology professors and my advisor Assoc. Prof. Dr. Tahsin BAŞARAN for their
invaluable advice and encouragement throughout the research. I would like to give my special
thanks to my advisor for his endless supporting, leading, patience and motivation who always
guided me during my higher education. He was always energetic and enthusiastic. He gave me
the freedom I needed to move on and moral support when I was writing this thesis. I also thank
my co-advisor Prof. Dr. Aytunç EREK from Mechanical Engineering Department of Dokuz
Eylul University for his grateful advice and smartness.
I would very much like to BEST Co. and R&D Center for giving me this opportunity,
continual supports and encouragement during this study. I would also like to extend my deepest
gratitude to Dr. Gökhan KALKAN, who is the R&D Manager of BEST Co. for his kind help,
endless effort to teach all his knowledge, generous advice and supporting me on my way to
realize my dreams and my higher education. He was always solution oriented, whenever
something went wrong or disappointing occurred. Special thanks to my workmates and friends,
not only for understanding and supporting, but also for sincere friendship.
To all the people that I have met, thank you for sharing life with me and also for your
friendship. I have a motto; ‘Share Knowledge!’. Because this thesis would not have been
possible without sharing.
 ABSTRACT

INVESTIGATION OF TRANSFORMER OIL FLOW AND HEAT

TRANSFER IN TRANSFORMER RADIATOR

Transformer losses are produced by the current passing through resistance on the
winding conductors. These losses are converted to heat energy as all electrical machines.
Transformer life depends on the aging of cellulose-based insulation material in winding.
Winding temperature must be kept below a certain temperature limit, because the excess heat in
the windings directly accelerates the aging of insulation material. Oil-immersed transformers
are cooled by using transformer oil. In this study, flow and heat transfer of transformer radiator
filled with mineral oil, silicone oil and ester oil were investigated. Oil in radiator was modelled
in 3D, thereby oil flow volume was created, heat transfer calculations were made with using
different type transformer oils and investigation of flow behavior is studied. Temperature
dependent density, thermal conductivity and viscosity values are obtained from oil suppliers.
Transformer oil database was composed by using curve fitting methods in MathCAD Software
to use in CFD (Computational Fluid Dynamics) analysis in Fluent software. Finally, the
pressure differences were obtained for each oil type. According to the velocity versus pressure
difference graph that was obtained from the results of this study, inertial resistance and viscous
resistance could be described which are required in future studies to use porous medium
approach. Number of elements and computation domain will be decreased by using porous
medium approach, then complete transformer model could be analyzed. Additionally, thermal
characteristics of transformer oils were obtained and compared in this study. Accordingly,
natural ester oil had the best heat transfer and pressure drop.

iv
 ÖZET

TRANSFORMATÖR RADYATÖRÜNDE YAĞ AKIŞININ VE ISI

TRANSFERİNİN İNCELENMESİ

Transformatör kayıpları sargı iletkeni üzerinden geçen direncin oluşturduğu akımdan

kaynaklanır. Tüm elektrik makinelerinde olduğu gibi, oluşan kayıplar ısı enerjisine dönüşür.
Transformatör ömrü, sargıda bulunan selüloz bazlı izolasyon malzemesinin yaşlanmasına
bağlıdır. Sargı içerisinde oluşan fazla ısı doğrudan doğruya izolasyon malzemesinin
yaşlanmasını hızlandırdığı için, sargı sıcaklığı önceden belirlenmiş olan sıcaklık sınırlarında
tutulur. Yağlı tip transformatörler, transformatör yağı kullanılarak soğutulmaktadır. Bu
çalışmada, mineral yağ, silikon yağ ve ester yağ kullanılan transformatör radyatöründe akış ve
ısı transferi incelenmiştir. Radyatör içerisinde bulunan transformatör yağı 3 boyutlu olarak
modellenmiş, böylece yağ akış hacmi oluşturulmuş, farklı tipte transformatör yağlarının ısı
transferi hesaplamaları yapılmış ve akışın gösterdiği davranış incelenmiştir. Yağ
tedarikçilerinden sıcaklığa bağlı olarak değişen yoğunluk, ısı sığası, ısıl iletkenlik ve viskozite
değerleri sağlanmıştır. Transformatör yağına ait veriler, ANSYS Fluent simulasyon yazılımında
HAD (Hesaplamalı Akışkanlar Dinamiği) analizlerinde kullanılmak üzere, MathCAD
yazılımında eğri uydurma yöntemleri ile oluşturulmuştur. Sonuç olarak, rher bir yağ için
radyatör boyunca oluşan basınç farkları elde edilmiştir. Bu çalışma neticesinde gelecek
çalışmalarda gözenekli ortam yaklaşımının kullanılabilmesi amacı ile hıza bağlı olarak değişen
basınç farkları grafiği elde edilmiştir. Böylece gözenekli ortam modellemesinde gerekli olan
atalet direnci ve viskoz direnç katsayıları tanımlanabilecektir. Gözenekli ortam yaklaşımı
kullanılarak, eleman sayısı ve çözüm süresi düşürülecek, böylece tüm bir transformatör analizi
yapılabilecektir. Ayrıca, bu çalışma ile transformatör yağlarının karakteristikleri de elde
edilerek, birbirleriyle karşılaştırılmıştır. Buna göre, doğal ester yağ, ısı transferi ve basınç
düşümü açısından diğer yağlara göre daha iyi bir performans göstermiştir.

v
 TABLE OF CONTENTS

CHAPTER 1. INTRODUCTION ..................................................................................... 1

1.1. What is Transformer? ............................................................................. 5
1.2. Transformer Losses ............................................................................... 14
1.3. Cooling Modes ...................................................................................... 14
1.3.1. ONAN (Oil Natural - Air Natural) ................................................ 16
1.3.2. ONAF (Oil Natural - Air Forced) ................................................. 17
1.3.3. OFAF (Oil Forced - Air Forced) ................................................... 18
1.3.4. ODAF (Oil Directed Air Forced) .................................................. 18

CHAPTER 2. LITERATURE REVIEW ........................................................................ 20

2.1. Transformer Life ................................................................................... 20
2.2. Reducing the Hot-spot Temperature ..................................................... 21
2.3. Further Developments of New Methods ............................................... 24
2.4. Comparison Between CFD Simulations and Experiments ................... 25
2.5. Porous Media Approach in CFD Simulations ...................................... 30
2.6. Complete and Slice Models .................................................................. 31
2.7. Transformer Winding............................................................................ 35
2.8. Transformer Cooling Fluids .................................................................. 37
2.9. Overview of Thesis According to Literature ........................................ 39

CHAPTER 3. TRANSFORMER OILS .......................................................................... 41

CHAPTER 4. NUMERICAL ANALYSIS ..................................................................... 48

4.1. Problem Statement ................................................................................ 50
4.2. Modelling .............................................................................................. 51
4.3. Meshing................................................................................................. 53
4.4. General Settings .................................................................................... 55
4.5. Viscous Model ...................................................................................... 55
4.6. Material Properties ................................................................................ 56

vi
 4.7. Boundary Conditions ............................................................................ 58
4.8. Solution ................................................................................................. 60

CHAPTER 5. RESULTS AND DISCUSSION .............................................................. 62

5.1. CFD Results .......................................................................................... 63
5.1.1. Temperature Distribution .............................................................. 63
5.1.2. Velocity Vectors............................................................................ 65
5.1.3. Velocity Streamlines ..................................................................... 66
5.1.4. Heat Flux ....................................................................................... 69
5.1.5. Comparison of Oils ....................................................................... 71
5.1.6. Dimensionless Numbers................................................................ 74
5.2. Mesh Independence .............................................................................. 82

CHAPTER 6. CONCLUSION ....................................................................................... 84

REFERENCES ............................................................................................................... 86

vii
 LIST OF FIGURES

Figure Page
Figure 1. Electrical Transmission. .................................................................................... 2
Figure 2. Transformer Core Steel. .................................................................................... 3
Figure 3. Power Transformer. ........................................................................................... 4
Figure 4. Oil-immersed Arc Furnace Transformer. .......................................................... 6
Figure 5. Dry-type Transformer........................................................................................ 6
Figure 6. Heat Transfer at Radiator Section. .................................................................... 7
Figure 7. Transformer Main Parts..................................................................................... 8
Figure 8. Transformer Core. ............................................................................................. 8
Figure 9. Laminated Coated Core. .................................................................................... 9
Figure 10. Step-lap and Books. ......................................................................................... 9
Figure 11. Transformer Winding. ................................................................................... 10
Figure 12. Winding Insulated Conductors. ..................................................................... 11
Figure 13. Transformer Tank. ......................................................................................... 12
Figure 14. Corrugated Transfromer Tank. ...................................................................... 13
Figure 15. Transformer Parts. ......................................................................................... 13
Figure 16. Transformer with Radiators and Fans. .......................................................... 15
Figure 17. Radiator. ........................................................................................................ 16
Figure 18. Transformer Radiators. .................................................................................. 17
Figure 19. Transformer and Radiators. ........................................................................... 18
Figure 20. OFAF (left) and ODAF (right) Cooling Modes. ........................................... 19
Figure 21. Temperature Distribution on Radial Spacers. ............................................... 21
Figure 22. Temperature Distribution of Detailed and Simplified Disc Model for Two
Types. ............................................................................................................. 23
Figure 23. Temperature Distribution of Disc for the 2D Model. .................................... 24
Figure 24. Radiator Model. ............................................................................................. 26
Figure 25. Cooling Capacity versus Flow Rate. ............................................................. 27
Figure 26. Temperature versus Distance From Radiator Top to Bottom. ...................... 28
Figure 27. Temperature Distribution at the Flow Rate of 0.001 m3/s. ........................... 29
Figure 28. Temperature Distribution at the Flow Rate of 0.004 m3/s. ........................... 29
Figure 29. Temperature Distribution in the Radiators. ................................................... 30

viii
Figure 30. Temperature Difference (T_(oil,top) – T_(oil,bottom)) for Different Fan
Configurations. ............................................................................................... 31
Figure 31. Developed Flow Domains for Second Transformer with Complete (left) and
Slice Model (right). ........................................................................................ 32
Figure 32. Position of the Thermocouples During the Experimental Tests and Geometry
of the Complete and Slice Model. .................................................................. 33
Figure 33. Oil Temperature Distribution in a Vertical Mid-plane of the Fins in Kelvin:
Transf-01 (left), Transf-02 (middle), Transf-03 (right).................................. 34
Figure 34. The Isotherms at the Top of Transformer for Different Inlet Velocities....... 36
Figure 35. Miscibility of Alternative Fluids at Ambient Temperature. .......................... 37
Figure 36. Transformer Oil Types. ................................................................................. 41
Figure 37. Duval Triangle. .............................................................................................. 42
Figure 38. Density Graph versus Temperature. .............................................................. 44
Figure 39. Specific Heat versus Temperature. ................................................................ 45
Figure 40. Thermal Conductivity versus Temperature. .................................................. 46
Figure 41. Dynamic Viscosity versus Temperature........................................................ 47
Figure 42. Imported radiator geometry in Design Modeler. ........................................... 48
Figure 43. Radiator model and directions. ...................................................................... 49
Figure 44. Transformer Radiator. ................................................................................... 51
Figure 45. Radiator Model in Catia V5. ......................................................................... 52
Figure 46. Simplified Radiator Model. ........................................................................... 52
Figure 47. Mesh of the Oil Ducts. .................................................................................. 53
Figure 48. Inlet/Outlet Mesh Type and Inflation. ........................................................... 54
Figure 49. Section of Oil Duct. ....................................................................................... 54
Figure 50. Mesh of the Radiator Model. ......................................................................... 55
Figure 51. Cross-section of Radiator Oil Duct (top) and ¼ Radiator Oil Duct
(bottom) ......................................................................................................... 56
Figure 52. Dynamic Viscosity versus Temperature Graph of the First Interval for
Natural Ester Transformer Oil........................................................................ 57
Figure 53. Dynamic Viscosity versus Temperature Graph of the Second Interval for
Natural Ester Transformer Oil........................................................................ 58
Figure 54. Boundary Conditions. .................................................................................... 58
Figure 55. BEST Co. Test Laboratory. ........................................................................... 59
Figure 56. Scaled Residuals. ........................................................................................... 61

ix
Figure 57. Temperature Distribution at the Inlet Zone with 0.05 m/s Inlet Velocity for
Natural Ester Oil............................................................................................. 63
Figure 58. Temperature Distribution at the Outlet Zone with 0.05 m/s Inlet Velocity for
Natural Ester Oil............................................................................................. 64
Figure 59. Temperature Distribution Along Simulated Quarter Radiator. ..................... 64
Figure 60. Velocity vectors at the inlet with constant 0.05 m/s inlet velocity................ 65
Figure 61. Velocity Vectors at the Outlet with Constant 0.05 m/s Inlet Velocity. ......... 65
Figure 62. Streamlines at Inlet of Natural Transformer Oil with 0.05 m/s..................... 66
Figure 63. Velocity Streamline with 0.05 m/s Inlet Velocity at Inlet Boundary Condition
for Natural Ester Transformer Oil. ................................................................. 67
Figure 64. Velocity Streamline with 0.05 m/s Inlet Velocity at Outlet Boundary
Condition for Natural Ester Transformer Oil. ................................................ 67
Figure 65. Recommended radiator entry zone. ............................................................... 68
Figure 66. Heat Flux at the Inlet Zone with 0.05 m/s Inlet Velocity for Natural Ester
Transformer Oil. ............................................................................................. 69
Figure 67. Heat Flux at the Inlet Zone with 0.05 m/s Inlet Velocity for Natural Ester
Transformer Oil. ............................................................................................. 70
Figure 68. Heat Flux with 0.05 m/s Inlet Velocity for Natural Ester Oil. ...................... 70
Figure 69. Velocity versus Pressure Drop Graph. .......................................................... 73
Figure 70. Velocity versus Heat Flux Graph. ................................................................. 74
Figure 71. Cross-sectional Area at Inlet. ........................................................................ 75
Figure 72. Cross-section Area of Oil Ducts. ................................................................... 76
Figure 73. Wetted Perimeter of Oil Duct Cross-section Area. ....................................... 77
Figure 74. Nusselt Number versus Reynolds Number Graph. ....................................... 80
Figure 75. Edge Sizing at Inlet Zone with Bias. ............................................................. 82
Figure 76. Edge Sizing at Inlet Zone without Bias. ........................................................ 83

x
 LIST OF TABLES

Table Page
Table 1. Transformer Coolant Types. ............................................................................. 15
Table 2. Circulation Types.............................................................................................. 16
Table 3. Experimental Temperature Measurements. ...................................................... 32
Table 4. Comparison Between Complete and Slice Models. ......................................... 34
Table 5. Basic Properties of Synthetic Ester, Natural Ester and Mineral Oil. ................ 38
Table 6. Duval's Triangle Fault Zone. ............................................................................ 42
Table 7. Natural Ester Transformer Oil Results. ............................................................ 71
Table 8. Mineral Transformer Oil Results. ..................................................................... 72
Table 9. Silicone Transformer Oil Results. .................................................................... 72
Table 10. Synthetic Transformer Oil Results. ................................................................ 72
Table 11. Reynolds Number of Natural Ester Transformer Oil Flow. ........................... 77
Table 12. Theoretical Nusselt Number for Natural Ester Transformer Oil Flow. .......... 78
Table 13. Theoretical Prandtl Number for Natural Ester Transformer Oil Flow. .......... 79
Table 14. Calculated Dimensionless Numbers at Different Velocities. ......................... 80
Table 15. Mesh Independence. ....................................................................................... 82

xi
 LIST OF SYMBOLS

Abbreviations
A Area [m2]
c Specific thermal capacity [J/kg.K]
Dh Hydraulic diameter [m]
g Gravitational acceleration [m/s2]
k Thermal conductivity [W/m.K]
I1 Current flowing through primary terminal [Ampere]
I2 Current flowing through secondary terminal [Ampere]
T Temperature [˚C or K]
T0 Reference temperature [˚C or K]
h Height [meter]
Tinlet Inlet temperature [˚C or K]
L Length [meter]
l Length [meter]
N1 Number of turns of the primary winding [-]
N2 Number of turns of the secondary winding [-]
S Distance [meter]
P Pressure [Pascal]
Ptotal Total cooling capacity of a radiator [Watt]
Pfin Cooling capacity of a fin [Watt]
Q Heat transfer [Watt]
Q oil Oil flow rate [m3/s]
R Resistance [Ohm]
t Time [Second]
u Vector along x axis [-]
V1 Voltage drop at the primary terminal [-]
V2 Voltage drop at the secondary terminal [-]
v Vector along y axis [-]
w Vector along z axis [-]

xii
Greek Letters
ρ Density [kg/m3]
ρe Electrical resistivity [ohm∙meter]
μ Dynamic viscosity [kg/m.s]
β Thermal expansion coefficient [-]
α Slope of the temperature rise [-]
ρ0 Density of oil at the reference temperature [kg/m3]
τwall Wall shear stress [Pascal]

xiii
 CHAPTER 1

INTRODUCTION

All the photos that have not any reference were used with the permission of
Balikesir Electromechanical Industrial Plants Co.
Energy is directly related with the quality of human lives. The world-wide
energy consumption is hugely increasing. In addition, dramatic increase in demand for
oil is seen in recent years.
Electricity is generated in power plants from hydrothermal, natural gas, wind
energy, solar energy and fossil fuel. These power sources rotate large turbines which are
connected to generators at their other ends. Generated electricity in the generators is
either 50 Hz. or 60 Hz. The maximum voltage level of the generated electricity is 20
kV. At this voltage level in order to transmit the electricity from the source to the end
consumer large cable cross-sectional areas are required. This is definitely an expensive
solution as the conductors using transmission lines are copper or aluminum. Moreover,
DC losses in other loss I2R losses are proportional to cross-sectional area of the used
conductor. Therefore, at low voltage levels the losses occurred in the long transmission
line generates excessive losses and the efficiency is reduced dramatically. As the
resistance is formulated as,

l
R = ρe (1.1)
A

where, ρe is electrical resistivity, l is the length of the transmission line and A is the
cross-sectional area of the current carrying conductor, the only solution to reduce the
total cost of the transmission line is to decrease the cross-sectional area of conductor.
Volt/turns of the transformer at the primary and the secondary terminals has to
be equal. This is formulated as,

V1 V2
= (1.2)
N1 N2

1
where, V1 is the voltage drop at the primary terminal, V2 is the voltage drop at the
secondary terminal, N1 and N2 are the number of turns of the primary and secondary
windings, respectively.
Another important equation of transformers relates current of the terminals to
the number of turns. This is formulated as,

N1 I1 = N2 I2 (1.3)

where, I1 and I2 are the currents flowing through primary and secondary terminals,
respectively.
From Equations (1.2) and (1.3), it can be seen that increasing the voltage at the
terminal reduces the current. In other words, current and voltage are inversely
proportional. This means that increasing the voltage level will generate substantial
amount of decrease in the line current. Therefore, the losses occurs on the transmission
line can be reduced.
Electrical energy is transmitted from power plants to electrical substations.
Electricity is more efficiently transmitted at high voltages. Power transformers transmit
the electricity at high voltages into substations. Distribution transformers convert the
electricity to lower voltages. Electricity transmission steps were visualized by Southern
Nevada Regional Professional Development Program as shown in Figure 1.

Figure 1. Electrical Transmission.

(Source: Retrieved October 3, 2015 from World Wide Web:
http://www.rpdp.net/sciencetips_v2/P12C6.htm)

2
 Transformers are static devices with high operating efficiency. The only
component which reduced transformer efficiency is the total losses occurring on them.
The losses are divided into two main subgroups. These are namely magnetic originated
and electrical originated losses. Electrical originated losses are I2R losses occurring on
both primary and secondary windings. These are relatively straightforward losses to
compute. The real challenge is to compute magnetic originated losses. For instance,
eddy losses due to changing magnetic field generate circulating currents on both copper
conductors and metallic parts. These losses are heavily dependent on conductor
dimensions, temperature of conductor, the frequency and the thickness of the conductor.
Another important loss occurs on magnetic path of the core. The core material shown in
Figure 2, is grain oriented electrical steel, where the magnetic domains are subjected to
sinusoidal magnetic flux. This will force magnetic domains to rotate. During this
rotation, magnetic domains generate heat energy. Additionally, the changing magnetic
field causes eddy currents in the core material according to Faraday’s law. These two
effects are called no-load losses and these losses occur only on magnetic path.

Figure 2. Transformer Core Steel.

The above mentioned no-load and load losses generate excessive heat on the
transformer. The generated heat should be dissipated to the surrounding atmospheric
medium in order to keep the hottest spot temperature below the allowed limit. This limit
is dependent on the temperature class of the material used. In power transformers
(Figure 3), natural cellulose based insulation materials are used due to their excellent
electrical insulation properties. These insulation materials suffer from a process called

3
thermal aging. Thermal aging degrades the mechanical properties of the cellulose based
insulation material and these materials become brittle over a certain temperature. Aging
is basically an attack of OH components of transformer oil to cellulose fibers. Cellulose
fibers become shorten as a result of this chemical process. This chemical process is
function of temperature and for A Class insulation materials the temperature limit is
118˚C.
Transformers are cooled by using different cooling modes according to type and
power of transformer. Transformer cooling systems have an important role in
transformer life.

Figure 3. Power Transformer.

This study provides information and guidance to researchers who are interested
in thermal and hydrodynamic analysis of transformers. The main aim of this study is to
investigate of a transformer radiator with different transformer oil types to obtain the
pressure difference with respect to velocity. Then, viscous resistance and inertial
resistance coefficients can be obtained easily to study with porous medium approach.
This study provides a detailed view and analysis about transformer radiators and the
thermal behavior of different types of transformer oils.

4
 General information about energy and transformer will be given in Chapter 1.
Additionally, transformer losses and cooling modes were covered. Chapter 2 presents
the studies on transformer life, reducing hot-spot temperature, further developments of
new methods, comparison between CFD simulations and experiments, porous media
approach in CFD simulations, complete and slice models, transformer winding and
transformer cooling fluids. Overview of this study according to presented literature
review was presented also in Chapter 2. Transformer oils and their properties were
explained in Chapter 3. Then, numerical analysis that covers the methodology of this
problem was discussed in Chapter 4. Problem statement, modelling, meshing, general
settings, viscous model, material properties, boundary conditions and solution were
considered in Chapter 4. Finally, Chapter 5 concludes this study by summarizing the
results and discussion. Dimensionless numbers were calculated in Chapter 5. Also,
correlation was developed between Reynolds number and Prandtl number to obtain
Nusselt number. In conclusion, the aim of study was summarized. Additionally, further
work developments were explained in conclusion chapter.

1.1. What is Transformer?

Transformers are electrical machines that transform alternating current energy

from one voltage to another voltage without a change in frequency. The first developed
transformer was designed in the early 1980s. The working principle of transformers
depends on the Faraday’s law of induction. Voltage is generated by moving the magnet
near a wire according to Faraday’s law. This voltage is called electromotive force (emf)
and denoted 𝜀. Electromotive force is directly proportional to the rate of change of
magnetic flux.
Transformers can be grouped into two classes according to cooling method; oil-
immersed transformers (Figure 4) and dry type transformers (Figure 5). In this study,
heat transfer and fluid flow were investigated in a radiator of an oil-immersed
transformer.

5
 Figure 4. Oil-immersed Arc Furnace Transformer.

Air is used for transformer cooling in dry-type transformers. Air is directly

forced into core and winding by using fans. This type of transformers have the power
limits because their losses are higher. Air is used for cooling fluid to dissipate heat.

Figure 5. Dry-type Transformer.

The generated heat is dissipated through the cooling systems according to the
transformer type. In oil-immersed transformers, excessive heat is dissipated through

6
transformer oil. The heat transfer from the conductors and the magnetic path is
transmitted to the transformer oil by conduction and convection. The heat transfer by
conduction is very small compared to convection heat transfer. The parameters that
effects heat transfer are conductivity of the insulation materials, winding cooling duct
dimensions, types of winding, disc width, spacer thickness, distance between discs,
overall winding dimensions, thermal properties of transformer oil type and the velocity
of transformer oil. The heat transmitted from windings to the transformer oil is
circulated into transformer radiators. The heat transfer in the radiators is due to
conduction, convection and radiation. Conduction heat transfer is extremely small
compared to convection type heat transfer as shown in Figure 6. Toil is the oil
temperature that is circulating in radiator, Twall is the temperature of radiator steel and
Tair is the ambient temperature. On the other hand, radiation heat transfer is negligible
in transformers according to literature.

Figure 6. Heat Transfer at Radiator Section.

The aged insulation material becomes brittle and ruptures, if the generated heat
is not dissipated to the surrounding medium. This exposes to conductor material to
touch each other and two adjacent conductors short circuit. This generates excessive
currents to flow between two conductors and the transformer explodes. This is a
catastrophic failure accompanied with a fire.
Transformers consist of four main parts as core, winding, tank and accessories
as shown in Figure 7.

7
 Figure 7. Transformer Main Parts.

1.1.1. Core

Magnetic flux is generated in winding by the current flowing through the

windings. Transformer core is used to create a magnetic flux path (Figure 8). The core
is laminated to reduce the eddy current losses in transformers (Figure 9).

Figure 8. Transformer Core.

8
 Laminated silicon steel core is made up of insulated thin laminations. Silicon
steel core provides low hysteresis loss and high permeability.

Figure 9. Laminated Coated Core.

Step-lap method is used to construct transformer core. Transformer core consists

of steps and books as shown in Figure 10.

Figure 10. Step-lap and Books.

9
1.1.2. Winding

The winding is composed of the wire conductors as shown in Figure 11, which
are wounded around the core, and these are covered with cellulose-based insulation
material.

Figure 11. Transformer Winding.

Conductors are insulated to avoid electrical breakdown, creepage and partial

discharge. Special paper is used for insulation of winding that is called insulation paper
as seen Figure 12.

10
 Figure 12. Winding Insulated Conductors.

The source of magnetic flux is primary winding. Magnetic flux tends to pass
through the core that has high magnetic permeability. The relative permeability of
transformer core material is nearly 20000 times higher than air. Thus, magnetic flux is
induced to secondary winding. Windings and the core together forms the active part
located in transformer tank. The main leakage field in transformers occurs between low
voltage and high voltage windings. As a result, excessive eddy losses and eddy currents
acts on these materials. In order to decrease the values of eddy currents, the transformer
conductors are divided into many small pieces and each piece is isolated from neighbor
conductors with special polymer based epoxies. This application helps reducing eddy
current and loss values in windings. However, every divided conductor has to be
transposed in order not to generate circulating currents due to resistive differences.

1.1.3. Tank

Transformer tank (Figure 13) is a mechanical element of transformer that

provides a protection by surrounding transformer oil and active parts. Most of the
transformer equipments are mounted on the transformer tank. Leakage magnetic fields
generate circulating eddy currents on the tank wall. Therefore, appropriate magnetic
shunts or magnetic screening has to be provided. Excessive heating will occur on the
tank wall, if magnetic shunts or flux rejecters are not provided. The temperature due to
these circulating currents can be as high as 300˚C locally. Therefore, conductors
carrying high currents generates additional heat source in transformers.

11
 Transformer tank is designed to withstand 2 bars of hydrostatic pressure.
Considering a power transformer dimensions which can reach values as high as 12
meters special stiffeners has to be provided. Otherwise, large and permanent deflections
occur on the transformer tank, which are not allowed according to IEC Standards.

Figure 13. Transformer Tank.

Corrugated tanks (Figure 14) are frequently used for distribution transformers as
cooling fins. Transformer losses are dissipated through the corrugated walls. However,
corrugated walls suffer from excessive internal pressure and limit the transformer sizes
due to their low mechanical strength. On the other hand, due to their flexibility they can
be work in a completely closed transformer design, namely hermetically sealed
transformer. In this study, hermetically sealed transformer design was not focused on.

12
 Figure 14. Corrugated Transfromer Tank.

1.1.4. Accessories

Tap changers, bushings, moisture holders, pressure safety valves, throttle valves,
buchholz relay and thermometers are some of the transformer accessories as seen in
Figure 15.

Figure 15. Transformer Parts.

13
1.2. Transformer Losses

Transformers are the kind of electrical machines that have energy losses as all
electrical machines. Two main types of losses occur in transformers; copper losses and
iron losses.
Copper losses are produced in transformer windings and defined with I2R loss.
Copper losses are the product of electrical currents in the conductors of the transformer
windings. Hysteresis loss and eddy current loss are both iron losses that occur in core
depending on ferromagnetic material properties. The magnetic flux passing through the
core generates circulating eddy currents because of the Faraday’s law of induction. This
circulating current generate additional losses and heats up the core. As the cross-
sectional area of core gets larger, the losses increase. In order to decrease of the amount
of dissipated losses, the core builds up from very thin laminations. The surface of each
lamination is coated. Thin and laminated iron sheets are used in transformer core to
decrease the eddy currents.

1.3. Cooling Modes

Cooling modes in oil-immersed transformer will be explained in this part.

Transformers work in specified temperature mentioned in loading guide for oil-
immersed power transformers part (Part 7) of IEC 60076-7 International Standards
(2005-12). All losses (iron losses and copper losses) are converted to heat energy in the
transformer. Copper losses are the main source of the heat energy. The occurred heat
energy increases the temperature of transformer. Transformer coolant is used to avoid
overheating in transformer. Insulation paper is damaged by excess heat which is
discarded by using transformer oil. Especially, radiators are used to increase the heat
transfer surface area in power transformers. Fans and pumps are also used in power
transformers as shown in Figure 16 that cannot be cooled using only radiators. Fans are
used to drive the air on the radiator outside surfaces and pumps are used to circulate the
oil into transformer tank or radiators.

14
 Figure 16. Transformer with Radiators and Fans.

The generated heat in the core and winding is transferred to the oil inside the
transformer. Transformer cooling systems are classified with four letter designation.
The first letter indicates the type of cooler. Second letter is to show whether the cooler
circulates natural or forced. Third letter designates the type of cooler that moving on the
outer cooling surfaces. The fourth letter shows the circulation type for the cooler
moving on the outer surfaces. The main types of cooling modes for oil-immersed
transformers can be listed as below in Table 1 and Table 2.

Table 1. Transformer Coolant Types.

Cooling Type Symbol

Oil O

Gas G

Water W

Air A

Solid S

15
 Table 2. Circulation Types.
Circulation Type Symbol
Natural N
Forced F
Directed D

1.3.1. ONAN (Oil Natural - Air Natural)

ONAN cooling mode is the basic cooling system in oil-immersed transformers.

The heated fluid flows upward and flows into the radiator from the inlet as shown in
Figure 17.

Figure 17. Radiator.

Oil flows through the radiator ducts and cooled oil is entered into the tank
(Figure 17). Oil is circulated in the space of winding and core. Air flows through the
outside surfaces of the radiators by natural convection. Also, oil collector pipes are
shown in Figure 18. Oil is collected from the upper oil collector pipe into radiator.
Then, cooled oil is transmitted into tank from the bottom oil collector pipe. Circulating

16
oil in transformer and radiator is very slow in ONAN mode. Basically, the principle of
thermosiphon effect provides the circulation of transformer oil in transformer and
radiator. Temperature difference between the transformer oil in tank and transformer oil
in radiator causes pressure difference, then this pressure difference circulates the
transformer oil.

Figure 18. Transformer Radiators.

1.3.2. ONAF (Oil Natural - Air Forced)

In ONAF mode, fans are blown to the cooling surfaces whereas the oil circulates
naturally inside the radiators. Fans are generally mounted below of the radiators. The air
flow rate is higher than air natural mode. ONAF mode is mostly used in power
transformer. ONAN/ONAF cooling modes combination can be also used in
transformers. In this type of cooling mode, natural cooling is used up to certain load.
Fans are worked automatically when the load exceeds the certain loads. Heat dissipation
rate in ONAF mode is higher than ONAN mode.

17
 1.3.3. OFAF (Oil Forced - Air Forced)

Pumps and fans are used to move coolant in OFAF mode. Oil is forced to
circulate in the transformer. Air is forced by using fans for cooling purposes of the
circulated oil.

1.3.4. ODAF (Oil Directed Air Forced)

Oil is directed into the winding and air blowing is obtained by using fans in
ODAF mode. Directed oil flows in oil ducts of winding. ODAF mode is generally used
in high rate power transformers.

Figure 19. Transformer and Radiators.

18
 Kulkarni and Khaparde (2005) remark the static electrification in large power
transformers. Higher the pump capacity the higher the heat transfer rate trend was used
until the some power transformers failed by static electrification phenomenon. Static
electrification phenomenon depends on moisture content of transformer oil,
temperature, flow rate, turbulence, surface conditions, pumps, orifices and AC/DC
fields. Therefore, pump selection in OFAF and ODAF cooling modes has an important
role in power transformer design. Consequently, higher the pump capacity does not
mean the higher heat transfer rate.
Additionally, OFAF and ODAF cooling modes are explained by Kulkarni and
Khaparde (2005). Transformer oil is forced into transformer tank in OFAF cooling
mode, while it is directed into the winding in ODAF cooling mode as seen in Figure 20.

Figure 20. OFAF (left) and ODAF (right) Cooling Modes.

(Source: Kulkarni and Khaparde, 2005)

19
 CHAPTER 2

LITERATURE REVIEW

This chapter will explore the literature that is relevant to understanding the
importance of thermal and hydrodynamic studies on the transformers. Each section will
summarize the main point of studies respectively. Transformer life, reducing the hot-
spot temperature, further developments of new methods, comparison between CFD
simulations and experiments, porous media approach in CFD simulations will be
covered. Furthermore, a study with complete and slice models is reviewed in this
chapter. Then, some of the studies on transformer winding hot spot temperature were
presented. Many studies have been performed regarding the thermal behavior of
transformer winding that has been presented in this chapter. On the other hand, not
many researches have been found in the literature on the thermal and hydrodynamics
behavior of transformer cooling systems. Thermal and hydrodynamic studies in
transformer winding will be investigated in the light of this study in the near future. In
this respect, this study plays an important role in characterization of thermal and
hydrodynamic properties of transformer cooling systems.
Oil-immersed transformers are preferred that have dielectric oil for cooling and
insulating as noted by Fernández et al. (2012). Mineral oil that has been widely used in
the power transformers, although is not biodegradable. Studies about transformer oils
and alternative cooling fluids are presented in this chapter.

2.1. Transformer Life

Winding temperature must be kept below a certain temperature limit, because

the excess heat in the windings directly accelerates the aging of insulation material.
Transformer life depends considerably on the aging of this cellulose-based insulation
material in winding.
Wittmaack (2014) was stated that chemical degradation in the electrical
insulation affects transformer life. Load losses were calculated by using Maxwell
electromagnetic field simulation software. The study aims at predicting the temperature

20
distribution on cellulose based and nomex insulating materials with CFD. The steady-
state CFD model used in the analysis contains curvilinear, non-orthogonal block
structured mesh elements. Temperature dependent material properties have been
utilized. Continuity, momentum and energy equations were solved by SIMPLE,
SIMPLEC or PISO algorithms. The study did not give any solid conclusions occurring
on insulation materials but mentions that pressure drop and temperature distribution can
be obtained. Figure 21 shows the temperature distribution on radial spacers.

Figure 21. Temperature Distribution on Radial Spacers.

(Source: Wittmaack, 2014)

2.2. Reducing the Hot-spot Temperature

CFD simulations have an important role in power transformer design, because of

the limitations of experimental measurements. Normal life expectancy of a power
transformer is about 20 years. Fonte et al. (2011) stated that the hot-spot temperature
can be reduced by as much 10˚C which has a strong direct impact on the power
transformer’s lifetime. 2D axis-symmetric CFD models were simplified. Flow and heat
transfer equations were solved by using Fluent 6.3 from ANSYS Inc. Power
transformers working with ODAF and ONAF cooling systems were simulated in this
study. Both 2D and 3D models were generated for ODAF cooled transformer model.
Hydrodynamic and thermal effects of radial spacers were not taken into account.
800000 elements were used with mesh refinement to construct grids in 2D CFD model.
On the other hand, 3D model was simulated with using 2450000 tetrahedral/hexahedral

21
elements. Only the 2D computational grid was created for ONAF cooled transformer.
Naphthenic oil was used as coolant in this power transformer and small temperature
difference within the oil was investigated. Boussinesq model was used to specify
density for the natural convection flow. Then, the calculated equation of temperature
based density is seen as below.

ρ = 868[1 − 6.40 × 10−4 (T − 293)] (2.1)

where ρ is used to indicate density of naphthenic transformer oil in kg/m3 and T is the
temperature in K.
Measurements of viscosity were performed by using cone/plate rheometer in the
temperature range 20˚C to 80˚C. Then, temperature based viscosity equation was
obtained as shown below.

3480
μ = 1.43 × 10−7 exp ( ) (2.2)
T

where μ is the viscosity of the naphthenic transformer oil in Pa∙s and T is the
temperature in K. Moreover, specific heat capacity and thermal conductivity were
assumed constant as cp = 2016 J kg −1 K −1 , and k oil = 0.126 W m−1 K −1 , respectively.
Disc model was simplified using calculated equivalent thermal resistance.
Simulations were studied with using two types of conductors to validate simplified
conduction model. Rectangular bar was used in Type 1 and copper transposed cable was
used in Type 2 as shown in Figure 22. The difference between detailed and simplified
disc model in both types was obtained as around 10% as seen in Figure 22.
Different boundary conditions were identified for both ODAF and ONAF
cooling systems. Boundary conditions were used equal for ODAF cooled transformer at
the inlet and outlet in 2D model simulations and the flow rate was obtained by
describing the pressure difference. On the other hand, flow rate was increased by the
pressure drop along the spacers in 3D disc model.

22
 Figure 22. Temperature Distribution of Detailed and Simplified Disc Model for Two
Types.
(Source: Fonte et al., 2011)

For ONAF cooling systems, where the oil flows by natural convection, pressure
is corrected at boundary conditions with calculating height-dependent density value.

ρ = ρ0 [1 − β(Tinlet + αh − T0 )] (2.3)

where ρ is the density, ρ0 is the density of oil at the reference temperature, T0 is the
reference temperature, h is the height, α is the slope of temperature rise, Tinlet is the
inlet temperature of the winding and β is the oil thermal expansion coefficient. Pressure
drop between the inlet and outlet was calculated as shown below.

h
∆Prelative = ∫h outlet ρgdh − ρop g(houtlet − hinlet ) (2.4)
inlet

According to the obtained results, 3D model was gave the best estimate for the
experimental temperatures. As a result of this study, the position and magnitude of the
hot-spot were determined on the third disc from the top as shown in Figure 23. As seen
in this study, CFD simulations of transformers provide many advantages.

23
 Figure 23. Temperature Distribution of Disc for the 2D Model.
(Source: Fonte et al., 2011)

Deviations were observed in both ODAF and ONAF cooling system because of
the using of a simplified heat conduction model.

2.3. Further Developments of New Methods

Simplified network approach is used for thermal design in literature. Joshi and
Deshmukh (2004) claimed that a new method used in developed new software is more
accurate than conventional method to define oil rise and winding rise. Temperature
based oil properties were defined at each node in network. Transformer winding was
modelled as disc and crossover winding. Oil guiding washers (OGW) are used to direct
oil flow in winding. Oil and air are circulated naturally in ONAN cooling system
whereas air is forced with fans in ONAF. In addition, oil is forced with using pump in
OFAF. The effects of fans and pumps were investigated. Radiators were modelled as
flat plated to simplify calculations in this study. Therefore, flow of oil in radiators
cannot be examined. Tank walls were also modelled as flat plates. Indeed, this study
was focused on the oil flow in winding. The methodology of thermal predictions of the

24
complete transformer model is discussed in this study. Average winding rise, top oil rise
and hot spots were predicted with different cooling systems for instance, ONAN, ONAF
and OFAF. Heat is transferred to air by both convection and radiation. Total heat
transfer is found by summing convection heat transfer and radiation heat transfer.

Q total = Q convection + Q radiation (2.5)

The external air flow was determined as laminar in the initial part of fin surfaces
according to calculated Reynolds number and the flow was followed by turbulent flow.
Heat is transferred from the tank and fins surfaces slowly, for this reason they
were assumed as isothermal surfaces. Pressure drop of oil moving toward the bottom
collector pipe increases inside the oil ducts of fins. Unlike the radiator cooling system,
oil pressure decreases with moving upwards in tank.
Moreover, buoyancy and pump create flow inside the windings. Network model
was used to calculate pressure drop and oil temperatures in winding. Pressure decreases
along the winding ducts by the effects of gravity, friction and bends. Local heat transfer
coefficient was calculated with using average temperatures and velocities around disc.
In some ducts, reversed oil flow was observed, then it was verified with using ANSYS
CFD simulation software. Calculated disc temperatures were used to determine the
average winding temperature in this study by using an iterative method. According to
this study, this new method is more confident than conventional method. At this point, it
is obviously clear that transformer cooling calculations should be improved with new
methods.

2.4. Comparison Between CFD Simulations and Experiments

Verification and validation are the most important processes to increase

confidence in CFD simulation. There is an example study from literature examined that
ONAN and ONAF cooling system with experiments. Cooling performance of radiators
used in oil-filled power transformers was studied with using two kinds of cooling
system by Kim et al. (2012). ONAN and ODAF cooling systems were used to
investigate the effects of non-direct and direct-forced oil flow. This study consists of

25
detailed calculations, CFD simulations and experimental results. Simulation was
modelled with 4 radiators that each has 40 fins as shown in Figure 24.

Figure 24. Radiator Model.

(Source: Kim et al., 2012)

Center to center length of each fin is indicated by L, which is equal to 3300 mm

and gap between fins are indicated by S, which is equal to 45 mm. The oil volume flow
rate was varied in the range from 1.0 × 10−3 m3 /s to 2.0 × 10−3 m3 /s.
Inlet temperature of radiator was assumed to be equal to the hot-spot
temperature. Calculations of the temperature distribution and cooling capacity were
verified with CFD simulations. Flow rate, inlet temperature and outlet temperature were
measured, then compared with calculations and the CFD simulation results for both
cooling methods.
Kim et al. stated that, temperature difference between inlet and outlet decreases
as an increasing in flow rate. Thus, this work is obviously showed that the cooling
capacity increases at high flow rate as shown in Figure 25.

26
 Figure 25. Cooling Capacity versus Flow Rate.
(Source: Kim et al., 2012)

Oil temperature increases at the distance from radiator top to bottom by

increasing in oil flow rate. Temperature versus distance graph with different flow rates
is shown below. As shown in Figure 26, it is seen that the oil temperature is higher for
4.0 × 10−3 m3 /s flow rate at the same distance. Although the flow rate was defined up
to 2.0 × 10−3 m3 /s, higher flow rate 4.0 × 10−3 m3 /s was obtained by using a pump.

27
 Figure 26. Temperature versus Distance From Radiator Top to Bottom.
(Source: Kim et al., 2012)

The total heat transfer coefficient was calculated using equation as shown below
in Equation (2.6).

Ptotal = ∑40
N=1 Pfin (N) = ρcp Q oil (Ttop − Tair )(1 − e
−(hp )
) (2.6)

where Pfin is the cooling capacity of a fin, Ptotal is the cooling capacity of the total
cooling capacity of a radiator, which have 40 fins. ρ and cp are the material properties
of oil, which are dependent on temperature. Q oil is indicated the oil volume flow rate.
Grid independent study is important for CFD simulations. Grid independency
was studied by using same boundary conditions to verify the accuracy of the results.
Over 7.3 million elements were used, then 8 million was chosen according to the result
of the grid independency study.
Flow was defined as steady state and incompressible. Viscous model was
constructed with using standard k-ɛ model. Boundary conditions were specified as mass
flow rate at the inlet and out flow at the outlet. Inlet temperature was set up 75˚C. The
convergence criteria were defined as under 1.0 × 10−6 . Outlet temperature was gained
at 60.39˚C with 1.0 × 10−3 m3 /s volume flow rate and 69.67˚C with 4.0 × 10−3 m3 /s

28
volume flow rate as shown in Figure 27 and Figure 28. Also, temperature distributions
at the given volume flow rates are seen in Figure 27 and Figure 28.

Figure 27. Temperature Distribution at the Flow Rate of 0.001 m3/s.

(Source: Kim et al., 2012)

Figure 28. Temperature Distribution at the Flow Rate of 0.004 m3/s.

(Source: Kim et al., 2012)

In experimental studies, T-type temperature sensors and ultrasound wave based

flow sensors were used to measure inlet temperature (Tinlet ), outlet temperature (Toutlet )
and oil volume flow rate (Q inlet ).

29
 There seems to be a general agreement between the analytical results and the
experimental results, and as a result, theoretical calculations are verified with
experimental studies.

2.5. Porous Media Approach in CFD Simulations

CFD model of power transformer requires a large number of mesh cells. Power
transformer is modelled with using porous medium approach to decrease solution time,
which is proportional to the number of mesh. Fdhila et al. (2011) focused on oil and air
flows and the heat transfer in fan-cooled transformer radiators with using ANSYS
Fluent 12.0 CFD simulation software. Symmetry boundary condition was used to model
of the radiator groups. Fans were modelled as a volume in this study. The effect of fan
position, fan size and oil flow rate were examined. The fan diameter varied between 0.8
m and 1.5 m. CFD model was simulated with using two, three and four fans.
Temperature distribution in the radiators with three fans is shown in Figure 29 below.
Air relative temperature distribution is showed at the left side and oil relative
temperature distribution is showed at the right side in the radiators.

Figure 29. Temperature Distribution in the Radiators.

(Source: Fdhila et al., 2011)

Four different configurations were studied to determine the temperature

differences between the inlet and the outlet of the radiators as shown in Figure 30. The
effects of the fan size and the locations on the temperature difference were also
analyzed. As a consequence, this research suggested that the porous medium approach
provided decrease in solution time.

30
 Figure 30. Temperature Difference (T_(oil,top) – T_(oil,bottom)) for Different Fan
Configurations.
(Source: Fdhila et al., 2011)

2.6. Complete and Slice Models

Gastelurrutia et al. (2009) studied on distribution transformers with corrugated

walls cooling ONAN (Oil Natural – Air Natural). Fluent V.6.3 software was used
implementing the CFD codes. Three geometries for ONAN cooling system of
distribution transformers were presented. First model (Trans-01) was handled 630 kVA.
Second model (Trans-02) was smaller than first model and the corrugated wall was
consisted more external fins. On the other hand, third model (Trans-03) was bigger and
had more and longer fins. Low voltage and high voltage coils and channels were
modelled with complete and slice models. Different flow domains and porous zone
were used as shown in Figure 31.

31
Figure 31. Developed Flow Domains for Second Transformer with Complete (left) and
Slice Model (right).
(Source: Gastelurrutia et al., 2009)

Temperature was measured for each model with different power losses as shown
in Table 3. The hottest surface was measured on the upper internal part of the fins (T9)
in Figure 32. On the other hand, the coldest point was measured at the base of the
transformer (T7) in Figure 32. In addition, temperature differences were evaluated
according to the experimental measurements. Temperature difference in the vertical
direction on the fins was calculated 14 K, and 2.5 K in the horizontal direction.

Table 3. Experimental Temperature Measurements.

(Source: Gastelurrutia et al., 2009)

32
 K-type thermocouples were used to measure the temperatures located on the
external surface of the transformer as seen in Figure 32. Also complete and slice models
are seen below. Radiation was neglected while the measuring temperatures.

Figure 32. Position of the Thermocouples During the Experimental Tests and Geometry
of the Complete and Slice Model.
(Source: Gastelurrutia et al., 2009)

Temperature differences between the complete and slice models for two
different transformers (Transf-01 and Transf-02) were presented by Gastelurrutia et al.
as shown in Table 4. Temperature difference between lid (T1) and the central fin (T11
and T12) was calculated as 1-2˚C. Also, 15-20˚C temperature difference was observed
at the base of the transformer (T7).

33
 Table 4. Comparison Between Complete and Slice Models.

(Source: Gastelurrutia et al., 2009)

The model of hollow fins was divided into 3 zones for each transformer model
as shown in Figure 33 below. Uniform temperature distribution was obtained at the first
zone, while temperature distribution was increasing in the second zone linearly. Then,
the sharp temperature drop was shown at the bottom part of the fins.

Figure 33. Oil Temperature Distribution in a Vertical Mid-plane of the Fins in Kelvin:
Transf-01 (left), Transf-02 (middle), Transf-03 (right).
(Source: Gastelurrutia et al., 2009)

Gastelurrutia et al. evaluated the influence of turbulence modelling.

Furthermore, grid independence was verified and the external thermal boundary
conditions were carried out. It was observed that the heat transfer coefficient varied in
the vertical direction.

34
2.7. Transformer Winding

In many research, thermal and hydrodynamic performance was studied in

transformer winding by researchers. The most important parameter is the hot spot
temperature in transformer design. Hot spot is the highest temperature occurred in the
winding and the primary reason for ageing of insulation materials. Ageing is a direct
function of temperature that occurs on transformers and IEC Standards defines the
aging rate of a transformer as a function of temperature. Winding is the heat source of
the transformer and that heat is dissipated by cooling equipments.
Radakovic and Sorgic (2010) were studied on the basics of detailed thermal-
hydraulic model for thermal design of oil power transformers. They were presented the
method for the calculation of temperatures inside the oil immersed power transformers,
based on detailed thermal-hydraulic model for natural, forced and directed oil flow.
Calculation safety factors can be reduced by accurate calculations. Therefore, cost of
production can be also reduced.
Skillen et al. (2011) were dealt with transformer losses and cooling modes for
the low-voltage winding geometry which has been considered. Also, the effect of inlet
mass flow rate on the solution has been investigated. It was found that the significant
local hot-spots were predicted via the use of CFD tools.
El-Wakil et al. (2006) performed a simulation of a step-down transformer
including core and windings by using Fluent Software. PISO (the pressure-implicit with
splitting of operators) method was used to solve equations. The inlet oil temperature of
winding was 338 K. They were focused on the cooling optimization of the power
transformer. Different geometries were simulated with six different flow rates of the
cooling oil at the entrance of winding as shown in Figure 34.

35
 Figure 34. The Isotherms at the Top of Transformer for Different Inlet Velocities.
(Source: El Wakil et al., 2006)

Heat transfer and fluid flow inside the winding oil cooling channels were studied
by using numerical method with six different cases. Finally, Case 1 was obtained as the
best geometry configuration, which has no insulations.

36
2.8. Transformer Cooling Fluids

Transformer oil is used for the purpose of cooling and insulating medium in oil-
immersed transformers. Generally, three types of transformer oil are used; mineral oil,
ester oil and silicone based oil.
Concern about contaminating effects of transformer oil on the environment is
increasing day by day. In this respects, researchers are focused on the alternative fluid
that will replace mineral oil in recent years. Fernández et al. (2012) compared main
properties of alternative fluids for power transformers. Mineral oils are widely used in
oil-immersed transformers although they are not environmentally friendly and they
contain high risk of fire. They classified the insulating fluids into four group; mineral
oils, high molecular weight hydrocarbons (HMWH), silicones and ester-based fluids. In
addition, miscibility of alternative fluids were presented as shown in Figure 35 below.

Figure 35. Miscibility of Alternative Fluids at Ambient Temperature.

(Source: Fernández et al., 2012)

Viscosity is known as the most important parameter for transformer cooling

systems. All alternative fluids that has presented by Fernández et al. (2012) are more
viscous than mineral oils. This is the critical point to determine the characteristics of
cooling systems. In addition, moisture is absorbed by alternative fluids more than
mineral oil. Moisture molecules are bounded on the winding insulation materials easily
while using traditional transformer oils. Additionally, vegetable based transformer oils
have tendency to absorb that moisture molecules contained within the insulation
material. As a result, moisture in the insulation material is diminished and strengthens
the aging withstand of insulation material. Therefore, using vegetable based transformer
oils enables higher overloading capacity of transformers.

37
 Rozga (2013) was presented a study on the properties of new environmentally
friendly transformer oils for power transformers in 1st Annual International
Interdisciplinary Conference. Basic properties of synthetic ester, natural ester and
mineral oil were summarized as shown in Table 5. It is obviously shown that natural
ester oil is more viscous than synthetic ester and mineral oil. Especially, at high
temperatures, viscosity of natural ester is 3 times higher than the mineral oil. Fire
resistance of transformer oils have significant role in safety regulations. Natural esters
have very high fire and flash points as shown in table below. Also, the environmental
impact of natural esters have drawn attention with 97% biodegradable value, while the
synthetic ester and mineral oil have 89% and 10%, respectively. Due to environmental
concerns, the future trend in transformer manufacturing is to increase the research on
biodegradable and fire-proof transformer oils. At the moment, natural ester based
transformer oils seems as the best option the industry have.

Table 5. Basic Properties of Synthetic Ester, Natural Ester and Mineral Oil.
Synthetic Natural
Physico-chemical properties Units Mineral
ester ester
Density at 20˚C kg/dm³ 0.97 0.92 0.88
Specific Heat at 20˚C J/kgK 1880 1848 1860
Thermal Conductivity at 20˚C W/mK 0.144 0.177 0.126
Kinematic Viscosity at 20˚C mm²/s 70 85 22
Kinematic Viscosity at 100˚C mm²/s 5.25 8.4 2.6
Pour Point ˚C -60 -21 -50
Fire Point ˚C 316 360 170
Flash Point ˚C 150
Fire Hazard Classification to
- K3 K2 0
IEC 61100 / IEC 61039
Biodegradability % 89 97 10
(Source: Rozga, 2013)

38
2.9. Overview of Thesis According to Literature

Determination of hot spot temperature in transformers is a challenge that has

been studied by various researchers. The basic problems encountered in these studies
are summarized below.
 The problem of determination of eddy currents in the windings requires 3D
electromagnetic simulations as the transformers do not possess any
symmetries. This is the electromagnetic part contribute to the phenomenon.
This requires solution of Maxwell’s equations in steady state and transient
conditions.
 Determination of eddy currents is not enough to provide final solution to the
identification of hot spot temperature. The problem is also dominated by
thermal and fluid dynamic laws. This part of the problem brings great
challenges as the transformer winding geometry is extremely complicated.
Many parameters for instance winding type, spacer dimensions, disc
dimensions, layer dimensions and cooling duct dimensions etc. further
complicates the problem.
 Coupled electromagnetic and CFD solvers are readily available in the
market today. However, in order to utilize these solvers, great
simplifications in transformer CFD models are required. The basic
simplification can be attained by finding ways to subtract the cooling
systems from the geometry. This will enable the saving from both the mesh
and the computation times.
 Geometry of the radiators is recently subject to changes as a result of new
transformer oils introduced into the market. To provide new and deeper
understanding related to these products requires new studies.

The abovementioned problems states that in order to provide complete coupled

simulation of a power transformer requires simplifying cooling systems. Moreover, new
radiator geometries and transformer oil performances has to be determined. In literature,
the CFD models are generated by assuming constant flux densities where the fringing
effects of electromagnetic flux lines are omitted. These studies also do not provide
coupled simulations. Performance of new transformer oils has not been studied in detail.

39
 Last but not least, modelling of transformer radiators with porous medium has
not been encountered in literature in details. In these studies constant oil density was
assumed and oil type and radiator geometry was not provided in detail. More
importantly viscous and inertial resistance coefficients, the most important parameters
in porous media approach, were not given apparently.
The work presented in this thesis provides an approximation to the performance
verification and simplification of cooling systems to analyze a transformer. This thesis
aims at providing guidance for complete coupled electromagnetic and CFD simulations.
The work presented here provides the first and the fundamental step of complete
coupled simulations. Hence, thermal and hydrodynamic characteristics of the
transformer cooling systems will be provided by this study. Then, all the cooling
systems of the transformer can be modelled as a porous medium in simulations to
decrease the element number and computation time. CFD simulations and
electromagnetic simulations should be generated to provide coupled solutions. All the
simulation results should be also verified by experimental studies.

40
 CHAPTER 3

TRANSFORMER OILS

Transformer oil is used in oil-filled transformers to provide electrical insulation

and heat transfer, thus it is called as insulating oil. Transformer oils can be grouped into
three main types in this thesis as shown in Figure 36: mineral, ester and silicone
transformer oils. Ester based transformer oil is divided into two groups: natural ester
and synthetic ester transformer oils. The most common type of transformer oil is
mineral oil. Mineral oils negatively impact environment. Most importantly, it has poorly
biodegradable (only 10%) as mentioned by Rozga (2013). Because of these reasons, it
was decided that natural ester oil behavior in transformer should be investigated to
collect more information for further developments. Also, dielectric properties of natural
ester transformer oil are similar or better than mineral oil.

Figure 36. Transformer Oil Types.

Transformer oil analysis provides information regarding the transformer

condition. It can be thought as a fingerprint of transformer. The properties of unused oil
can be compared with the oil sample of transformer oil in service. A famous method to
detect incipient falls is called dissolved gas analysis (DGA). Degradation of oil at
cellulose insulation material generates certain gases. These gases are hydrogen, oxygen,
carbon monoxide, carbon dioxide, ethylene and ethane etc. can be used to determine the
type of fault and the cause of fault. A famous duval triangle method compares the
volume of the generated gases and gives a description of the type of fault, fault
temperature and the fault’s origin as shown in Figure 37.

41
 Figure 37. Duval Triangle.
(Source: Arora, 2013)

DGA also gives the trend of gases formation in the transformer oil leading to
identify the type of fault. For instance, increase in percentage value of CH4 while
percentage of the other gases remain the same means that there is a high level of partial
discharge (PD) on insulation materials. Another example can be demonstrated as C2H2
has 3% and C2H4 has 30% volume content. According to duval triangle mentioned by
Arora (2013), the intersection of the lines lies in the area of T2 colored by orange. This
means that the fault has a thermal fault with temperature 300 – 700 ˚C as shown in
Table 6.

Table 6. Duval's Triangle Fault Zone.

(Source: Arora, 2013)

42
 On line DGA analyze systems are integrated into modern transformers. These
systems enable early detection of the faults prior to catastrophic failures. By monitoring
the transformers in service enables users to identify possible fault types. This ensures
that the manufacturer of the transformer can be responded to the fault before final
failure. This explains the importance of transformer oil because transformer oil tests can
be considered as finger print of transformer. It gives the health of transformer during its
service life.
Furthermore, besides fault detection transformer oil serves as cooling and
dielectric insulating medium. From heat transfer point of view, transformer oil enables
to cool down the both the windings and the core through natural circulation. In order to
increase the capacity of transformer further, pumps and fans can be added into cooling
system. Cooling the transformer efficiently provides higher MVA ratings. 25 MVA
natural circulating transformer can be increased up to 30 MVA just by adding cooling
fans below the radiators. Therefore, efficient cooling increases transformer power rating
as well as enables optimization in power transformer design.
In this study, flow and heat transfer of a transformer radiator filled with mineral
oil, natural ester oil, synthetic ester oil and silicone oil were investigated. Density,
specific heat, thermal conductivity and viscosity properties for the most common
transformer oil types used in simulations were presented in this chapter. Also, the
equations were obtained depending on the temperature for each property of each oil
types to use it in Fluent while defining the material properties. All the graphs had drawn
by using MathCAD.
Density, specific heat and thermal conductivity properties of the transformer oil
are constructed using curve fitting techniques from the manufacturer’s test data. These
data are glugged into Fluent software by using piecewise-linear profile. Physical
properties are defined at the 233 K and 383 K while using piecewise-linear profile. On
the other hand, piecewise-polynomial profile is used to define viscosity. ANSYS Fluent
software program uses the dynamic viscosity property. Therefore, dynamic viscosity
graphs and equations were discussed in this study.

43
3.1. Density

As shown in Figure 38, density of the transformer oils decreases linearly with
increase in temperature. Mineral transformer oil has the lowest density value, whereas
the synthetic ester oil has the highest.

Figure 38. Density Graph versus Temperature.

3.2. Specific Heat

Specific heat is a kind of physical property of matter that defines the required
heat to change a unit mass of a matter by one degree Celsius. As shown in Figure 39,
specific heat values slightly increase with temperature. Specific heat versus temperature
graph shows that the mineral oil has the highest specific heat values.

44
 Figure 39. Specific Heat versus Temperature.

3.3. Thermal Conductivity

Thermal conductivity is a physical property of matter that related with

conduction heat transfer. As shown below in Figure 40, thermal conductivity property
of transformer oils linearly decreases with temperature. Silicone oil thermal
conductivity value is constant, 0.151 W/mK. As a result, thermal conductivity versus
temperature graph shows that natural ester oil and synthetic oil both have the same
thermal conductivity values, whereas mineral oil has the lowest.

45
 Figure 40. Thermal Conductivity versus Temperature.

3.4. Dynamic Viscosity

Although viscosity graphs were obtained exponential. Then the equations were
defined in MathCAD using Levenberg – Marquardt non-linear method as shown in
Equation 3.1.

C 1 2 1 3 1 4
i+1 +C
[Ci +T+273.15 i+2 ∙(T+273.15) +Ci+3 ∙(T+273.15) +Ci+4 ∙(T+273.15) ]
μoil (T) = e (3.1)

where ci shows the coefficients that obtained from Levenberg – Marquardt non-linear
method, T is the reference temperature and μoil is the dynamic viscosity of oil.
From dynamic viscosity versus temperature graph (Figure 41) it is clear that
viscosity decreases with increasing temperature. Viscosity has the highest value at the
low temperatures. There is a significant decrease in dynamic viscosity values between -
20˚C and 0˚C.

46
 Figure 41. Dynamic Viscosity versus Temperature.

Although the viscosity equation was obtained, it is required that some special
codes should be written to define exponential function in Fluent. Special codes that are
known as user-defined functions (UDFs) allow the Fluent users to define special
boundary conditions, material properties and source terms. However, it is required to
have very well programming skills. Also, using UDFs could be complicated in some
cases. Because of these reasons, viscosity graph was divided into two parts to have two
different polynomial graphs. Equations of the each of viscosity graph were created by
using curve fitting techniques in MathCAD. Then the coefficients were used in Fluent
to define material property for each range by using piecewise-polynomial profile. First
range was defined between 233 K and 303 K. Then, the second range was constructed
between 303 K and 383 K. Detailed information about material properties was given in
next chapter.

47
 CHAPTER 4

NUMERICAL ANALYSIS

ANSYS Fluent is CFD simulation software to model fluid flow and heat
transfer. This study illustrates the fluid flow and heat transfer problem in a section of
transformer radiator. CFD simulation consists of four main parts as modelling, meshing,
solution and results. ANSYS Fluent CFD solver uses finite volume method. Modelling
and meshing works are specified as pre-processing and solution and results are defined
as post-processing.
In this study, a section of a radiator was investigated. Methodologically,
symmetry condition was used and quarter of the flow volume was created to investigate
the fluid flow and heat transfer in radiator section.
Radiator geometry was modelled in Catia V5 software, and then imported into
ANSYS Design Modeler module (Figure 42) to obtain a flow volume. Appropriate
mesh elements were selected for each part.

Figure 42. Imported Radiator Geometry in Design Modeler.

48
 In setup module, temperature dependent material properties of transformer oil
were defined. Boundary conditions and solution methods were selected. Results were
obtained and visualized.
Computational fluid dynamics (CFD) calculations are based on three main
equations; continuity, momentum and energy equations. The form of the continuity
equation for laminar flow is shown in Equation (4.1). Steady-state flow regime was
studied in this study. Fluid properties of the oil circulating through the radiator oil ducts
do not change over time at any point. Partial derivatives of all quantities with respect to
time are zero in steady-state flow regime. Radiator and directions are seen in Figure 43.

∂(ρu) ∂(ρv) ∂(ρw)

+ + =0 (4.1)
∂x ∂y ∂z

Figure 43. Radiator Model and Directions.

x, y, z-momentum equations are seen below in Equation (4.2)-(4.4), respectively.

x-momentum
∂ ∂ ∂ ∂P ∂2 u ∂2 u ∂2 u
(ρuu) + (ρvu) + (ρwu) = − + μ [ 2 + 2 + 2] (4.2)
∂x ∂y ∂z ∂x ∂x ∂y ∂z

49
y-momentum
∂ ∂ ∂ ∂P ∂2 v ∂2 v ∂2 v
(ρuv) + (ρvv) + (ρwv) = − + μ [ 2 + 2 + 2] (4.3)
∂x ∂y ∂z ∂y ∂x ∂y ∂z

z-momentum

∂ ∂ ∂ ∂P ∂2 w ∂2 w ∂2 w
(ρuw) + (ρvw) + (ρww) = − − ρg + μ [ 2 + 2 + 2 ] (4.4)
∂x ∂y ∂z ∂x ∂x ∂y ∂z

The energy equation of studied flow is as shown in Equation 4.5.

⃗ ∙ ∇)T] = k∇2 T + (τ̿ ∙ ∇)V

ρcp [(V ⃗ (4.5)

4.1. Problem Statement

Simulation of power transformer requires a simplification of cooling systems

because of mesh number and computation time. Porous model offers great amount of
simplifications to the complete cooling system analysis. In order to quantify quarter of a
single section of a radiator requires 2014258 element numbers. All the cooling system
can be modelled as a single body by using porous medium approach. This enables
extreme savings in terms of element number by this way. Therefore, one aim of this
study is to enable complete transformer coupled electromagnetic and CFD model.
In order to achieve this modelling technique, single section of a transformer
radiator is modelled and the pressure drop value is determined over a wide range of inlet
velocity values at yearly average ambient temperature. Yearly average temperature is
chosen because the transformer manufacturers design their system based on this
temperature value.
Another important point of this study is to cover the thermal behavior of
different and relatively new transformer oils. Their thermal properties are generated
with MathCAD software to model these oil types in order to input these material
properties into Fluent CFD Solver.
There are different transformer radiator manufacturers in the market. Every
radiator manufacturer provides different solutions with different radiator geometries

50
before starting modelling phase one section radiator samples were provided from every
manufacturer. Transformer radiator and oil ducts shapes are shown in Figure 44 below.

Figure 44. Transformer Radiator.

Radiators were cut with water jets to study and determine the oil duct
geometries. Moreover, their cooling duct CAD geometries were obtained in order to
compare actual geometry with the CAD geometries. No conflict between the CAD
geometries and the actual geometries were found. Based on these studies actual
modelling of one section transformer radiator was generated with ANSYS Design
Modeler.

4.2. Modelling

Design Modeler is parametric geometry software that provides modelling

features for ANSYS analysis. Transformer radiator geometry was modelled in CATIA
V5 software as shown in Figure 45. The modelled radiator has the 3200 mm length and
520 mm wide. Geometry was imported into Design Modeler and modified to have a
quarter of flow volume.

51
 Figure 45. Radiator Model in Catia V5.

Radiator model was divided into parts according to grid element types. Inlet and
outlet zones were simplified in this study as shown in Figure 46.

Figure 46. Simplified Radiator Model.

52
4.3. Meshing

Mesh (grid) module is the most important part of CFD simulations. In order to
calculate the fluid flow and heat transfer of all body, domains are splitted into small
parts, which are called mesh elements. ANSYS Fluent solver uses finite volume method
to govern fluid flow and heat transfer equations. Boundary conditions should be defined
before meshing operations.
In this study, inlet and outlet boundaries were defined as inlet-velocity and
outlet-pressure, respectively. Symmetry boundary conditions were used by two
symmetry planes to have a quarter flow volume.
Physical preferences were set to CFD and Fluent. Geometry was meshed by
using different types of grid. Radiator oil ducts, inlet and outlet parts were meshed by
using sweep mesh method with quad elements (Figure 47 and Figure 48). Quad meshes
provide high quality solutions in sweepable bodies. Sweep mesh is a meshing method
that has a high accuracy in CFD calculations. Inflation is defined to create thin elements
near no-slip walls. Face zones were defined in boundary conditions. On the other hand,
material properties influence the cells.

Figure 47. Mesh of the Oil Ducts.

53
 Figure 48. Inlet/Outlet Mesh Type and Inflation.

Inflation at the no-slip wall condition and quad mesh elements are shown in
Figure 49.

Figure 49. Section of Oil Duct.

Accuracy of the simulation is directly depends on the mesh quality. Skewness is

an important parameter that shows the mesh quality. 2014258 mesh elements were used
with 0.897 skewness as shown in Figure 50. Value of the skewness can be accepted up
to 0.95.

54
 Figure 50. Mesh of the Radiator Model.

4.4. General Settings

Model settings, material properties, cell zone conditions, boundary conditions

and reference values are the setup procedures. Solution methods, solution controls,
monitors, initialization and calculations are performed in solution.
Pressure-based solver was used which is applicable for incompressible flow.
Inlet and outlet area are equal and inlet velocity was defined constant. Therefore, mass
flow rate at inlet and outlet are equal in each case. The velocity of transformer oil is
independent of time at any point. Thus, steady-state flow regime was assumed in this
study. Acceleration due to gravity was defined -9.81 m/s2 along (-) z axis.

4.5. Viscous Model

The maximum thickness of oil duct is 9.1 mm as shown in Figure 51 and the
height of the radiator is 3200 mm. Thickness of the oil duct is very small compared to
the length of radiator. On the other hand, transformer oils have the high viscosity.
Because of these conditions, flow is assumed as laminar.

55
 Figure 51. Cross-section of Radiator Oil Duct (top) and ¼ Radiator Oil Duct (bottom).

4.6. Material Properties

Temperature dependent material properties of transformer oil (density, specific

heat, conductivity and viscosity) were defined in materials option.
Density of the transformer oils decrease linearly with increase in temperature,
whereas the specific thermal capacity increases as shown in Chapter 3. Thermal
conductivity also decreases with increasing temperature. Piecewise-linear method was
used to define density, specific thermal capacity and thermal conductivity of
transformer oils.
Viscosity graph is obtained exponential and its equation is defined in MathCAD
using Levenberg – Marquardt non-linear method. However, special codes required to
define this kind of complicated equations in Fluent. ANSYS Fluent allows defining
undefined property with using user-defined functions (UDF). User needs programming
skills to define UDF code.
In order to use UDF codes, piecewise-polynomial method in Fluent is used to
define viscosity property. Viscosity graphs are divided into two parts and the equations
are obtained for each part. Divided viscosity graphs are shown below. Fourth degree
polynomial equation was obtained for natural ester transformer oil viscosity as shown in
Equation (4.6) and Equation (4.7).

μ(T) = 0.000000000081T 4 − 0.00000014036T 3 + 0.000091453076T 2

−0.0265799317713T + 2.9142493540421 (4.6)

56
 μ(T) = 0.0000000245382T 4 − 0.0000312372486T 3 + 0.0149171842259T 2
−3.1680682038423T + 252.5684734582901 (4.7)

Obtained polynomial equations and dynamic viscosity values for natural ester
transformer oil are seen in Figure 52 and Figure 53. First interval of graph is defined
from 253 K to 333 K, and second interval is from 333 K to 473 K. Red points indicates
the measured viscosity values that are given and blue lines shows the equation results.

Figure 52. Dynamic Viscosity versus Temperature Graph of the First Interval for
Natural Ester Transformer Oil.

57
 Figure 53. Dynamic Viscosity versus Temperature Graph of the Second Interval for
Natural Ester Transformer Oil.

Moreover, material properties of radiator steel were selected constant. Density,

specific heat and thermal conductivity of radiator steel were 7850 kg/m3, 490 J/kg K,
46 W/m K, respectively.

4.7. Boundary Conditions

Inlet velocity, outlet pressure and symmetry boundary conditions are defined as
shown in Figure 54.

Figure 54. Boundary Conditions.

58
 Inlet velocity and temperature conditions were defined based on the following
principle. In transformers, various numbers of radiators are placed side by side.
According to the measurements that has been carried out in Balikesir Electromechanical
Industrial Plants Corporation Test Laboratories (Figure 55) the following velocity
profiles for ONAN cooling mode was measured. The average oil flow velocity of 20
radiators placed side by side was found to range between 0.01 m/s to 0.035 m/s based
on the loading and ambient temperature conditions. Ambient temperature at these
measurements was 20˚C. This temperature value corresponds to 80˚C top oil
temperature.

Figure 55. BEST Co. Test Laboratory.

In the simulations, the values for velocity and ambient temperature are taken
into account based on these considerations. Moreover, in IEC 60076-2 Standard yearly
average ambient temperature for transformer design is also given as 20˚C. Therefore,
velocity profile at this temperature correspond both the standards and the
measurements.
The aim of the study is to determine pressure drop, outlet temperature and heat
flux of a transformer radiator based on different velocity values. For instance,
overloading of transformer generates more heat and this might change the velocity

59
value at the inlet. In order to cover this difference and to determine the radiator pressure
drop values for a wider range of velocities simulations were extended to cover 0.05 m/s.
0.01 to 0.05 m/s velocities and 353 K temperature were defined in velocity inlet
boundary condition.
Outlet pressure boundary condition was defined to investigate the flow and heat
transfer at outlet. Then, other surfaces were defined as wall automatically by Fluent.
Heat transfer coefficient, free stream temperature, wall thickness and material settings
were identified in wall settings. Thermal heat transfer coefficient was assumed as
6 W/m2K in this study. Ambient temperature was 298 K.

4.8. Solution

Steady-state problem was investigated in this study, SIMPLE (Semi-Implicit

Method for Pressure-Linked Equations) algorithm was selected in solution methods.
Velocity and pressure gradients were calculated to solve momentum equation. Then
pressure field was calculated by using under-relaxation factor for pressure.
Firstly, only the flow was investigated by closing the energy equations under
Solution Controls menu to simplify iterations. After converged flow simulation, energy
equation was opened.
Iteration results were observed by adding required variable into Surface
Monitors. In this study, inlet pressure, outlet pressure, outlet temperature, inlet density,
outlet density and heat flux at wall were defined in Surface Monitors. Variables were
reported by using area-weighted average report type.
Before starting calculations, solution is initialized using standard initialization.
Iterations continue and the monitors were followed until the solution converges. As seen
in Figure 56, solution was converged after nearly 1200 iterations. Convergence criteria
were 10-4 for the continuity and momentum equations and 10-9 for energy equation.

60
Figure 56. Scaled Residuals.

61
 CHAPTER 5

RESULTS AND DISCUSSION

ANSYS Fluent CFD simulations use finite volume method to solve continuity,
momentum and energy equations. CFD flow volume was created to simulate
transformer oil flow in radiator.
Fluid flows are classified into two groups as compressible and incompressible
flows. Incompressibility is a fluid dynamics property that describes the flow. In this
study, incompressible flow was studied. Incompressible flows have small changes in
density. Liquids are generally classified as incompressible fluids. Velocity inlet is a
boundary condition in ANSYS Fluent CFD software. Velocity inlet boundary condition
was used at inlet with constant velocity. Different velocity values were applied at this
boundary. Also, inlet temperature was specified as 353 K for all simulations.
Convection heat transfer thermal condition was defined at wall. Then, heat transfer
coefficient was assumed as 6 W/m2K. Thermal heat transfer coefficient characteristics
on radiator inner and outer surfaces should be studied in the future works.
Flow was assumed as uniform at inlet. Additionally, symmetry boundary
conditions were used to reduce computational domain. On the other hand, radiator
boundary condition was not used for this problem. Because pressure drop and heat
transfer coefficient characteristics does not known. One of the most important purpose
of this study is to determine the characteristics of radiators that can be used in future
works.
The most important limitation of the study is the mesh procedure. Researchers
prefer the Catia or Inventor to design geometry instead of ANSYS Design Modeler
module. Then, geometry should be updated in Design Modeler and also in Mesh module
to avoid creating sharp edges. High performance computers are required to perform the
calculations in 3D CFD simulations.

62
5.1. CFD Results

Results module is the CFD post-processing module in ANSYS software to

visualize the flow and heat transfer in the simulation model. In this section, natural ester
transformer oil will be investigated with visualizing temperature distribution, velocity
vectors, heat flux and velocity streamlines of one quarter of transformer radiator
section. According to the results, geometry of transformer radiators can be improved.
Especially, at the corners heat transfer rate is lowest. Sharp corners can be rounded.
Moreover, Reynolds number, Nusselt number and Prandtl number were calculated
theoretically.

5.1.1. Temperature Distribution

Inlet temperature was set to 353 K at inlet boundary for all numerical analysis.
The detailed temperature distributions around inlet and outlet zones are seen in
Figure 57 and Figure 58. According to legend, red zone is around 353 K and blue zone
is around 332 K. Also, temperature distribution along the simulated quarter model is
seen in Figure 59.

Figure 57. Temperature Distribution at the Inlet Zone with 0.05 m/s Inlet Velocity for
Natural Ester Oil.

63
Figure 58. Temperature Distribution at the Outlet Zone with 0.05 m/s Inlet Velocity for
Natural Ester Oil.

Figure 59. Temperature Distribution Along Simulated Quarter Radiator.

64
5.1.2. Velocity Vectors

Additionally, velocity profile of natural ester transformer oil was investigated

that has 0.05 m/s at inlet. Velocity vectors were defined as normal to boundary to avoid
the reversed flows at outlet boundary as seen in Figure 60 and Figure 61.

Figure 60. Velocity vectors at the inlet with constant 0.05 m/s inlet velocity.

Figure 61. Velocity Vectors at the Outlet with Constant 0.05 m/s Inlet Velocity.

65
5.1.3. Velocity Streamlines

Streamlines were sketched for this flow to create velocity profile. Although the
inlet velocity was 0.05 m/s, it is seen that velocity is decreased to nearly zero inside the
oil ducts as shown in Figure 62. This velocity decreasing because of the small thickness
that oil ducts have. Especially, at low temperatures natural transformer oil cannot flow
inside the radiator with this traditional method. Velocity streamlines at inlet and outlet
zones were shown in Figure 63 and Figure 64, respectively.

Figure 62. Streamlines at Inlet of Natural Transformer Oil with 0.05 m/s.

66
Figure 63. Velocity Streamline with 0.05 m/s Inlet Velocity at Inlet Boundary Condition
for Natural Ester Transformer Oil.

Figure 64. Velocity Streamline with 0.05 m/s Inlet Velocity at Outlet Boundary
Condition for Natural Ester Transformer Oil.

67
 The flow furthest from the inlet region in y direction can be accelerated by
introducing an appropriate gap in the entry zone as shown in Figure 65. The grey
shaded zone in Figure 65 will decrease the cross-sectional area and hence the velocity
of streamlines along y direction will increase. This will increase the efficiency of the
transformer radiators and suggested as an improvement based on the outcomes of this
study.

Figure 65. Recommended radiator entry zone.

68
5.1.4. Heat Flux

Heat flux at inlet zone and outlet zone were visualized by using 0.05 m/s
velocity for natural ester transformer oil as shown in Figure 66 and Figure 67,
respectively. Average heat flux is 284.9 W/m2. Heat flux on radiator surface was shown
in Figure 68.

Figure 66. Heat Flux at the Inlet Zone with 0.05 m/s Inlet Velocity for Natural Ester
Transformer Oil.

69
Figure 67. Heat Flux at the Inlet Zone with 0.05 m/s Inlet Velocity for Natural Ester
Transformer Oil.

Figure 68. Heat Flux with 0.05 m/s Inlet Velocity for Natural Ester Oil.

70
5.1.5. Comparison of Oils

As shown in Table 7 and Table 8, the outlet temperature of radiators is the same
which were simulated with natural ester and mineral transformer oil, respectively.
Besides that, heat flux of each simulated radiator is the same. On the other hand,
pressure difference between inlet and outlet of the radiator simulated with natural ester
transformer oil is more than mineral oil. Silicone transformer oil has the most pressure
drop as shown in Table 9. Although mineral oil has the lowest pressure drop and the
highest heat flux, they should be replaced by natural ester transformer oils according to
environmental concerns. Pressure difference increases by increasing the inlet velocity
and heat flux also increases. Besides, temperature difference between inlet and outlet is
decreasing by increasing velocity. Transformer oil temperature should be kept in
working temperature limits. Because of that, minimum temperature difference and
maximum heat flux is preferred in transformer cooling. The most heat flux with
minimum temperature difference was observed in natural ester transformer oil as shown
in Table 7. Additionally, temperature difference between the inlet and outlet of
transformer winding has to be kept as low as possible according to Kim et al. (1987).
The difference between inlet and outlet temperature of the transformer winding can be
kept minimum, if mass flow rate of radiator can be increased. This can be achieved by
increasing value of pressure drop along a closed circulating loop.

Table 7. Natural Ester Transformer Oil Results.

Temperature
Pressure Outlet
Velocity Heat flux Difference
Difference Temperature
(inlet-outlet)
[m/s] [Pa] [W/m²] [K] [K]
0.02 94.9 258.7 337.1 15.9
0.03 127.5 272.6 341.7 11.3
0.04 159.3 280.1 344.1 8.9
0.05 190.8 284.9 345.7 7.4

71
 Table 8. Mineral Transformer Oil Results.
Temperature
Pressure Outlet
Velocity Heat flux Difference
Difference Temperature
(inlet-outlet)
[m/s] [Pa] [W/m²] [K] [K]
0.02 37.2 258.4 337.0 16.1
0.03 47.0 272.0 341.6 11.4
0.04 56.0 279.2 344.1 8.9
0.05 64.6 283.7 345.7 7.3

Table 9. Silicone Transformer Oil Results.

Temperature
Pressure Outlet
Velocity Heat flux Difference
Difference Temperature
(inlet-outlet)
[m/s] [Pa] [W/m²] [K] [K]
0.02 151.8 240.6 331.7 21.3
0.03 206.6 259.2 337.6 15.4
0.04 260.0 269.3 340.8 12.2
0.05 312.9 275.8 343.0 10.1

Table 10. Synthetic Transformer Oil Results.

Temperature
Pressure Outlet
Velocity Heat flux Difference
Difference Temperature
(inlet-outlet)
[m/s] [Pa] [W/m²] [K] [K]
0.02 80.4 253.5 335.7 17.3
0.03 103.9 268.4 340.6 12.4
0.04 126.5 276.5 343.3 9.7
0.05 148.7 281.5 345.0 8.0

The flow of an incompressible Newtonian fluid was studied. The thickness of

radiator is very small compared to length. Therefore, fluid flow was defined as laminar

72
as in the case of a real transformer radiator flow. Additionally, thermal properties of oil
and low velocity affect the flow regime.
Velocity versus pressure difference graph was drawn for each oil types as shown
in Figure 69. It is seen that pressure difference is approximately the same for natural
ester and synthetic oil. Mineral oil has the minimum pressure difference and silicone oil
has the highest.
As shown in Figure 70, natural ester oil has the maximum heat transfer rate and
it is nearly same as mineral oil. The lowest heat transfer rate is seen for silicone oil.
Also, heat transfer increases with velocity of transformer oil.

Figure 69. Velocity versus Pressure Drop Graph.

73
 Figure 70. Velocity versus Heat Flux Graph.

5.1.6. Dimensionless Numbers

This section includes the dimensionless number of flow in radiator with natural
ester transformer oil. As mentioned last section, natural ester transformer oil has the
best heat flux. Natural ester transformer oil was investigated in this section.
Reynolds numbers of natural ester transformer oil flow were calculated for each
case as shown in Table 11. Reynolds number is a dimensionless number in fluid
mechanics that defines the flow regime. Ratio of inertial forces to viscous forces gives
the Reynold number as shown in Equation (5.1), where ρ is the average density of
transformer oil, v is the average velocity, Dh is the hydraulic diameter and μ is the
average dynamic viscosity. It can be seen from the Reynolds numbers obtained from
Equation (5.1) that the major forces that affect the flow are the viscous forces.

inertial forces ρvDh

Re = = (5.1)
viscous forces μ

74
 Dynamic viscosity is defined by molecular viscosity in Fluent software. Average
velocity through oil ducts were calculated by using Equation (5.2) that is known as
continuity equation, where, A1 and A2 are the cross-sectional areas of inlet and oil
ducts, also v1 and v2 are the velocity of transformer oil at inlet and oil ducts,
respectively.

A1 v1 = A2 v2 (5.2)

Cross-sectional area at inlet and oil ducts are 282.96 mm2 and 727.29 mm2 as shown in
Figure 71 and Figure 72, respectively inlet and oil ducts are shown in green in the
figures.

Figure 71. Cross-sectional Area at Inlet.

75
 Figure 72. Cross-section Area of Oil Ducts.

Hydraulic diameter is used for non-circular ducts and it is calculated by using

Equation (5.3), where, A is the cross-sectional area of oil ducts and P is the wetted
perimeter of the cross-section.

4A
Dh = (5.3)
P

Cross-section area of oil ducts and wetted perimeter of the cross-section are shown with
green color in Figure 72 and Figure 73, respectively. For this case, cross-sectional area
of oil ducts is 727.3 mm2 and wetted perimeter of the cross-section is 429.5 mm. Then,
hydraulic diameter was calculated as shown below by using Equation (5.3).

4(727.3 mm2 )
Dh =
429.5 mm

Dh = 6.77 mm.

76
 Figure 73. Wetted Perimeter of Oil Duct Cross-section Area.

Calculated Reynolds numbers were very small as shown in Table 11 below.

Accordingly, viscous forces directly affect the heat energy dissipation in this study.

Table 11. Reynolds Number of Natural Ester Transformer Oil Flow.

Velocity Average Velocity Average Average Reynolds
through Oil Ducts Density Viscosity Number
[m/s] [m/s] [kg/m3] [kg/ms] [-]
0.02 0.0078 885.0303 0.0134 3.479
0.03 0.0117 883.4450 0.0127 5.497
0.04 0.0156 882.6011 0.0123 7.560
0.05 0.0195 882.0687 0.0121 9.601

Moreover, calculated Nusselt number and Prandtl number are seen in Table 12
and Table 13, respectively. Nusselt number is a dimensionless number that is the ratio
of convective heat transfer to conductive heat transfer as shown in Equation (5.4),
where h is the heat transfer coefficient at the film temperature, L is the characteristic
length and k is the thermal conductivity of the transformer oil at film temperature. In
this study, characteristic length, Dh , was calculated as 6.77 mm. Therefore, theoretical
Nusselt number was calculated by using Equation (5.5) as shown in Table 12.

77
 hL
Nu = (5.4)
k

hDh
Nu = (5.5)
k

Reference values were adjusted in Fluent software so that temperature was given
as film temperature and length was given as characteristic length. Film temperature was
calculated by using Equation (5.6) to obtain the physical properties. All the related
physical properties (density, specific heat, thermal conductivity and viscosity) were
evaluated at film temperature as seen in Table 12.

Twall + Tfluid
Tfilm = (5.6)
2

Table 12. Theoretical Nusselt Number for Natural Ester Transformer Oil Flow.
Inlet Film Thermal Heat Transfer Theoretical Nusselt
Velocity Temperature Conductivity Coefficient Number
[m/s] [K] [W/m.K] [W/m2.K] [-]
0.02 342.2816 0.1517 111.36 4.97
0.03 344.6465 0.1513 112.77 5.05
0.04 345.9115 0.1510 114.20 5.12
0.05 346.7161 0.1508 115.72 5.20

Furthermore, theoretical Prandtl number was calculated as seen in Table 13.

Prandtl number is the ratio of momentum diffusivity to thermal diffusivity as shown in
Equation (5.7), where μ is the dynamic viscosity, cp is the specific heat and k is the
thermal conductivity of oil.

μcp
Pr = (5.7)
k

Prandtl number is the main parameter to define relationship between the velocity
boundary layer and temperature boundary layer. As shown in calculated Prandtl

78
numbers in Table 13, viscous diffusion is dominant for transformer oils. Generally,
transformer oils have high Prandtl number. Therefore, it can be concluded that the
thermal boundary layer of transformer oils is located at the viscous region close to wall.

Table 13. Theoretical Prandtl Number for Natural Ester Transformer Oil Flow.
Inlet Film Thermal Dynamic Specific Theoretical
Velocity Temperature Conductivity Viscosity Heat Prandtl Number
[m/s] [K] [W/m.K] [W/m2.K] [J/kg.K] [-]
0.02 342.2816 0.1517 0.01332 2348.805 206.38
0.03 344.6465 0.1513 0.00127 2354.176 201.25
0.04 345.9115 0.1510 0.01259 2359.159 196.59
0.05 346.7161 0.1508 0.01237 2362.368 193.67

Using the calculated data, a graph between the Nusselt number and Reynolds
number was plotted for different velocities. Influence of Reynolds number on Nusselt
number was shown in Figure 74. Equation (5.8) shows that the Nusselt number
increases linearly as the Reynolds number increases. In Figure 74, blue points indicate
the theoretical Nusselt number and the red line obtained by using Equation (5.8).

Nu = 0.0372 Re + 4.8419 (5.8)

79
 Figure 74. Nusselt Number versus Reynolds Number Graph.

Calculated Nusselt number, Reynolds number and Prandtl number are shown in
Table 14 for different velocities.

Table 14. Calculated Dimensionless Numbers at Different Velocities.

Inlet Velocity Nusselt Number Reynolds Number Prandtl Number
[m/s] [-] [-] [-]
0.02 4.97 3.479 206.38
0.03 5.05 5.497 201.25
0.04 5.12 7.560 196.59
0.05 5.20 9.601 193.67

The Nusselt number depends on Reynolds number and Prandtl number as shown
in Equation (5.9), where a, b and c are constants. Reynolds number is a function of
density, velocity, hydraulic diameter and dynamic viscosity as Re=Re(ρ, v, Dh , μ).
Prandtl number is function of dynamic viscosity, specific heat and thermal conductivity
as Pr=Pr(μ, cp , k).

80
 Nu = C Rea Pr b (5.9)

Least-squares method was used to determine the constants a, b and c. Equation (5.9)
was separated into two parts to use power law as seen in Equation (5.10) and Equation
(5.11), where C = √C1 C2 .

Nu = C1 Rea (5.10)

Nu = C2 Pr 𝑏 (5.11)

Power law was applied to find correlation for Equation (5.10) in MathCAD. The
logarithm was taken for both sides of the Equation (5.12) as shown below to use linear
regression. Then, u, v and w were defined as u=ln(Nu), v=ln(C1 ) and w=ln(Re). Thus,
nonlinear equation was defined as a linear equation as shown in Equation (5.13). The
matrix form equation was constituted to find constants a and C1 as shown in
Equation (5.14). Consequently, a and C1 were obtained as 0.021785 and 4.7,
respectively. These steps were applied for the Equation (5.11). Hence, b and C2 were
obtained as 0.68956 and 196.4 respectively.

ln(Nu) = ln(C1 ) + a ln(Re) (5.12)

u = v + aw (5.13)

n ∑ ln(Re) ∑ ln(Nu)
ln(C1 )
[ ][ ]= (5.14)
2 𝑎 ∑ ln(Re) ln(Nu)
∑ ln(Re) ∑ ln(Re)
[ ]

Equation (5.10) was multiplied by Equation (5.11) then the final correlation was
obtained as shown in Equation (5.15).

Nu = 30.41 Re0.021785 Pr −0.344780 (5.15)

81
5.2. Mesh Independence

All analysis studied in this thesis were simulated with 2014258 mesh elements
(mesh type – 3) and 0.897 skewness. Mesh independence was also studied with natural
transformer oil to validate the results as shown in Table 15.

Table 15. Mesh Independence.

Number of
Outlet Pressure
Type Mesh Velocity Heat Flux Skewness
Temperature Difference
Elements
# [-] [m/s] [K] [Pa] [W/m²] [-]
1 8178016 0.05 345.7 190.0 285.9 0.931
2 2143041 0.05 345.7 190.8 284.9 0.899
3 2014258 0.05 345.7 190.8 284.9 0.897
4 1370706 0.05 345.9 191.6 285 0.899

Sweep method was used to generate mesh of oil ducts, inlet zone and outlet zone
in all mesh types. Oil ducts were meshed by using sweep bias, which has the biggest
element at the middle. Mesh size controls were used on inlet and outlet zones both with
(Figure 75) and without bias (Figure 76) in type – 1 and type – 3. Then, skewness was
obtained 0.931 and 0.897 for mesh type – 1 and mesh type – 2, respectively. On the
other hand, edge sizing controls were not used in mesh type – 4.

Figure 75. Edge Sizing at Inlet Zone with Bias.

82
Figure 76. Edge Sizing at Inlet Zone without Bias.

83
 CHAPTER 6

CONCLUSION

This chapter summarizes the thesis, discusses its findings and contributions,
shows the limitations, and also presents the future work recommendations.
CFD (Computational Fluid Dynamics) simulation of a power transformer
requires significant number of elements. Because of this reason, supercomputers are
needed to perform finite volume based simulation programs. At this point, some
approaches and assumptions can be useful to simplify problems. Porous media approach
is a method to solve complex geometries in Fluent. Flow characteristics and material
properties should be known in detail.
Transformer life directly depends on the aging of cellulose based insulation
material in winding. The hottest temperature of winding can be determined by using too
much element number. For this reason, velocity versus pressure difference graphs can
be created to use porous media approach. Then, cooling systems can be defined as
porous media in CFD simulations of transformers.
Mineral oils are the most common oil used in oil immersed transformers.
However, they have high risk of fire with 170 ˚C. Also, they are not environmentally
friendly. On the other hand, natural ester transformer oils are both environmentally
friendly and they have 360 ˚C fire point. Natural ester oils are fully biodegradable. They
have 97% biodegradability as seen in Table 5. Especially in recent years, environmental
issues have become more important at all around the world. Therefore, non-renewable
transformer cooling fluids will be replaced with renewable alternative cooling fluids in
the near future. Moreover, ester oil and synthetic transformer oils are preferred for wind
turbines in Europe and US.
The aim of this study is investigation of transformer radiator and behavior of
different types of transformer oil. This study also examines the different types of
transformer oil. Radiator model was simulated with using different transformer oils.
Outlet temperature, pressure difference, heat flux and temperature difference between
inlet and outlet of transformer were presented in Table 7, Table 8, Table 9 and Table 10
with natural ester, mineral, silicone and synthetic transformer oils, respectively. This

84
study also provides an approach and methodology to researchers, who studies on CFD
simulations of transformers. Transformer CFD simulations can be solved by using this
methodology.
First step of solution algorithm that is used for transformer hot-spot temperature
value was accomplished by this study. Investigation of radiator, oil ducts and cooling
system performance will be studied by using this method for cold start works in
transformers. Also, improvements to the transformer radiators will be studied.
Investigation on radiator groups will be studied to specify the hot-spot locations and
cooling performance in the light of this research. Finally, oil flow and heat transfer in
winding will be presented in detail. Then, this study will provide a methodology to
simulate a transformer in ANSYS Fluent CFD software.
The outcomes of this study will be verified in a special test room. This is a state
supported project where the thermal and hydrodynamic performances of transformer
cooling systems can be tested in this special test room located in Balikesir
Electromechanical Industrial Plants Corporation’s facilities. This test room along with
this thesis will provide better understanding of transformer cooling systems.

85
 REFERENCES

Arora, R.K. (2013, December). Different DGA techniques for monitoring of

transformers. International Journal of Electronics and Electrical Engineering,
(1), No.4.

El Wakil, N., Chereches, N.C.&Padet,J. (2006). Numerical study of heat transfer and
fluid flow in a power transformer. International Journal of Thermal Sciences,
45, 615-626.

Fdhila, R.B., Kranenborg, J., Laneryd, T., Olsson, C., Samuelsson, B., Gustafsson,
A.&Lundin, L. (2011). Thermal modelling of power transformer radiators using
a porous medium based CFD approach. Second International Conference on
Computational Methods for Thermal Problems, China.

Fernández, I., Ortiz, A. Delgado, F., Renedo, C.&Pérez,S. (2012). Comparative

evaluation of alternative fluids for power transformers. Electric Power Systems
Research, 98, 58-69.

Fonte, C. M., Lopes, J. C. & Dias, M. M. (2011, 6-9 June). CFD analysis of core type
power transformers. 21st International Conference on Electricity Distribution,
Frankfurt.

Gastelurrutia et al. (2009). Numerical modelling of natural convection of oil inside

distribution transformers. Applied Thermal Engineering, (31), 493-505.

IEC 60076-7 International Standards. Loading guide for oil-immersed power

transformers , Part 7, 2005-12.

Joshi, K.U., & Deshmukh, N. K. (2004). Thermal analysis of oil cooled transformer.
Cigre 2004 Session, Paris, France.

Karsai, K., Kerényi, D.&Kiss, L. (1987). Large power transformers. Hungary, Elsevier.

Kim, M., Cho, S.M. & Kim, J.K. (2012). Prediction and evaluation of the cooling
performance of radiators used in oil-filled power transformer applications with
non-direct and direct-oil-forced flow. Experimental Thermal and Fluid Science,
(44), 392-397.

86
Kulkarni, S.V. & Khaparde, S.A. (2005). Transformer engineering design and practice.
India, Mercel Dekker Inc.

Radakovic, Z. & Sorgic, M. (2010). Basics of detailed thermal-hydraulic model for

thermal design of oil power transformers. IEEE Transactions on Power
Delivery, 25(2), 790-802.

Rozga, P. (2013). Properties of new environmentally friendly biodegradable insulating

fluids for power transformers. 1st Annual International Interdisciplinary
Conference, AIIC 2013, 24-26 April, Portugal.

Skillen, A., Revell, A., Iacovides, H.&Wu, W. (2011). Numerical prediction of local
hot-spot phenomena in transformer windings. Applied Thermal Engineering, 36,
96-105.

Southern Nevada Regional Professional Development Program. (n.d.).Physical science.

Retrieved October 3, 2015 from the World Wide Web:
http://www.rpdp.net/sciencetips_v2/P12C6.htm

Wittmaack, R. (2014). Thermal design of power transformers via CFD. 11th. World
Congress on Computational Mechanics (WCCM XI).

87