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

net/publication/318274980

CFD Study of Fluid Flow and Temperature Distributions in a Power

Transformer Winding

Conference Paper · June 2017

DOI: 10.1109/ICDL.2017.8124674

CITATIONS READS

8 210

4 authors, including:

Stefan Tenbohlen Raphael Lebreton

Universität Stuttgart General Electric
378 PUBLICATIONS   3,690 CITATIONS    7 PUBLICATIONS   33 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Optimization of different technologies of EMC Filter for High Voltage Cable Harness in Electric Vehicle View project

Failure Analysis View project

All content following this page was uploaded by Saeed Khandan Siar on 17 January 2018.

The user has requested enhancement of the downloaded file.

 19th IEEE International Conference on Dielectric Liquids (ICDL), Manchester, United Kingdom, 25 – 29 June, 2017

CFD Study of Fluid Flow and Temperature

Distributions in a Power Transformer Winding
Khandan. S, Tenbohlen. S Breuer. C Lebreton. R
Institute of Power Transmission and R&D Engineer Power transformers R&D Engineer Power transformers
High Voltage Technology (IEH) CCM, GE Grid Solutions, TICC, GE Grid Solutions
University of Stuttgart Moenchengladbach, Germany Paris, France
Stuttgart, Germany christian.breuer@ge.com raphael.lebreton@ge.com
saeed.khandan@ieh.uni-stuttgart.de
stefan.tenbohlen@ieh.uni-stuttgart.de

Abstract—The focus of the present numerical study is on the method in power transformer windings and in [10], they
temperature distribution and the fluid flow of an oil cooled measured the oil speed in horizontal channels of an OD cooled
winding of a power transformer. The experimental setup consists winding model in comparison to a 3D CFD model to give an
of three passes of a zig-zag disc type winding, in which losses are insight of flow behaviors in the OD cooling method. A.
modeled by heating cartridges in each winding segment. Precise Weinläder and S. Tenbohlen [11] investigated the thermal
temperature sensors measure the temperature of each turn. The performance of transformer windings by simulating and
laboratory setup allows the exact control of the boundary measuring temperature distributions over a whole pass of a
conditions, e.g. the oil flow rate and the inlet temperature. winding segment. In this contribution, the experimental model
Furthermore, a simulation is carried out with a free open source
of a winding is explained and its thermal performance is
computational fluid dynamics solver OpenFOAM and validated
with the experimental results. The good agreement of the
investigated numerically with numerical CFD method. The
simulation results with experimental measurements validates the results are compared with the experimental measurements and
model. The model utilizes the turbulent flow for the different mass the hot-spot values for different inlet flow rates of oil are
flow rates of the oil. illustrated. Moreover, the numerical approach in this paper is
focused on efficient, free and fast implementation of complex
fluid flow and coupled conjugated heat transfer in windings to
Keywords— CFD, Conjugated Heat Transfer, Power Transformers, extend and optimize numerical modeling; therefore,
Temperature Distribution OpenFOAM as the free open source CFD software is used for
I. INTRODUCTION 2D investigations of thermal behaviors in the winding.
Power transformers are valuable devices in power systems. II. EXPERIMENTAL SETUP
The hot-spot temperature of a power transformer is the most
critical factor for its aging, which should be held under a In order to obtain measurement data for validation of
recommended value determined by the insulation class of the numerical results, a winding model for experimental setup is
transformer. Inadequate cooling leads to the excessive increase prepared, which allows measuring the temperature at every
in winding hot-spot temperatures, which can cause winding conductor element and detecting the location of hot-spot in
insulation and oil quality deteriorations and consequently each pass of the winding. Due to the similarities in thermal
accelerated aging of the transformer. Therefore, the cooling behaviors and the fluid flow patterns in the entire winding of
system of a power transformer plays an important role for the common transformer design, the experimental winding
dissipating the heat generated in the windings to the model used a section of 8° in the circumferential direction of an
surrounding ambient. In this context, the optimal design of the entire winding turn, so that all the details of a winding can be
cooling system in addition to its proper operation is necessary observed in a section. Furthermore, the corresponding losses
to avoid operating failures and thermal damages. per complete winding turn are referenced proportionally.
Meanwhile, the actual applied heating power per conductor in
Traditional methods of transformer winding thermal design the model and the oil flow rates passing the winding models are
utilized simple models based on hot-spot factors [1, 2] and 8°/360° = 1/45 times the referenced values pointed out in the
thermo-hydraulic networks [3-5] corroborated with figures captions. In this contribution, the inlet temperature is
measurements on winding models to simulate the effect of set to ϑin = 80°C while the oil flow rates are set to ṁoil = 9 kg/s,
winding losses [6]. Torriano et al. [7] presented a detailed and 18.0 kg/s. Consequently, the fluid behaviors for selected
numerical CFD study based on parameters affecting the flow rates lead to turbulent flow pattern simulation, which are
temperature distributions in the disc-type winding. Amoiralis et noticeable for investigations. The losses per winding turn are
al. [8] wrote an IEEE transaction on literatures based on kept at Plosses, i = 360 W/turn. Fig. 1a shows a top view of the
transformer design and optimization. Tenbohlen et al. [9] used symmetrical section of the transformer winding, which the
2D and 3D numerical simulations with experimental validation experimental winding model is based upon. Moreover, some
to evaluate the thermal behavior of oil directed (OD) cooling characteristic dimensions of the geometry are given.

1
978-1-5090-4877-9/17/$31.00 ©2017 IEEE
 19th IEEE International Conference on Dielectric Liquids (ICDL), Manchester, United Kingdom, 25 – 29 June, 2017

(b)

a) Top view of a winding segment. b) Cross section of the winding.

Fig. 1. Design dimensions of the modelled winding [9].

Fig. 2a shows the chosen winding model layout in a

schematic view. On the right, Fig. 2b provides a reference for
the pass, disc and conductor numbering which the presented
post analysis relies upon. (a) (c)
Fig. 3. a) Experimental winding model in tank during operation, b) View on
cartridges head inside conductors, c) View of sensors exits.

Accurate measurements of the individual conductor

temperatures are accomplished with temperature sensors
located inside the conductor. Next to a representative winding
design, the operating conditions of the modelled winding are
selected accordingly with respect to the winding type and
dimensions. While the oil inlet temperature is set via a
controlled flow heater, the initially provided oil flow rate from
a gear pump is limited to a defined value by a controlled valve
processing data from a flow meter.

III. CFD APPROACH

In the present study, a finite volume method is used to
a) Schematic layout
b) Numbering of discs and conductors investigate the flow behavior and temperature distribution in
in investigated passes the winding model. OpenFOAM is employed to implement the
Fig. 2. Components of a conductor model for the experimental winding numerical calculations for which the Navier-stocks equations
model [8]. and the continuity equations are solved. OpenFOAM is freely
available open source software containing several C++
As illustrated in Fig. 2, the winding model has a total of 18 libraries and applications which can numerically solve heat
discs and 18 channels. The two remaining channels and discs transfer and momentum equations. Table 1 gives two different
are post-conditioning discs which are not considered in the operation points corresponding to the mass flow rates equal to
results. To represent the inlet and exit conditions at each ṁoil = 9 kg/s, ṁoil = 18 kg/s and with reference inlet
investigated pass of the experimental winding model, several temperature of 80°C. For the determination of the Reynolds
additional discs in pass 0 are installed below the respective numbers, the reference flow rate is calculated from the
passes, as illustrated in Fig. 2. The typical paper wrapping is smallest hydraulic diameter through which the entire oil
applied around conductors in transformer windings but in the should pass, inside the vertical channel on the inner winding
presented model, solid blocks of plastic are used as an diameter side located across the oil washer. Table 1 gives the
electrical insulation, which are located only between the calculated Reynolds numbers in different operation conditions.
conductors of a disc.
TABLE 1
Fig. 3a contains a photo of the experimental winding INVESTIGATED OPERATING CONDITIONS AND RESPECTIVE REYNOLDS.
model during operation. Although this picture gives no
Operating Mass flow inlet temperature Reynolds numbers
detailed insight into the chosen mechanical design of the points rate [kg/s] [°C] [1]
winding model, comprising its sealing and insulation concept,
it offers a first impression of the required technical efforts. 1 9 80 2000
Fig. 3b shows holes for the temperature sensors filled partly 2 18 80 4000
with heat-conducting paste before the sensors were injected
and pushed towards the bottom of the drilling. The openings The behavior of the fluid flow in the operating points 1 and
were sealed then impeding oil from entering the drill holes, as 2 indicate the analysis at higher Reynolds numbers, where the
shown in Fig. 3c. Heating cartridges are employed to inject application of turbulence model is required to achieve accurate
into each conductor model through a borehole, providing the results with CFD simulations since the laminar character of
heating power normally dissipated in the winding turns of a the oil is affected by high oil flow rates with high inlet
power transformer during operation. temperature. Turbulent numerical models are defined by an

2
978-1-5090-4877-9/17/$31.00 ©2017 IEEE
 19th IEEE International Conference on Dielectric Liquids (ICDL), Manchester, United Kingdom, 25 – 29 June, 2017

identical solver setup, which applies the SST (Shear Stress consideration for modeling selected oil flow rates becomes
Transport) turbulence model developed by Menter [12] with questionable [9]. Because of the same inlet temperature in
its transitioning onset model at different operating conditions. both operating conditions, the Reynolds number of the oil
The SST turbulence model is a two-equation eddy-viscosity flow is only dependent on the flow rate and properties in same
model which has used k-ω formulation near wall regions of inlet temperatures are similar. As illustrated in Fig. 5, the
the boundary layers makes the model directly usable for the deviations of the maximum temperature gradients in 2D
viscous sub-layer, hence the SST k-ω model can be used as a simulations in the third and the second discs in 9 kg/s arise up
Low-Reynolds turbulence model without any extra damping to 50% of measurement results in simulation. Although the
functions. The possibility of switching to k-ε behavior and results with OpenFOAM have deviations with measurement
thereby avoids the common k-ω problem made SST results, but the 2D simulations predicted correctly behaviors of
formulation more accurate and reliable for the fluid flows than temperature distributions within conductors of each turn.
the standard k-w model [12].
For preparing the mesh for OpenFOAM, firstly stl-format
of geometry in Solid works is prepared. The meshing process
is comprised of five steps; the refinement is done according to
refinement levels set in SnappyHexMesh dictionary first. In
the second step, cells are snapped in order to get smooth
surfaces in five levels and, eventually in the last step, surfaces
layers are created according to the value set in the
SnappyHexMesh dictionary, and the mapFields utility maps
the results from each step of solving as initial conditions for
the next step. Therefore, for simulations with OpenFOAM, the
domain of winding is solved five times to achieve the results.
After four levels of meshing, the independency of the domain a. ṁoil = 9.0 kg/s, ϑin = 80°C and P losses, i = 360 W/turn.
on meshing is verified, so that five steps of calculations are
selected. In Fig. 4, the fine mesh of one pass of the winding in
front view is shown.

b. ṁoil = 18.0 kg/s, ϑin = 80°C and P losses, i = 360 W/turn.

Fig. 4. Computational mesh of the fluid and solid domains in Fig. 5. Comparison of the measured temperature gradients with numerically
SnappyHexMesh. determined 2D OpenFOAM results in the second pass at two oil flow rates
with specific losses per winding turn of Plosses, i = 360W/turn.

IV. RESULTS AND DISCUSSIONS B. Results of the experimental and numerical winding model
analysis
A. Temperature gradients between numerical methods and The average temperature in the winding and the location of
experimental setup the maximum temperature with the value in a pass differ quite
To verify the numerical calculations, the temperature significantly between flow rates. In Table 2, the average and
gradients between oil and conductors at all investigated maximum winding temperature with the location of hot-spot
operating conditions in the second pass are presented. Fig. 5a for the mass flow rates of 9 kg/s and the 18 kg/s are shown,
and 5b show the deviation of temperature gradients at various respectively.
flow rates. To calculate the local temperature gradient at a TABLE II. AVERAGE AND MAXIMUM WINDING TEMPERATURE WITH THE
specific conductor, the oil temperature at the inlet of the pass LOCATION OF HOT-SPOT IN THE SECOND PASS OF THE WINDING AT ṀOIL = 9.0
of that conductor is applied, which are determined KG/S, AND ṀOIL = 18.0 KG/S, ΘIN = 80°C, PLOSSES, I = 360W/TURN.
thermodynamically. The main reason of displaying
Flow
temperature gradients in comparison to absolute temperatures Rate
Model Tavg discs (°C) Thot-spot (°C) Hot-spot location
is obtaining the possibilities of related results at different inlet OpenFOAM 92.91 101.12 Disc 3
9
temperatures. Because of the thermal dependency of the oil
kg/s Experiment 88.74 93.13 Disc 4
properties, the laminar character of the oil flow is affected at
high oil temperatures and speeds. For that reason, the 18 OpenFOAM 89.16 96.80 Disc 2
presented numerical investigations include the application of kg/s Experiment 86.43 90.51 Disc 4
turbulent pattern for both flow rates since a laminar

3
978-1-5090-4877-9/17/$31.00 ©2017 IEEE
 19th IEEE International Conference on Dielectric Liquids (ICDL), Manchester, United Kingdom, 25 – 29 June, 2017

As indicated in Table 1. the predicted locations of hot-spot

in turbulent cases with OpenFOAM in comparison with
experimental results are different. The main reason for the
deviation with predicting the location is backflow in the first
few horizontal channels in 2D simulations. The mechanism for 9 kg/s
the generation of local hot-spot for high oil mass flow rates as
an example in 18 kg/s can be explained by purely
hydrodynamic arguments. The high flow velocity oil in the
horizontal channels below the washers acts as a jet into the
vertical channel and causes a separation of flow. Such
interaction prevents the oil from entering the first horizontal
channels. This flow behavior can cause a significant
overheating of the bottom disc in the passes which sometimes 18 kg/s
causes local hot-spot in the winding. In general, by 3D
simulations and experimental investigations, by incrementing
the flow rate and velocity distributions, the hot-spot
temperatures will be dropped, respectively [9].
Pass 1 Pass 2
Fig. 6 shows the velocity distributions at flow rates 9 kg/s Fig. 7. Comparison of the oil temperature distributions in pass 1 and pass
and 18 kg/s and Fig. 7 presents the oil flow temperature, for 2 within 2D CFD winding models in OpenFOAM at ṁoil = 9.0 kg/s and
both flow rates. The enhancement temperature gradient in the ṁoil = 18.0, ϑin = 80°C and P losses, i = 360 W/turn.
left vertical channel in pass 2 and reverse flow at the outlet
regions of the first horizontal channels cause also the local REFERENCES
hot-spot in the left side of pass 2. [1] P. H. G. Allen, and D. J. Allan, “Layer-type transformer winding
cooling factors derived from analogue solution of the governing
equations,” Proc. IEE, no.110, pp. 523-534, 1963.
[2] P. H. G. Allen and A. H. Finn, “Transformer winding thermal design by
computer,” Proc. IEE, no. 51, pp. 589-599, 1963. B. Smith, “An
approach to graphs of linear forms (Unpublished work style),”
unpublished.
9 kg/s
[3] L. Imre, I. Szabo, and A. Bitai, “Determination of the steady state
temperature field in naturally oil-cooled disc-type transformers,” Proc.
6th International. Heat Transfer Conf. II, Toronto, Canada, pp. 123-128,
1978.
[4] L. Imre, I. Szabo, and A. Bitai, “Thermal transient analysis of OFAF
transformers by numerical methods, numerical methods in thermal
problems: Proc. 2nd Int. Conf., R. W. Lewis, K. Morgan, and B. A.
Schrefler (Eds.), Swansea U.K., pp. 899-909, 1981.
[5] A. J. Oliver, “Estimation of transformer winding temperatures and
coolant flows using a general network method,” Proc. Inst. Elect. Eng.,
18 kg/s vol. 127, pp. 395-405, 1980.
[6] S. Szpiro, P. H. G. Allen and C. W. Richards, “Coolant distribution in
disc type transformer winding horizontal ducts and its influence on coil
temperatures” ,Proc. 7th International Heat Transfer Conf., Munich,
Germany, pp. 251-256, 1982.
Pass 1 Pass 2 [7] F. Torriano, M. Chaaban, and P. Picher , “Numerical study of
Fig. 6. Comparison of the velocity distributions in pass 1 and pass 2 parameters affecting the temperature distribution in a disc-type
transformer winding,” Applied Thermal Engineering , 30(2010)2034-
within 2D CFD winding models in OpenFOAM at ṁoil = 9.0 kg/s and
2044, 2010.
ṁoil = 18.0, ϑin = 80°C and P losses, i = 360 W/turn.
[8] E. I. Amoiralis, M.A. Tsili, and A.G. Kladas. , “Transformer design and
V. CONCLUSION optimization: a literature survey,” IEEE Transaction on Power Delivery,
vol. 24, no. 4, pp. 1999–2024, 2001.
In this contribution, 2D simulations with OpenFOAM, [9] S. Tenbohlen, C. Breuer, F. Devaux, R.Lebreton, N.Schmidt, and T.
presented to determine the oil velocity and temperature Stirl, “Evaluation of the thermal performance of transformer windings
distributions followed by an analysis of the hot -spot and its by numerical investigations and measurements”, CIGRE 2016, Paris,
A2-309, 2016.
location. Moreover, the average temperature in the second
[10] S. Tenbohlen, N.Schmidt, C. Breuer, S. Khandan , and R. Lebreton,
pass was studied. The presented windings models, which have “Optical and numerical investigation of oil flow velocities inside a
been validated by experimental results, provide useful insight zigzag cooled winding model”, ARWtr 2016, International Workshop,
into the horizontal cooling channels and the flow behavior Spain, 2016.
within the winding. The evaluation of CFD numerical [11] A. Weinläder, W. Wu, S. Tenbohlen and Z. Wang , “Prediction of the
solutions on thermal modeling of power transformers winding oil flow distribution in oil-immersed transformer windings by network
modelling and computational fluid dynamics,” IET electric power
with OpenFOAM, free open source software is aimed and applications, vol. 6, no. 2, pp. 82-90, 2012.
considered in this contribution. The results are verified with [12] F. R. Menter, “Two-Equation Eddy-Viscosity Turbulence Models for
experimental setup and the distributions of temperature. Engineering Applications,” AIAA Journal 32 (8), 1994, pp. 1598–1605.

4
978-1-5090-4877-9/17/$31.00 ©2017 IEEE

View publication stats