Download as pdf or txt
Download as pdf or txt
You are on page 1of 9

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

net/publication/358331161

Thermal Analysis of Natural Cooling Type Distribution Transformer Retrofilled


With Natural Ester Oil

Article in IEEE Transactions on Dielectrics and Electrical Insulation · February 2022


DOI: 10.1109/TDEI.2022.3148447

CITATIONS READS
15 371

5 authors, including:

Anupam Dixit C. Ekanayake


The University of Queensland The University of Queensland
9 PUBLICATIONS 31 CITATIONS 105 PUBLICATIONS 2,538 CITATIONS

SEE PROFILE SEE PROFILE

Hui Ma Tapan Kumar Saha


The University of Queensland The University of Queensland
131 PUBLICATIONS 1,893 CITATIONS 666 PUBLICATIONS 16,187 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Anupam Dixit on 16 February 2022.

The user has requested enhancement of the downloaded file.


This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TDEI.2022.3148447, IEEE
Transactions on Dielectrics and Electrical Insulation

Thermal Analysis of Natural Cooling Type Distribution


Transformer Retrofilled with Natural Ester Oil
Anupam Dixit, Chandima Ekanayake, Hui Ma
Tapan K. Saha and Muhammad Hafeez Ansari
School of Information Technology and Electrical Engineering, The University of Queensland,
Brisbane, QLD 4072, Australia

ABSTRACT
Retrofilling distribution transformers with natural ester has attracted an increasing
interest among the utilities. Many retrofilled transformers have layered winding
structure and natural cooling mode. Thermal performance of such type of transformers
has not been thoroughly studied. To ensure the temperature rise in a layered winding
type retrofilled transformer complying with IEC-60076-14, it is necessary to analyse the
temperature distribution inside the transformer. In this paper, the temperature
distribution of a layered winding natural cooling type distribution transformer is
investigated under retrofilling scenario. Computational fluid dynamics (CFD) analysis is
performed to compare the cooling performance of the transformer filled with mineral oil
and that of the transformer filled with natural ester oil. The temperature distribution
along the height of the transformer winding is investigated through fiber optic sensors.
Moreover, dimensional analysis method is applied to understand the influence of several
parameters (oil characteristics and dimension of cooling ducts) on temperature
distribution and hot spot factor. The findings of this paper can support decision making
of retrofilling of layered winding natural cooling mode transformers.
Index Terms — Layer-type winding, natural ester oil, natural cooling, retrofilling,
thermal modelling
ester oil did not have desirable thermal performance due to its
higher viscosity. The oil flow rates and temperature distribution
1 INTRODUCTION in a naturally cooled transformer considering five types of
dielectric liquid were investigated in [9]. The above studies are
THE biodegradable and high flash point characteristics of mainly based on theoretical analysis and simulation.
natural ester oil has encouraged the retrofilling of mineral oil Temperature distribution in actual retrofilled transformers were
filled distribution transformers with natural ester oil. Over ten measured using fiber optic sensors before and after transformer
thousand transformers are retrofilled worldwide [1]. However, retrofilling [10, 11]. A rise in temperature after retrofilling was
due to the different physical and chemical properties of natural reported. However, the same conclusion cannot be readily
ester oil compared to the mineral oil, the oil flow rate and heat applied to transformers which have different winding
dissipation through the cooling ducts can vary [2, 3]. Therefore, geometries or are retrofilled with different types of oils.
a transformer may experience different thermal stresses after The winding geometry is one of the main factors affecting the
retrofilling. This may affect the transformer’s ageing pattern temperature distribution in a transformer. In disc type
and subsequently its remaining useful life [4, 5]. transformer windings, temperature distribution is influenced by
Thermal performance under retrofilling scenario for natural the rate of oil flow in radial and axial ducts whereas in layered
cooling mode was explored in [6]. The authors considered the type transformer windings it is influenced by rate of oil flow in
mineral oil and synthetic ester oil under retrofilling scenario. axial ducts and the space between the winding layers. Number
Thermal modelling of transformer considering three ester based of winding layers, arrangement of washers, number of discs per
oils was investigated in [7]. Oil temperature dependent pass and size of cooling ducts are some other parameters which
properties were evaluated and then used in computational fluid affect the rate of oil flow and hence, the temperature
dynamics (CFD) analysis to determine the rated top oil and hot distribution in a transformer [12-14]. Most of investigations
spot temperatures. Ambient temperature profile was also reported in the literature focused on thermal modelling of disc
considered in the analysis. In [8], oil flow rate and thermal winding transformers. They are summarized below.
performance of transformers filled with three types of oil Equivalent thermal hydraulic network based model and
(mineral hydrocarbon oil, gas to liquid hydrocarbon oil and thermal-electric analogy based models were implemented in
synthetic ester oil) were studied. It was found that the synthetic [15-17]. Though these models consume less computational
resources, the accuracy of the estimated temperatures is
Manuscript received on 20 June 2021, in final form 12 January 2022, compromised. In contrast, CFD models can provide more
accepted 15 January 2022. Corresponding author: H. Ma.
accurate temperature estimation at the cost of high

1070-9878 (c) 2021 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: University of Queensland. Downloaded on February 16,2022 at 05:43:59 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TDEI.2022.3148447, IEEE
Transactions on Dielectrics and Electrical Insulation

computational resources. In [18, 24], authors performed the each layer and seven layers of HV winding with 109 turns in
CFD simulations to investigate the oil flow rate and temperature each layer. There are two vertical ducts of width 3 mm for
distribution of natural cooling mode disc type winding cooling. Also, some spaces are provided between the layers and
transformer. Another CFD simulation to study the effect of turns in the winding. The detailed winding dimensions
models and input boundary conditions on temperature (digested from the manufacturer’s data sheet) are presented in
distribution of transformer was reported in [19]. It is concluded Table 1. Also, the inside view of the twin transformers and the
that higher oil flow rate could cause more uniform temperature experiment setup are shown in Figure 1.
distribution across the winding leading to a more efficient
cooling. In [20], a 3D CFD model was developed to estimate
the oil flow rate and temperature distribution considering
uniform and non-uniform power loss cases.
In [21], CFD parametric sweep and dimensional analysis
were conducted to interpret and investigate the hot spot factor.
The authors investigated the correlation between the hot spot
factor and dimensionless numbers such as Reynolds number
(Re) and Prandtl number (Pr). It was concluded that the
dimensionless numbers were non-linearly related to the hot spot
factor in case of winding with non-uniform losses. In [22], the
hot spot factor was experimentally investigated to determine the (a) (b)
hot spot temperature in disc winding transformers. In [23], an Figure 1. Twin transformers used in this study and experiment setup (a) an
inside view of the transformer’s core-winding geometry (b) experiment setup
approach to improve the thermal modelling of naturally cooled
transformers was proposed. Thermal model parameters were
The windings of both transformers are equipped with ten
estimated using the Levenberg-Marquardt algorithm.
bragg grating based fiber optic temperature sensors, which
The temperature distribution in a transformer is also
measure the temperature based on bragg wavelength. The
governed by its loading level and cooling mode. In forced
sensors are distributed at irregular intervals in winding ducts
cooling modes, the temperature rise is controlled using pumps
along the winding height. The illustration of the transformer
and fans. However, in natural cooling modes, the temperature
with fiber optic cable and temperature sensors arrangement are
rise is mainly controlled through the natural convection of oil.
provided in Figure 2. The sensors have temperature uncertainty
Also, the oil circulation depends on buoyancy forces and oil
of ±0.5 °C.
characteristics [6, 24]. In [6], thermal performance was
analyzed for forced and natural cooling modes of transformers Table 1. Dimension of winding components (from manufacturer’s data sheet)
when filled with three types of oil. It was found that in forced Component Dimension (in mm)
cooling modes, the flow rate of synthetic ester oil was more Winding height 185
uniform than that of the mineral oil. In natural cooling mode, LV winding (one turn height and width) 1.6 x 7.1
the hot spot temperature of synthetic ester oil filled transformer HV winding (one turn height and width) 1.4 x 1.4
LV winding insulation thickness 0.15
was higher than that of the other two types of oil. HV winding insulation thickness 0.087
An accurate temperature distribution in retrofilled Vertical ducts width 3
transformers will help utilities to appropriately load Space between LV winding layers 0.52
transformers to operate them within their thermal limits. To the Space between HV winding layers 0.4
Space between LV winding turns 1.214
best of the authors knowledge, temperature distribution in Space between HV winding turns 0.212
naturally cooled layered winding transformers in retrofilling
scenario have not been sufficiently explored.
In this paper, thermal modelling of naturally cooled layered
winding distribution transformers in retrofilling scenario is
investigated. To simulate the retrofilling scenario, twin
transformers with identical layered winding geometries filled
with mineral and natural ester oil respectively are used. CFD
simulations are conducted to investigate temperature
distribution in these two naturally cooled transformers.
Moreover, dimensional analysis is explored to study the
applicability of dimensionless parameters (e.g, Richardson
number and Reynolds number) to study the temperature
distribution in naturally cooled layered winding transformers. Figure 2. The transformer construction and fiber optic cable and temperature
sensors arrangement along winding height

2 EXPERIMENTAL SETUP
3 CFD SIMULATION
Twin 5 kVA transformers (230 V/2 kV) having identical winding
geometries are used in this study. One transformer is filled with The continuity equation, Navier-Stokes equations and energy
Shell Diala mineral oil and the other is filled with FR3 natural ester. equation are solved for the laminar flow of oil using the
Each transformer has four layers of LV winding with 22 turns in ANSYS®. A 2D model of the whole winding is implemented.

1070-9878 (c) 2021 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: University of Queensland. Downloaded on February 16,2022 at 05:43:59 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TDEI.2022.3148447, IEEE
Transactions on Dielectrics and Electrical Insulation

A section of the winding with two turns of low voltage (LV) Table 4. Relationship between oils physical properties and temperature
and six turns of high voltage (HV) is shown in Figure 3. 𝛿(𝑇) = 𝑦𝑜 + 𝑎. 𝑇
Density
(g/m2), 𝛿:
For mineral oil: yo = 0.896, a = -6.64 x 10-4
For FR3 oil: yo = 0.934, a = -6.72 x 10-4
Specific heat 𝐶𝑝 = 𝑎 + 𝑏 ∙ 𝜃
(kJ/(kg.K)), Cp:
For mineral oil: a = 1.755, b = 0.00415
For FR3 oil: a = 2.07, b = 0.00396
For mineral oil:
Viscosity (cSt), 5.1𝑋106
𝑣: For T < 20 °C, 𝑣 =
(𝜃+48.1)3
3.78𝑋104
For T > 20 °C, 𝑣 = − 0.24
(𝜃+28)2
For FR3 oil:
Figure 3. A section of winding geometry designed for simulation 7.227−2.82.log (𝜃+273.15)
𝑣 = 1010 − 0.7
The characteristics of copper conductors and insulation paper
Thermal 𝑘 = 𝑦0 + 𝑎 ∙ e−𝑏∙𝜃
conductivity
are presented in Table 2. The LV and HV windings insulation (W/(m•K)), k: For mineral oil:
paper thickness is set as 0.2 mm to get better mesh quality. Also, yo = 0.0757, a = 0.041, b = 0.006
the insulation paper wrapped the conductors are assumed to be For FR3 oil:
a thermally coupled wall. An oil natural cooling mode in a yo = 0.137, a = 0.036, b = 0.0136
closed tank is simulated. This is similar to the experimental
setup, in which oil flow and heat transfer are mutually coupled.
Oils viscosity vs temperature
Table 2. Properties of copper and paper insulation
1000
Copper Insulation FR3
Properties Kinematic Viscosity [cSt]
conductor MO
Density (kg/m2) 8978 930
Specific heat (J/(kg.K)) 381 1340 100
Thermal conductivity (W/(m•K)) 401 0.155

An initial ANSYS® analysis is performed to identify the


10
optimum number of meshing elements that can achieve
desirable accuracy while take less computational time. A mesh
independent analysis is performed as shown in Table 3. Mesh 4
is selected for simulation because of its desirable mesh quality 1
-20 30 80 130
and better solution convergence. Mesh 4 has an element size of Temperature [°C]
1.9x10-4 m which generates a mesh with 909,816 elements.
Figure 4. Comparison of oils viscosity with temperature [Source: Cargill Ltd]
Mesh 4 gives an orthogonal quality of about 0.95 for more than
99 % of the elements. Also, more than 99 % of the elements
have skewness of the order of 10-10. The solution converged 4 RESULTS
with velocities and continuity residuals of the order of 10 -5 For cylindrical symmetrical design of transformer winding, a
while energy residual of the order of 10 -11. 2D representation of plane of symmetry can save computational
resources without losing essential information of oil flow and
Table 3. Mesh refinement study results temperature distribution. Therefore, a 2D geometry of the
Mesh 1 Mesh 2 Mesh 3 Mesh 4 winding arrangement is adopted for CFD simulation in
No. of elements 381,529 528,871 832,086 909,816 ANSYS®. To ensure the mass and energy conservation, flux
Hot spot temperature 56.44 °C 57.97 °C 58.43 °C 58.24 °C
Simulation time 87 min 130 min 132 min 147 min reports for mass flow rate and heat transfer rate are checked
after each simulation. Moreover, the software inbuilt ‘SIMPLE’
In natural cooling mode, the oil flow rate and heat transfer are algorithm has been used for the pressure velocity coupling.
determined by temperature dependent characteristics of oil. The In this study the oil is not entering or leaving the transformer
relationship between oils characteristic and temperature tank like in the transformers with radiator arrangement, rather
(provided by the oil suppliers) are presented in Table 4. the oil maintains a buoyancy driven flow within the tank. For
Oil viscosity is the most dominant characteristic which such natural convection flows in a closed tank, defining inlet
governs the transformer thermal performance. The variation of and outlet locations and boundary conditions would not be very
viscosity with temperature for FR3 oil and mineral oil is shown appropriate. Therefore, the boundary conditions have been
in Figure 4. It is clear from the figure that the viscosity of FR3 considered for the wall of the transformer tank, through which
oil is higher than the mineral oil at all the temperatures. the heat dissipate to the ambience. Since the heat dissipates
Therefore, in retrofilled transformers, it is of prime importance through the wall by convection as well as radiation, combined
to check the mobility of natural ester oil in the windings which convection and external radiation thermal condition is
were originally designed to suit for mineral oil. simulated. The thermal boundary conditions for transformer

1070-9878 (c) 2021 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: University of Queensland. Downloaded on February 16,2022 at 05:43:59 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TDEI.2022.3148447, IEEE
Transactions on Dielectrics and Electrical Insulation

wall are summarized in Table 5. Also, thermal boundary 4.2 TRANSFORMER FILLED WITH FR3 OIL
conditions are considered as ‘coupled’ for the walls that form The simulation and experimental results for the transformer
the interface between two regions such as the interface between filled with FR3 oil are shown in Figure 6. It can be observed
copper conductors and insulation paper, and between insulation that the the temperatures obtained from simulation and the fiber
paper and oil medium. optic measurements follow a similar trend, i.e. the temperature
continuously rises along the height of the winding. However,
Table 5. Thermal boundary conditions for transformer tank wall
the simulation result of the FR3 oil filled transformer shows
Parameter Value
that, as compared to mineral oil filled transformer, the
Heat transfer coefficient (W/(m2•K)) 5
Free stream temperature (°C) 26.85
maximum winding temperature shift towards the middle of the
External emissivity 0.85 winding. This could be due to the higher viscosity of FR3 oil.
External radiation temperature (°C) 26.85
Temperature distribution in FR3 oil filled
The HV and LV windings power losses used in the simulation transformer
are 38 W and 34 W respectively, which are obtained from the 58
transformer’s data sheet. The temperature is calculated in total 56

Temperature (in °C)


at 110 points along the winding height for both LV and HV
winding’s layers. The average value of temperature for all the 54
winding layers are considered assuming that there is 52
insignificant variation of temperature between two points at the
same height. Also, ten fiber optic sensor measurements are used 50
to obtain the temperature along the winding height.
48 Fiber optic sensor
Simulation
46
4.1 TRANSFORMER FILLED WITH MINERAL OIL 0 50 100 150 200
Bottom Winding height (in mm) Top
The fiber optic sensor measurements show a continuous
temperature rise from the bottom to the top of the winding, as Figure 6. Winding temperature distribution in FR3 oil filled transformer
shown in Figure 5. However, the simulation result shows that the
temperature firstly rises to a maximum value and then drops The lower oil temperature at the bottom of the winding keeps
slightly near the top of the winding. The drop in winding lower viscosity and density of oil resulting in lower
temperature near the top may because of the high oil flow rate. Due thermosyphon forces and a lower oil flow rate. As a result, the
to the absence of fiber optic sensors at the top of the winding, the maximum temperature occurs at the middle part of the winding,
above temperature drop near the winding top could not be further which reduces the viscosity and density of oil in the vicinity of
verified from the measurement. the middle part of the winding. This in turn could develop
higher thermosyphon forces and oil flow rate in the upper part
of the winding resulting in lower winding temperature at the
Temperature distribution in mineral oil filled top.
transformer The shift in the location of the maximum winding temperature
57 in FR3 oil filled transformer could not be observed because
56 limited number of fiber optic sensors are placed on the winding.
Temperature (in °C)

55 Also, there are no sensors placed at the top and bottom ends of
54 the winding. However, the temperature measured by the
53 available sensors follows the pattern of temperature rising from
52 the bottom towards the top of the winding. The placement of a
51 large number of sensors in close vicinity throughout the
50 winding could obtain a more accurate winding temperature
Fiber optic sensor
49 distribution and capture the shift in the location of maximum
Simulation
48 winding temperature.
0 50 100 150 200 For FR3 oil filled transformer, the maximum winding
Bottom Winding height (in mm) Top
temperature measured by fiber optic sensor is about 56 °C and
Figure 5. Winding temperature distribution in mineral oil filled transformer the temperature gradient of about 9 °C is noted. This indicates
that the winding temperature for FR3 oil filled transformer is
The fiber optic measurement shows a temperature gradient of about 2 °C higher than the mineral oil filled transformer under
about 5 °C from the bottom to the top of the winding and the the same loading conditions. As per simulation results,
maximum winding temperature is about 54 °C. The maximum temperature rises by about 1 °C when mineral oil is replaced
winding temperature obtained through simulation is 56 °C. The with FR3 oil for the same winding geometry and loading
trend of temperature distribution obtained from both conditions. Such temperature rise is mainly due to the higher
measurement and simulation model shows that the hot spot is viscosity of FR3 oil, which reduces down the oil flow
near the top of the winding. circulation through the windings.

1070-9878 (c) 2021 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: University of Queensland. Downloaded on February 16,2022 at 05:43:59 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TDEI.2022.3148447, IEEE
Transactions on Dielectrics and Electrical Insulation

5 DISCUSSION duct is adjacent to the LV winding and receives more heat as


compared to the second duct.
5.1 OIL FLOW RATE At the bottom of the winding, the ratio of oil flow rate in
mineral oil filled transformer to that in FR3 oil filled
Since there are no pumps or radiators attached to the twin transformer is 1.31. However, at the top of the winding, this
transformers investigated in this study, the oil circulation in ratio is 1.87. On an average, the oil flow rate in mineral oil filled
these transformers are mainly controlled by the thermosyphon transformer is 1.59 times of the oil flow rate in FR3 oil filled
forces. The heat from the winding is transferred to the oil transformer. This value is very close to the values mentioned in
through convection and the hot oil moves upward. Once the oil [6]. However, the oil flow rate is also dependent on the winding
cools down, it moves downward. losses, geometry, volume of oil, and the oil characteristics.
In layered winding transformers oil flows through vertical Therefore, further investigations are needed.
cooling ducts and the space between winding layers and turns.
As illustrated in Figure 7, the maximum volume of oil flows
5.2 TEMPERATURE DISTRIBUTION
between the outer layer of HV winding and the tank. A
significant amount of oil also flows through the vertical cooling For both the mineral oil filled and FR3 oil filled transformers
ducts and the space between LV winding and the core. the hot spots do not occur at the top of the winding, instead
However, the dimensions of the space provided between the between 65 % to 90 % height of the winding. The whirlpool of
winding layers and turns are not enough for an adequate oil flow oil near the winding top (Figure 7) maintains a high oil flow
and therefore higher temperature develops between the layers rate and facilitates the heat dissipation from the winding. Since
and turns of the winding. in natural cooling mode the oil flow rate and temperature
distribution are mutually coupled, the higher temperature at
about 70 % of winding height may cause more thermosyphon
forces. This may also increase the oil flow rate at the top of the
winding resulting in a decrease of temperature there.
Even though the power losses are less in the LV winding as
compared to the HV winding, the smaller surface area of LV
winding lead to the dissipation of more heat flux per unit of its
surface area. The temperature contour obtained from simulation
is shown in Figure 9. It can be seen that the hot spot temperature
in the FR3 oil filled transformer is about 1.7 °C more than the
mineral oil filled transformer.
The rise in temperature along the winding height is because
of low thermal diffusion rate and higher viscosity of oils, which
causes the generation of hot streaks [19]. The hot streaks are
Figure 7. Oil flow velocity vectors for layered winding geometry clearly visible in the layers of the winding, specially near the
top of the winding as shown in Figure 9. These hot streaks are
The simulation results for the oil flow rate in mineral oil the result of separation of the thermal boundary layers which
filled and FR3 oil filled transformers are shown in Figure 8. develops on the winding turns. Since the oil flow rate is
significantly small between the layers of the winding, the hot
streaks generated from the bottom winding turns do not vanish
and add to the hot streaks generated from the upper winding
turns. This results in a low heat transfer rate in the upper side of
the winding and hence, rise in the temperature.
Though the layered winding transformers investigated in this
paper have relatively small height, a significant temperature
gradient in the winding has been observed. For high voltage
rating distribution transformers, more hot streaks can be
developed due to the increase of their winding heights.
Subsequently, this can lead to a higher rise of temperature of
the transformer after retrofilling.
The power transformers (more than 100 kV rating) with disc
type winding could experience 20 °C and even more rise in hot
spot temperature after retrofilling [10, 11]. Layered winding
natural cooled distribution transformers with lower voltage
Figure 8. Oil velocities in ducts of mineral oil filled and natural ester filled
transformers
ratings may also experience a high rise in temperature, which is
greatly dependent on their winding geometry and size of
It can be observed that the variation of oil flow rate from the cooling ducts. Therefore, necessary feasibility study should be
bottom to the top of the winding is more in the first duct as conducted before retrofilling large distribution transformers
compared to the second duct. This could be because the first operating at higher voltage levels.

1070-9878 (c) 2021 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: University of Queensland. Downloaded on February 16,2022 at 05:43:59 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TDEI.2022.3148447, IEEE
Transactions on Dielectrics and Electrical Insulation

𝜕𝑇 ∗ 𝜕𝑇 ∗ 1 1 𝜕 𝜕𝑇 ∗ 𝜕2 𝑇 ∗
𝑣𝑟∗ + 𝑣𝑧∗ = ( (𝑟 ∗ )+ ) (4)
𝜕𝑟 ∗ 𝜕𝑧 ∗ 𝑅𝑒.𝑃𝑟 𝑟 ∗ 𝜕𝑟 ∗ 𝜕𝑟 ∗ 𝜕𝑧 ∗ 2

where 𝑟 ∗ , 𝑧 ∗ , 𝑣𝑟∗ , 𝑣𝑧∗ , and 𝑝∗ are dimensionless variables given


as:
𝑟 𝑧
𝑟∗ = and 𝑧∗ = (5)
2×𝑊𝑑𝑢𝑐𝑡 2×𝑊𝑑𝑢𝑐𝑡
𝑣𝑟 𝑣𝑧 𝑝
𝑣𝑟∗ = and 𝑣𝑧∗ = and 𝑝 = ∗
(6)
𝑣𝑏 𝑣𝑏 𝑣𝑏2

where ‘𝑟’ is the radial direction coordinate, ‘𝑧’ is the axial


direction coordinate, ‘𝑣𝑟 ’ is the radial velocity (m/s), ‘𝑣𝑧 ’ is the
axial velocity (m/s), ‘𝑝’ is the static pressure (Pa), ‘𝑊𝑑𝑢𝑐𝑡 ’ is
the width of cooling duct (m), ‘𝑣𝑏 ’ is the oil velocity at the
bottom side of winding (m/s), and ‘’ is the oil density (kg/m3).
Reynolds number (Re), Prandtl number (Pr), and Richardson
number (Ri) obtained from the dimensionless equations are
expressed as [24]:

𝜌𝑣𝑏 (2×𝑊𝑑𝑢𝑐𝑡 )
𝑅𝑒 = (7)
𝜇
𝜇𝐶𝑝
𝑃𝑟 = (8)
Figure 9. Temperature contour for mineral oil filled (left) and FR3 oil filled 𝑘
𝐺𝑟
transformers (right) 𝑅𝑖 = (9)
𝑅𝑒2

6 DIMENSIONAL ANALYSIS where ‘𝜇’ is oil viscosity (Pa.s), ‘𝐶𝑝 ’ is oil specific heat
For a given transformer geometry, a number of influencing (J/(kg.K)), and ‘𝑘’ is oil thermal conductivity (W/(m.K)). ‘Gr’
parameters such as oil types, oil properties and oil flow rates is the Grashof number, which is expressed as [24]:
etc. affect the temperature distribution of the transformer
winding. During the decision-making of transformer 𝑇 +𝑇
𝑔𝛽𝜌2 (𝑇𝑤 − 𝑏2 𝑡 )(2×𝑊𝑑𝑢𝑐𝑡 )3
retrofilling, it is necessary to investigate different combinations 𝐺𝑟 = (10)
𝜇2
of these parameters on transformer thermal performance before
and after retrofilling. The number of the combinations of these
where, ‘Tt’ is the oil temperature at the winding top, ‘Tb’ is the
parameters could be very high and may introduce analysis
oil temperature at the winding bottom, ‘Tw’ is the average
burdens.
winding temperature, ‘𝑔’ is gravitational acceleration (m/s2),
Applying dimensional analysis allows for the influencing
and ‘𝛽’ is oil volume thermal expansion coefficient (K-1). ‘𝑇 ∗ ’
parameters to be grouped into fewer dimensionless variables.
The influence of these dimensionless variables on temperature is the dimensionless temperature which is expressed as [24]:
distribution can give the same results as the influence of 𝑇−𝑇𝑡
different combinations of parameters [18]. Using the 𝑇∗ = (11)
𝑇𝑤 −(𝑇𝑡 +𝑇𝑏 )/2
dimensional analysis, a generalized result can be obtained for a
given transformer geometry, no matter which type of oil will be The expression in Equation (11) is similar to hot spot factor
used for retrofilling. expression [24]:
The oil flow and heat transfer phenomenon in natural cooling 𝑇ℎ𝑠 −𝑇𝑡
𝐻= (12)
mode are strongly coupled and are governed by the laws of 𝑇𝑤 −(𝑇𝑡 +𝑇𝑏 )/2
conservation of mass, momentum, and energy. The dimensional
analysis has been applied on these conservation equations to except the hot spot temperature ‘Ths’ has been replaced by any
obtain the dimensionless variables. The dimensionless form of temperature in fluid domain. The maximum value of
laws of conservation of mass, momentum and energy for fluid dimensionless temperature is nothing but the hot spot factor
flow in cylindrical coordinates are given as below [24]. only [18].
Firstly, three cases are simulated with different oil velocities
1 𝜕(𝑟 ∗ 𝑣𝑟∗ ) 𝜕𝑣𝑧∗ at the inlet of winding while keeping the ‘Re’ same by varying
𝑟 ∗ 𝜕𝑟 ∗
+ 𝜕𝑧 ∗
=0 (1) the density and viscosity of oil. The purpose is to verify the
applicability of dimensional analysis. The dimensionless
𝜕𝑣𝑟∗ 𝜕𝑣𝑟∗ 𝜕𝑝∗ 1 𝜕 1 𝜕(𝑟 ∗ 𝑣𝑟∗ ) 𝜕2 𝑣𝑟∗ temperature is determined at the same location for all the three
𝑣𝑟∗ + 𝑣𝑧∗ = − + ( ( )+ ) (2)
𝜕𝑟 ∗ 𝜕𝑧 ∗ 𝜕𝑟 ∗ 𝑅𝑒 𝜕𝑟 ∗ 𝑟 ∗ 𝜕𝑟 ∗ 𝜕𝑧 ∗ 2 cases. The repetition of dimensionless temperature value for
𝜕𝑣𝑧∗ 𝜕𝑣𝑧∗ 1 1 𝜕 𝜕𝑣𝑧∗ 𝜕2 𝑣𝑧∗
each case, as shown in Table 6, proofs the applicability of the
𝑣𝑟∗ + 𝑣𝑧∗ = 𝑅𝑖. 𝑇 ∗ + ( (𝑟 ∗ )+ ) (3) method. As per the dimensionless equations (1) - (4), the
𝜕𝑟 ∗ 𝜕𝑧 ∗ 𝑅𝑒 𝑟 ∗ 𝜕𝑟 ∗ 𝜕𝑟 ∗ 𝜕𝑧 ∗ 2

1070-9878 (c) 2021 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: University of Queensland. Downloaded on February 16,2022 at 05:43:59 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TDEI.2022.3148447, IEEE
Transactions on Dielectrics and Electrical Insulation

temperature distribution is mainly a function of Ri, Re and Pr when oil stagnates at some locations which again results in
[18]. higher values of dimensionless temperature. A minimum
dimensionless temperature is achieved where heat transfer and
Table 6: Simulation test cases
oil flow rate are balanced giving an optimal ‘Ri’ [24].
Case 1 Case 2 Case 3
In this paper, the values of ‘Ri’ is changed by varying the
Oil velocity (m/s) 0.02 0.03 0.04 gravitational acceleration term. Due to the low voltage rating of
Dimensionless temperature 1.2 1.2 1.2
the transformers in this study, the thermosyphon forces are
smaller and the resulting oil flow rates are in the order of 10-4,
In this paper, only the influence of ‘Re’ and ‘Ri’ on which give very high values of ‘Ri’. Such large ‘Ri’ represent
temperature distribution has been analysed. Influence of ‘Pr’ on the complete natural cooling mode. For the layered winding
temperature distribution is less as compared to ‘Re’ and geometry in this study, it is observed that the dimensionless
therefore not included in the study [21, 24]. temperature shows the similar pattern along the winding height
For the layered winding geometry studied in this paper, oil
for each ‘Ri’ value, i.e. it firstly increases, reaches to a
velocity, oil characteristics (density and viscosity) and the size
maximum value and then decreases, as shown in Figure 11.
of cooling ducts are grouped together. The ‘Re’ value is varied
from 8 to 97 by changing the oil viscosity since it is the most
Ri Vs Dimensionless temperature
dominating parameter of oil which may govern the temperature 2

Dimensionless temperature
distribution in retrofilled transformers. The dimensionless
temperature is calculated along the winding height for each ‘Re’ 1.5

value, as shown in Figure 10.


1

Re Vs Dimensionless temperature Ri = 9485 Ri = 9455


0.5
5 Ri = 9508 Ri = 9816
Dimensionless temperature

Ri = 10071 Ri = 10377
0 Ri = 10835 Ri = 11436
Ri = 12218 Ri = 13498
0 Ri = 15789
-0.5
Location of maximum 0
7
15
22
30
37
45
52
59
67
74
82
89
97
104
112
119
126
134
141
149
156
164
171
184
dimensionless temperature
Winding height (in mm)
-5
Figure 11. Influence of Ri on dimensionless temperature (non-linear x-axis)
Re = 8 Re = 11
Re = 23 Re = 42
Re = 62 Re = 97 For the twin transformers used in this study, change of oil
-10
from mineral to FR3 reduces the ‘Ri’ value from 15114 for the
106
113
121
128
136
143
151
158
171
184
0
7
15
22
30
37
45
52
59
67
74
82
91
99

mineral oil filled unit to 11368 for the FR3 oil filled unit. The
Winding height (in mm)
maximum dimensionless temperature shifting toward the
Figure 10. Influence of Re on dimensionless temperature (non-linear x-axis)
middle side of the winding as shown in Figure 11 coincides with
the maximum winding temperature shifting as shown in Figure
On increasing the ‘Re’ value, it can be observed from Figure 5 and 6.
10 that the dimensionless temperature on the bottom side of
A database consisting of a range of ‘Re’ and ‘Ri’ values and
winding decreases while it increases on the top side of winding.
the corresponding dimensionless temperatures can be
Also, it is noted that the maximum value of dimensionless developed. This can help to predict the temperature distribution
temperature increases with the increase of ‘Re’. However, the for a fleet of large layered winding natural cooled distribution
location of maximum dimensionless temperature remains the transformers under retrofill scenario. It is worth mentioning that
same with the increase in ‘Re’. the number and location of ducts may have different influence
The maximum dimensionless temperature does not vary
on maximum dimensionless temperature with the variations to
significantly when ‘Re’ values are smaller than eight. For the
‘Re’ and ‘Ri’. This has not been adequately proven and
twin transformers used in this study, due to their small size and therefore, further investigation is required.
the natural cooling modes, ‘Re’ is less than two and the
corresponding maximum dimensionless temperature is 1.72.
However, for high voltage rating distribution transformers, the 7 CONCLUSION
‘Re’ may vary significantly and hence, the maximum The thermal performance of a layered winding natural cooled
dimensionless temperature. distribution transformer was investigated in retrofilled scenario
In naturally cooled transformers the heat transfer and oil flow by both CFD simulations and fiber optic sensor measurements.
rate are strongly coupled. The ‘Ri’ value shows the relative The study showed that the location of maximum winding
effect of heat transfer and oil flow rate. A lower value of ‘Ri’ temperature is largely dependent on the oil physical
implies less heat transfer and higher oil flow rate. This usually characteristics. A rise in maximum winding temperature was
does not occur in naturally cooled transformers as oil flow rate expected in retrofilled scenario, which was evident from both
depends on thermosyphon forces, to which the heat transfer is measurement and simulation results. However, there was only
responsible [24]. Such scenario will lead to a higher value of about 2 °C rise in maximum winding temperature due to the
dimensionless temperature. A higher value of ‘Ri’ implies more small size of the transformers used in this study. As per the CFD
heat transfer and less oil flow rate. This represents the situations simulations more hot streaks could be formed in the natural

1070-9878 (c) 2021 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: University of Queensland. Downloaded on February 16,2022 at 05:43:59 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TDEI.2022.3148447, IEEE
Transactions on Dielectrics and Electrical Insulation

ester filled units causing higher winding temperature gradients distribution and temperature distribution in disc type ON cooled
transformers," in 2017 IEEE 19th International Conference on Dielectric
compared to that of the mineral oil filled units. Both
Liquids (ICDL), 25-29 June 2017 2017, pp. 1-4, doi:
measurement and simulation indicated that the winding 10.1109/ICDL.2017.8124677.
temperature gradient of the natural ester filled unit was more [9] R. Lecuna, F. Delgado, A. Ortiz, P. B. Castro, I. Fernandez, and C. J.
than 1.25 times compared to that of the mineral oil filled unit Renedo, "Thermal-fluid characterization of alternative liquids of power
transformers: A numerical approach," IEEE Transactions on Dielectrics
used in this study. Therefore, when retrofilling large layered
and Electrical Insulation, vol. 22, no. 5, pp. 2522-2529, 2015, doi:
winding distribution transformers the utilities should 10.1109/tdei.2015.004793.
thoroughly investigate the impact of hot streaks on temperature [10] D. Kweon, K. Koo, J. Woo, and Y. Kim, "Hot spot temperature for 154
distribution in distribution transformers. kV transformer filled with mineral oil and natural ester fluid," IEEE
Transactions on Dielectrics and Electrical Insulation, vol. 19, no. 3, pp.
To improve the efficiency of analysing a large number of
1013-1020, 2012, doi: 10.1109/TDEI.2012.6215107.
different parameters that affect the transformer thermal [11] R. Girgis, M. Bernesj¨o, and G. K. Frimpong, "Detailed performance of
performance, the dimensional analysis was performed in this a 50 MVA transformer filled with a natural ester fluid versus mineral
paper. Influence of Reynolds number and Richardson number oil," presented at the CIGRE 2010, 2010. [Online]. Available: https://e-
cigre.org/publication/A2-107_2010-detailed-performance-of-a-50-mva-
on the dimensionless temperature has been investigated. For the
transformer-filled-with-a-natural-ester-fluid-versus-mineral-oil.
layered winding geometry used in this study, it is observed that [12] F. Torriano, H. Campelo, M. Quintela, P. Labbé, and P. Picher,
an increase in Reynolds number results in an increase of the "Numerical and experimental thermofluid investigation of different disc-
maximum dimensionless temperature (hot spot factor) with the type power transformer winding arrangements," International Journal of
Heat and Fluid Flow, vol. 69, pp. 62-72, 2018, doi:
location of maximum dimensionless temperature remaining
10.1016/j.ijheatfluidflow.2017.11.007.
almost the same. In the case of the Richardson number, the [13] M. Lee, H. A. Abdullah, J. C. Jofriet, and D. Patel, "Thermal modeling
magnitude of maximum dimensionless temperature does not of disc-type winding for ventilated dry-type transformers," Electric
follow any distinguishable pattern. However, with the increases Power Systems Research, vol. 80, no. 1, pp. 121-129, 2010, doi:
10.1016/j.epsr.2009.08.007.
of the Richardson number, the location of maximum
[14] M. Daghrah, Z. D. Wang, Q. Liu, D. Walker, P. W. R. Smith, and P.
dimensionless temperature shifts toward the top side of the Mavrommatis, "Design of experimenal setup to study factors affecting
winding. hot spot temperature in disc type winding transformers," in IET
The results of dimensional analysis presented in this paper International Conference on Resilience of Transmission and
Distribution Networks (RTDN) 2015, 22-24 Sept. 2015 2015, pp. 1-6,
may help to better understand the thermal performance of
doi: 10.1049/cp.2015.0899.
layered winding natural cooled transformers and support the [15] Z. R. Radakovic and M. S. Sorgic, "Basics of Detailed Thermal-
decision on their retrofilling. However, more CFD simulations Hydraulic Model for Thermal Design of Oil Power Transformers," IEEE
need to be performed considering different layered winding Transactions on Power Delivery, vol. 25, no. 2, pp. 790-802, 2010, doi:
10.1109/tpwrd.2009.2033076.
geometries to obtain the universally applicable results.
[16] G. Swift, T. S. Molinski, and W. Lehn, "A fundamental approach to
transformer thermal modeling. I. Theory and equivalent circuit," IEEE
REFERENCES Transactions on Power Delivery, vol. 16, no. 2, pp. 171-175, 2001, doi:
10.1109/61.915478.
[1] J. Valmus, "Natural ester dielectric fluid overview," NWPPA EO ETF [17] W. H. Tang, Q. H. Wu, and Z. J. Richardson, "A simplified transformer
Meeting, vol. Spokane, Wa, 2016. [Online]. Available: thermal model based on thermal-electric analogy," IEEE Transactions
https://www.nwppa.org/wp-content/uploads/NWPPA-ETF-FR3- on Power Delivery, vol. 19, no. 3, pp. 1112-1119, 2004, doi:
Presentation.pdf. 10.1109/tpwrd.2003.822968.
[2] D. M. Mehta, P. Kundu, A. Chowdhury, V. K. Lakhiani, and A. S. Jhala, [18] X. Zhang, M. Daghrah, Z. Wang, and Q. Liu, "Flow and temperature
"A review of critical evaluation of natural ester vis-a-vis mineral oil distributions in a disc type winding-Part II: Natural cooling modes,"
insulating liquid for use in transformers: Part II," IEEE Transactions on Applied Thermal Engineering, vol. 165, 2020, doi:
Dielectrics and Electrical Insulation, vol. 23, no. 3, pp. 1705-1712, 10.1016/j.applthermaleng.2019.114616.
2016, doi: 10.1109/tdei.2016.005371. [19] F. Torriano, M. Chaaban, and P. Picher, "Numerical study of parameters
[3] U. M. Rao, H. Pulluri, and N. G. Kumar, "Performance analysis of affecting the temperature distribution in a disc-type transformer
transformer oil/paper insulation with ester and mixed dielectric fluids," winding," Applied Thermal Engineering, vol. 30, no. 14-15, pp. 2034-
IEEE Transactions on Dielectrics and Electrical Insulation, vol. 25, no. 2044, 2010, doi: 10.1016/j.applthermaleng.2010.05.004.
5, pp. 1853-1862, 2018, doi: 10.1109/tdei.2018.007224. [20] F. Torriano, P. Picher, and M. Chaaban, "Numerical investigation of 3D
[4] J. Tokunaga, H. Koide, K. Mogami, and T. Hikosaka, "Comparative flow and thermal effects in a disc-type transformer winding," Applied
studies on the aging of thermally upgraded paper insulation in palm fatty Thermal Engineering, vol. 40, pp. 121-131, 2012, doi:
acid ester, mineral oil, and natural ester," IEEE Transactions on 10.1016/j.applthermaleng.2012.02.011.
Dielectrics and Electrical Insulation, vol. 23, no. 1, pp. 258-265, 2016, [21] X. Zhang, Z. Wang, and Q. Liu, "Interpretation of Hot Spot Factor for
doi: 10.1109/tdei.2015.005502. Transformers in OD Cooling Modes," IEEE Transactions on Power
[5] R. Liao, J. Hao, G. Chen, Z. Ma, and L. Yang, "A comparative study of Delivery, vol. 33, no. 3, pp. 1071-1080, 2018, doi:
physicochemical, dielectric and thermal properties of pressboard 10.1109/tpwrd.2017.2710087.
insulation impregnated with natural ester and mineral oil," IEEE [22] M. Daghrah, Z. D. Wang, Q. Liu, D. Walker, C. Krause, and G. Wilson,
Transactions on Dielectrics and Electrical Insulation, vol. 18, no. 5, pp. "Experimental investigation of hot spot factor for assessing hot spot
1626-1637, 2011, doi: 10.1109/TDEI.2011.6032833. temperature in transformers," in 2016 International Conference on
[6] M. Daghrah, Z. Wang, Q. Liu, A. Hilker, and A. Gyore, "Experimental Condition Monitoring and Diagnosis (CMD), 25-28 Sept. 2016 2016, pp.
Study of the Influence of Different Liquids on the Transformer Cooling 948-951, doi: 10.1109/CMD.2016.7757981.
Performance," IEEE Transactions on Power Delivery, vol. 34, no. 2, pp. [23] S. Najar, J.-F. Tissier, S. Cauet, and E. Etien, "Improving thermal model
588-595, 2019, doi: 10.1109/tpwrd.2019.2895533. for oil temperature estimation in power distribution transformers,"
[7] M. M. M. Salama, D.-E. A. Mansour, M. Daghrah, S. M. Abdelkasoud, Applied Thermal Engineering, vol. 119, pp. 73-78, 2017, doi:
and A. A. Abbas, "Thermal performance of transformers filled with 10.1016/j.applthermaleng.2017.03.061.
environmentally friendly oils under various loading conditions," [24] X. Zhang, Z. Wang, Q. Liu, P. Jarman, and M. Negro, "Numerical
International Journal of Electrical Power & Energy Systems, vol. 118, investigation of oil flow and temperature distributions for ON
2020, doi: 10.1016/j.ijepes.2019.105743. transformer windings," Applied Thermal Engineering, vol. 130, pp. 1-9,
[8] X. Zhang, Z. Wang, Q. Liu, M. Negro, A. Gyore, and P. W. R. Smith, 2018, doi: 10.1016/j.applthermaleng.2017.10.092.
"Numerical investigation of influences of liquid types on flow

1070-9878 (c) 2021 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: University of Queensland. Downloaded on February 16,2022 at 05:43:59 UTC from IEEE Xplore. Restrictions apply.
View publication stats

You might also like