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

net/publication/327752627

CFD Investigation of Temperature Distributions by Non-uniform Heat Losses

inside Windings

Conference Paper · September 2018

DOI: 10.1109/ICHVE.2018.8642236

CITATIONS READS

3 213

2 authors, including:

Stefan Tenbohlen
Universität Stuttgart
378 PUBLICATIONS   3,690 CITATIONS   

SEE PROFILE

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

Particle triggered PD in HVDC GIS View project

Smart Grids View project

All content following this page was uploaded by Saeed Khandan Siar on 19 September 2018.

The user has requested enhancement of the downloaded file.

 CFD Investigation of Temperature Distributions by
Non-uniform Heat Losses inside Windings
S. Khandan, S. Tenbohlen
Institute of Power Transmission and High Voltage Technology (IEH), University of Stuttgart, Stuttgart, Germany.
saeed.khandan@ieh.uni-stuttgart.de

Abstract: this paper shows the application of (Computational distribution layer-type transformers; whereas on the oil-
Fluid Dynamics) CFD method to predict the hot-spot temperature directed cooling mode (OD), [9-11] have considered high
(HST) at different operating conditions inside the disc-type velocities to predict eddies of fluid flow inside channels and
winding. It is very well-known that electrical heat losses are not even reversed flow in horizontal channels of windings.
distributed uniformly inside windings of power transformers due In terms of experimental investigations, a comprehensive
to the flux leakages in the winding; therefore, the influences of the study of the oil flow distribution inside a disc-type winding
non-uniform heat losses in the conductors of the winding are model was carried out by Allen et al. [12]. The publications by
investigated and compared with the uniform heat losses Tenbohlen et al. [13-14] addressed the behavior of fluid flow
distribution at different mass flow rates. Moreover, experimental in horizontal ducts and consulted accurate CFD simulations to
verifications are provided to support 3D CFD simulations in the
investigate of thermal aspects of OD cooling conditions.
oil-directed (OD) cooling modes. The accurate numerical results
Zhang et al. [15-16] investigated a conventional disc-type
of CFD simulations include the average temperature in each pass,
winding model operated under ON conditions and performed
the hot-spot location (HSL) at different dimensionless eddy losses
measurements within the disc-type winding models.
factor (Qeddy). The presented results give a deep insight into the
thermal aspects of the disc-type windings enabling the designer to Determination of the lifetime in power transformers depends
optimize the cooling conditions of power transformer windings. directly on HST in winding and non-uniform heat losses
distributions have remarkable effects on HST and HSL of the
I. INTRODUCTION winding. Meanwhile, in order to have accurate investigations
Power transformers belong to the most important and and detailed insights on thermal aspects of windings,
valuable assets in power systems. The lifetime of the significant influences of non-uniform heat losses distributions
transformers mainly depends on the lifetime of its insulation. on thermal behavior are considered as a high interest. In two
Since the temperature is the dominated factor for the aging rate comprehensive contributions, the discussions on the hot-spot
of paper insulation, the large number of researches are focused factor were presented [18-19].
on the thermal conditions of power transformers. The active Two main mechanisms contributing heat within the winding
part inside the transformer insulation system developing the can be defined. First of all, load currents dissipate heat with the
maximum temperature during operation is commonly referred quadratic dependency of I2R losses on the loading rate. The
to as the hot-spot. Two factors primarily determine its location second loss mechanism relates to basic operating principles of
and temperature. First of all, the amount of locally dissipated power transformers, the transmission of electrical energy via
heat resulting from the distribution of electrical losses has to be induction linked to the electromagnetic flux. Within
taken into account. Moreover, local heat transfer characteristics electromagnetic field, leakage flux induces currents in all
resulting from cooling fluid flow distribution and thermal electrically conducting materials proportional to the loading
aspects of applied insulation need to be considered. In order to rate. The corresponding ohmic losses commonly referred to as
investigate the temperature distribution of the windings, stray losses are therefore also proportional to the quadratic
different approaches can be used. value of loading rate and are not limited to the windings and
In recent years, most of the numerical CFD contributions core. Within the windings, alternating leakage flux induces
have been focused on investigating the location of the hot-spot eddy currents perpendicular to the axial direction of conductors.
[1-3] or concentrated on the parameters affecting the In contrast to the I2R losses that rather constant per
temperature distributions and fluid flow in the windings [4-6]. conductor element, the stray losses connected to the magnetic
Besides, cooling of windings of power transformers relies on field contribute a non-uniform distribution of heat sources over
the circulation of dielectric coolant liquid between the radiators the winding. Auxiliary losses refer to the fans, pumps and
and windings. Different methods of cooling have determined to cooling equipment are not included in the total losses and
investigate the fluid behaviors. Based on the natural cooling usually eliminate in windings design [20]. In addition to our
mode (ON), [7-8] have performed the CFD solutions to previous work [14], in sections II and III in the present study, a
simulate the natural convection heat transfer within the power detailed review of the experimental procedure and the
transformers and displayed the oil flow behavior inside appropriated boundary conditions are explained. Finally, the
accurate 3D CFD results are represented to show the influence

978-1-5386-5086-8/18/$31.00@2018 European Union

of non-uniform electrical heat losses distribution over the To supply the non-uniform electrical heat losses distributed
winding model. Furthermore, the thermal aspect of the winding in the winding turns, heating cartridges are installed into each
model at different operating conditions are considered to define conductor. Heating cartridges are supplied by a PC-controlled
the temperature rise of the winding at different loading DC power supply. Fluid guides divide the winding into three
conditions. passes due to dividing the cooling fluid similarly into horizontal
channels in each pass. The main contribution of fluid guides is
II. DESCRIPTION OF AN EXPERIMENTAL DISC-TYPE WINDING
MODEL to force the oil to enter a pass through only one vertical channel
and leave the pass via the opposite channel, leading to a zig-zag
The design of disc-type transformer windings shows a flow distribution. Moreover, the fluid guides cause more
strong symmetry in circumferential direction [13]. To display consistent flow distribution inside the horizontal channels and
this design characteristics more clearly, Fig. 1 depicts a as a result, enhance the heat transfer on the disc surfaces.
magnified three-dimensional view of a segment with two main To ensure the amount of oil flow rate at the model entrance, a
investigated passes. The entire of the model consists of three digital controlled valve, and a flow meter are employed to
passes. The first pass is the pre-conditioning pass, whose task provide a specific rate of oil flow inside the model. Besides, the
is to obtain stable fluid flow for the investigation. At the top inlet oil temperature is set by a flow heater. Additionally, two
end of the model, there is a post-conditioning pass to provide adjacent sensors monitor the temperature at the entrance.
steady flow conditions at the outlet part during measurements. Moreover, the equipped temperature sensors capture the
Each pass has 6 discs and 4 turns and totally the experimental temperature inside each conductor, therefore a proper control on
winding model is designed with 20 discs. Besides, the given main operating parameters is applied to the laboratory setup.
enlargement includes the conductors and discs numberings In this study, the inlet temperature is set to ϑin = 80°C while
which are used for presenting results. The spacer geometry the inlet oil flow rate is set to ṁoil = 3 kg/s and ṁoil = 18 kg/s,
formed in order to match the corresponding width of the represent the laminar and turbulent flows for measurements
horizontal channels at the inner and outer winding diameter. and the numerical investigations. Furthermore, in all types of
cooling methods, the distribution approach of heat losses
affects directly on the thermal aspects of the winding. Since the
generated electrical heat losses inside windings are distributed
non-uniformly, detail investigations on the HTS, the HSL, and
AT in a winding model at different operation conditions cause
prospective overview for manufacturers and designers.
III. NUMERICAL METHOD
In order to solve CFD problems using numerical methods,
the domain should be discretized. In this study, the numerical
geometries of the winding model are created and discretized
using ICEM CFD software. Afterward, the discretized domains
are solved utilizing CFX. Additionally, thermal interactions are
Fig. 1. Details of the winding model including horizontal and vertical considered between fluid and solids inside the domain.
cooling channels and the numbering of the conductors for post-processing. Furthermore, the domains are discretized very fine to capture
the temperature gradients and velocity behaviors nearby the
fluid and solid interfaces. A steady-state condition without
mesh dependency is achieved by the calculation process which
no change in solutions is obtained within two subsequent time
steps during calculations. Outer walls of the winding model are
considered adiabatic; therefore, neither heat nor flow leakage is
possible through outer walls. Finally, the sufficient number of
Fig. 2. Cross section of winding model and technical dimensions. cells for the 3D CFD approach is accomplished on approx.
The main purpose of the investigated experimental winding 200×106 cells.
model in this contribution is to deliver measurement data for the Thermal boundary conditions are assigned according to the
validation of numerical results. Since the thermal aspects of the experimental setup conditions and allow the fluid properties to
winding model are of highest interest, the model allows vary by the temperatures. Comparing the uniform and the non-
temperature measurements at every conductor element to ensure uniform operational conditions, 32 W heat loss is set uniformly
identifying hot-spot and its location, equipped with a PT-100 inside each disc at the winding model, consequently, because
temperature sensor. The modification of the spacer geometry of having 4 turns at each disc, in the case of uniform heat loss
from a rectangular to a trapezoid shape is carried out distribution, 8 W/turn is integrated for each conductor. Table 1
asymmetrically towards the channel center. More details on the shows four different conditions for the distribution of heat
manufacturing processes are clarified in [13-14]. losses within the winding model of the power transformer,
tagged as dimensionless heat losses factor Qeddy.
 TABLE I
DETAILS OF HEAT LOSSES DISTRIBUTION IN DISCS GIVEN AT PASS 2 (W).
Qeddy Disc 1 Disc 2 Disc 3 Disc 4 Disc 5 Disc 6
1.0 32 32 32 32 32 32
1.5 32 35.2 38.4 41.6 44.3 48
2.0 32 47 51.2 55.4 59.1 64
2.5 32 58.7 64 69.3 73.8 80
As expected, there is an increase in heat losses at the top
discs due to the increased contribution of the radial leakage
flux at the end of the winding. In order to show the ratio of the
heat losses in different turns, L1 to L4 correspond to the inner
turn towards the outer turn of the winding which is referenced
in Fig. 1; accordingly. Table 2 shows the distribution of losses
over the conductors of discs in the radial direction in the last
pass at Qeddy = 2. It is notable that not only the heat losses
increase at the top pass of the winding but also the conductors
in the same discs experience different heat losses due to the
different radial heat flux densities [18]; [19]. a. ṁoil =3 kg/s, ϑin = 80°C, pass 2
TABLE II Fig. 3. Comparison of the measured temperature gradients with numerically
RATIO OF THE HEAT LOSSES COMPARED WITH UNIFORM OPERATION CONDITION
determined 3D CFD results at the different conditions and Qeddy factors.
FOR EACH TURNS AT Q = 2.

L1 L2 L3 L4
Q ave, top disc / Q ave 1.81 1.87 2.05 2.27
Q ave, 2 top discs/ Q ave 1.75 1.81 1.98 2.17
Q ave, 3 top discs/ Q ave 1.68 1.73 1.88 2.09
Q ave, 4 top discs/ Q ave 1.64 1.68 1.82 2.01
Q ave, 5 top discs/ Q ave 1.58 1.63 1.78 1.96
Q ave, 6 top discs/ Q ave 1.48 1.52 1.65 1.81
a. Pass 2, Q=1 b. Pass 2, Q=1.5
IV. RESULTS AND DISCUSSIONS
Numerical CFD simulations were conducted with the
specified boundary conditions according to the designed
experimental setup to give detail insight into the thermal
behavior of the winding in power transformer. At the first
prior, this section proposes to discuss the validation of the CFD
results and then illustrates the thermal behavior of the winding c. Pass 2, Q=2 d. Pass 2, Q=2.5
Fig. 4. Temperature distributions determined with the 3D CFD winding
model and the HST and the HSL at different eddy losses
model at ṁoil =3 kg/s, ϑin = 80°C, and different Qeddy factors.
factors.
To validate the CFD numerical results, Fig. 3 depicts the
temperature gradients between oil and conductors by non-
uniform and uniform heat losses distribution. It should be
noted that the inlet temperature of the cooling oil at each pass
is calculated thermodynamically. Moreover, the temperature
gradients are based on the difference between the volume
averaged local temperature at each conductor and the
a. Pass 2, Q=1 b. Pass 2, Q=1.5
calculated temperature at the inlet of the considered pass.
The comparisons of the measurements and 3D CFD results
predict the temperature gradients at the considered pass in the
winding by the accuracy of less than 2 K deviations with
measurements. In Fig. 4 and Fig. 5, temperature distributions
determined with 3D CFD results at two different inlet flow
rates are illustrated.
It is noticeable that the rate of the oil fluid flow has directly c. Pass 2, Q=2 d. Pass 2, Q=2.5
influence the cooling condition of the conductors. By the non- Fig. 5. Temperature distributions determined with the 3D CFD winding
uniform heat losses, despite the higher heat losses at the top model at ṁoil =18 kg/s, ϑin = 80°C, and different Qeddy factors.
disc of the winding, HSL locates at the middle region of the Table 3 shows the detailed view of the pass 2 for both
pass (disc 3 or disc 4). Due to having a lower share of the uniform and non-uniform heat losses distribution by the flow
cooling oil in the lower channels, the average temperature of rates of 3 kg/s at the inlet temperature of 80°C. Based on the
the middle discs of the winding model is higher. different Qeddy factors, the temperature distributions at different
operating points are illustrated. It is worthwhile to note that [3] A. Weinlaeder and S. Tenbohlen, “Thermal-hydraulic investigation of
heat losses directly cause higher AT and the HST, respectively. transformer windings by CFD-modelling and measurements”, 16th ISH,
South Africa, 2009.
The AT and the HSL are quite different between each [4] J. Mufuta and E. Van, “Modeling of the mixed convection in the
operating conditions. It can be seen that the predicted windings of a disc-type power transformer”, International Journal of
temperatures have a good agreement with the measured values. Applied Thermal Engineering, vol. 20, pp. 417-437, 2000.
Additionally, the conductors nearby the vertical channels [5] E. Rahimpour, M. Barati, and M. Schafer, “Investigation of parameters
affecting the temperature rise in windings with zigzag cooling flow path”,
have a bigger surface for cooling, therefore by the high ratio of Applied Thermal Engineering, vol. 27, no. 11, pp. 1923–, 2007.
the heat loss, HSL is not located at the last outer turn of the [6] F. Torriano, M. Chaaban and P. Picher, “Numerical study of parameters
winding model. It should be considered that although ATs affecting the temperature distribution in a disc-type transformer
increase at different operating conditions, the HSLs are placed winding,” International Journal of Applied Thermal Engineering, vol.
30, pp. 234-244, 2010.
at disc 3 or disc 4 by the different Qeddy factor. Therefore, the [7] J. Gastelurrutia, J.C. Ramos, G.S. Larraona, A. Rivas and J. Izagirre,
HST depends on the cooling conditions of the pass and the “Numerical models of natural convection of distribution
ratios of the heat losses has no remarkable effect on the HSL. transformers” ,International Journal of Applied Thermal Engineering
Vol. 31, pp. 493-505, 2011.
TABLE III [8] X. Zhang, Z. Wang, Q. Liu, M. Negro, A. Gyore and P. Smith,
DETAILS ABOUT TEMPERATURE DISTRIBUTION AND HOT-SPOT LOCATION IN “Numerical investigation of influences of liquid types on flow
NON-UNIFORM AND UNIFORM HEAT LOSSES DISTRIBUTION AT MASS FLOW RATE distribution and temperature distribution in disc type ON cooled
3KG/S IN PASS 2. transformers”, International IEEE conference on dielectric liquids,
Qeddy=1 Qeddy =1.5 ICDL 2017, UK.
Operating [9] E.J. Kranenborg, C.O. Olsson, B.R. Samuelson, L/A. Lundin, and R.M.
HST HSL AT HST HSL AT Missing, “Numerical study on mixed convection and thermal streaking
Condition
(°C) (°C) (°C) (°C) (°C) (°C) in power transformer windings”, Proc. of the 5th European Thermal-
Measured 98.9 Disc3 93.7 103 Disc4 96 Sciences Conference, The Netherlands, 2008.
CFD 99.4 Disc4 94.8 104 Disc4 96.7 [10] A. Skillen, A. Revell, H. Lacovides and W. Wu, “Numerical prediction
Qeddy =2 Qeddy =2.5 of local hot-spot phenomena in transformer windings”, International
Operating Journal of Applied Thermal Engineering, vol. 36, pp. 96-105, 2012.
HST HSL AT HST HSL AT
Condition [11] X. Zhang, Z. D. Wang, and Q. Liu, “Prediction of pressure drop and flow
(°C) (°C) (°C) (°C) (°C) (°C)
Measured 107 Disc4 97.7 112 Disc4 99.8 distribution in disc type transformer windings under an OD cooling
CFD 108 Disc4 98.4 113 Disc4 88.8 mode”, IEEE Transactions on Power Delivery, vol. 32, no. 4, pp. 1655–
1663, 2017.
V. CONCLUSION [12] P.H.G. Allen, O. Szpiro, and E. Campero, “ Thermal analysis of power
transformer winding”, Electric Machines and Power Systems, 6(1):1-
In this contribution, the thermal behavior of the disc-type 11.1981.
winding of the power transformers are investigated and the [13] S. Tenbohlen, C. Breuer, F. Devaux, R.Lebreton, N. Schmidt, and T.
Stirl, “Evaluation of the thermal performance of transformer windings
influence of the non-uniform heat losses on the temperature by numerical investigations and measurements”, CIGRE 2016, Paris,
distributions, and the hot-spot locations are considered. 3D A2-309, 2016.
CFD winding model are explained and the CFD results are [14] S. Tenbohlen, N. Schmidt, C. Breuer, S. Khandan, and R. Lebreton,
validated by the measurements. The presented winding model “Investigation of thermal behaviour of an oil directed cooled transformer
winding”, Vol. 33, No. 3, 1091-1098, 2017.
provide a detailed insight into the thermal conditions of the [15] J. Zhang, X. Li, “Coolant flow distribution and pressure loss in ONAN
conductors. Accurate numerical simulations are mainly transformer winding-part I: Theory and model development”, IEEE
proposed whose deviations from measurements are acceptable. Transactions on Power Delivery, Vol. 19, No. 1, 186-193, 2004.
The investigations show the increment of ATs and HSTs at the [16] J. Zhang, X. Li, “Coolant flow distribution and pressure loss in ONAN
transformer winding-part II: Optimization of design parameter”, IEEE
different operating conditions due to the non-uniformly heat Transactions on Power Delivery, Vol. 19, No. 1, 194-199, 2004.
losses distributions compare to the uniform heat losses [17] M. Daghrah, Z. Wang, Q. Liu, D. Walker, C. Krause and G. Wilson,
distributions by the OD cooling mode. It is observed that the “ Experimental investigation of hot-spot factor for accessing hot-spot
radial leakage fluxes at the ends of the winding increased at temperature in transformers”, International Conference on Condition
Monitoring and Diagnosis, CMD 2016, 2016.
this region which have a significant influence on the AT and [18] J. Lapworth, P. Picher, J. Channet, and et al. “Cigré brochure 659:
the HST in the winding model. The numerical results Transformer thermal modelling”, CIGRé Working Group A2.38, A2.38,
investigates the thermal behavior of the winding model at Tech. Rep., 2016.
different operating conditions indicate proper predictions by [19] Z. Radakovic, U. Radoman, and P. Kostic, “Decomposition of the Hot-
spot factor,” IEEE Transactions on Power Delivery., vol. 30, pp. 403-
numerical deep considerations on the electrical heat losses. 411, 2015.
VI. REFERENCES [20] M. Heathcote, J & P transformer book, Thirteenth ed.: Newnes, 2007.

[1] S. Khandan, S. Tenbohlen, C. Breuer, and R. Lebreton, “CFD study of

fluid flow and temperature distributions in a power transformer
winding”, International IEEE conference on dielectric liquids, ICDL
2017, UK, available online, DOI: 10.1109/ICDL.2017.8124674.
[2] X. Zhang, Z. D. Wang, and Q. Liu, “Interpretation of Hot-spot factor for
transformers in OD cooling modes”, IEEE Transactions on Power
Delivery, available online, 2017.

View publication stats