Investigation of Thermal Behavior of An Oil Directed Cooled Transformer Winding

See discussions, stats, and author profiles for this publication at: https://www.researchgate.
net/publication/317352452
Investigation of Thermal Behavior of an Oil Directed Cooled Transformer

Winding
Article in IEEE Transactions on Power Delivery · June 2017

DOI: 10.1109/TPWRD.2017.2711786
CITATIONS READS
18 547
5 authors, including:
Stefan Tenbohlen Raphael Lebreton

Universität Stuttgart General Electric
378 PUBLICATIONS 3,689 CITATIONS 7 PUBLICATIONS 33 CITATIONS
SEE PROFILE SEE PROFILE
Some of the authors of this publication are also working on these related projects:
Particle triggered PD in HVDC GIS View project
Frequency Response 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.

IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 00, NO. 00, 2017 1
1 Investigation of Thermal Behavior of an Oil-Directed

2 Cooled Transformer Winding
3 Stefan Tenbohlen, Senior Member, IEEE, Nicolas Schmidt, Christian Breuer, Saeed Khandan, and Raphael Lebreton
4 Abstract—In this contribution, the oil speed in horizontal chan- determine the hot-spot value. A Literature survey based on IEEE 39
5 nels of an oil-directed cooled winding is investigated by experi- Transactions in [1] revealed the continued interest in the appli- 40
6 mental and numerical methods. The presented winding model of- cation of advanced techniques for transformers design optimiza- 41
7 fers insight into the horizontal cooling channels perpendicular to
8 the main oil-flow direction. To visualize oil flow, tracing particles tion. For the investigation of the temperature distribution within 42
of
9 were added to the cooling oil and illuminated by high-power light- the windings of an OD cooled transformer, different approaches 43
10 emitting diodes. The particle velocities were then determined by can be pursued [2]. 44
11 taking photographs with a defined exposure time. The design of The importance of winding hotspot determination is reflected 45
12 the sophisticated winding model is described in the contribution. in published efforts to predict the hot-spot temperature in the 46
13 In addition to the experimental results, this contribution presents
14 a comparison with respective numerical results from 2-D and 3-D windings. The thermal-hydraulic networks (THN) can be used 47
15 computational fluid dynamics calculations. Finally, numerical re- for describing the entire oil circuit in order to investigate natural 48
16 sults from the winding model concerning the oil-flow distribution circulations [3]–[5]. However, it has been shown in the literature 49
Q1
17
18
19
20
21
22
23
24
inside the winding at various operating conditions are presented.
enabling the designer to optimize the cooling of the power trans-

former windings.
Index Terms—Power transformer, oil-directed cooling, optical
investigation, thermal modelling, computational fluid dynamics,
oil-flow rate.
ro
The investigation indicated a strong nonuniform oil-flow distribu-
tion on the horizontal channels. The presented results give a deep
insight into the oil flow and temperature behavior of windings,
that comparing with the THN, the CFD method provides better
accuracy in prediction of the temperature distributions and fluid
behaviors [6]–[8].
Thermal modeling offers a detailed view of temperature dis-
tribution over the whole winding and temperature values at the
characteristic points. The originally developed algorithm ther-
mal modeling network for temperature calculations is based on
characteristic temperatures.
50
51
52
53
54
55
56
57
EP
25
An efficient thermal design offers the opportunity to signifi- 58
26 I. INTRODUCTION cantly reduce costs in the production process and to improve the 59
OWER transformers are key components applied in trans- thermal performance [9]. Due to the numerical simulations with 60
27
28
29
P mission and distribution systems. Their power rating and
life cycle characteristics are strongly dependent on thermal as-
CFD techniques, a better comprehension of the cooling mecha-
nism of core type transformers has developed to high levels.
61
62
30 pects. The higher the evolving temperature levels at a given In recent years, most of the numerical models are considered 63
31 loading rate are settling, the faster the insulation materials will in 2D approach. El Wakil et al. [10] have performed a 2D sim- 64
32 age at that loading rate. Therefore, a profound knowledge about ulation of a step-down 3-phase power transformer that included 65
33 the temperature distribution inside a transformer is crucial for the core, primary and secondary windings using a commercial 66
34 an appropriate assessment of the component’s power rating and finite volume code. Skillen et al. have [11] investigated on a 2D 67
IEE
35 for that reason offers the opportunity to cut costs in the pro- axisymmetric model the local hot-spot in detail within the low- 68
36 duction process. In the past, and still today to some extent, voltage winding and predicted with CFD simulations the hot 69
37 analytical and empirical methods have been used to predict the streak locations. Mufuta et al. [12] have presented 2D results of 70
38 temperature distribution in the transformer components and to the flow pattern as a function of the Reynolds and Grashof num- 71
bers as well as of geometric parameters. Gastelurrutia et al. [13] 72
have performed the CFD solutions for an ONAN distribution 73
Manuscript received December 19, 2016; revised March 14, 2017; ac-
cepted May 20, 2017. Date of publication; date of current version. Paper no. layer-type transformer. Their model simulated the natural con- 74
TPWRD-01484-2016. (Corresponding author: Stefan Tenbohlen.) vection behavior within the power transformers and displayed 75
S. Tenbohlen, N. Schmidt, and S. Khandan are with the Institute of the oil flow behavior inside distribution layer-type transformers. 76
Power Transmission and High Voltage Technology, University of Stuttgart,
Stuttgart 70569, Germany (e-mail: stefan.tenbohlen@ieh.uni-stuttgart.de; A 2D CFD model is used also to recognize the significant ef- 77
nicolas.schmidt@ieh.uni-stuttgart.de; saeed.khandan@ieh.uni-stuttgart.de). fects of buoyancy term and hot oil streak formations, the latter of 78
C. Breuer is with the R&D Engineer Power Transformers Department, GE which could more or less worsen downstream oil temperatures 79
Grid Solution, 41065 Mönchengladbach, Germany (e-mail: christian.breuer@
ge.com). [14]. Dimensional analyses are conducted in 2D OD cooled 80
Q2
R. Lebreton is with the R&D Engineer Power Transformers Department, GE winding model to identify the independent dimensionless vari- 81
Grid Solution, Puteaux 92900, France (e-mail: raphael.lebreton@ge.com). ables that effect pressure drops in the windings [15]. Torriano 82
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org. et al. [16] have focused on the parameters affecting on the tem- 83
Digital Object Identifier 10.1109/TPWRD.2017.2711786 perature distributions and fluid flow in the winding. 84
0885-8977 © 2017 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.
2 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 00, NO. 00, 2017
Fig. 1. Basic winding design and alignment of spacers creating symmetrical

sections with periodic cooling channel system [20].
85 Besides the 2D approach for simulating thermal behaviors of

86 windings, 3D approach is also employed to determine the tem-
of
87 perature distributions and fluid flow in recent CFD calculations.
88 A low-voltage winding has been simulated and the results show
89 that the performances of the 3D numerical computations are sig-
90 nificantly better than that of the 2D models [17]. Smolka et al.
Fig. 2. Design characteristics of the modelled winding, geometry and the
91 [18] have presented full geometry CFD analyses coupled to elec- winding model [20]. a) Top view on winding segment. b) Cross section of wind-
92 tromagnetic simulations to examine the specific power losses ing model. c) Cross section of winding model with given channel dimensions
93 within the coil and core of a dry-type transformer. Campelo and insulation thickness applied between conductors resembling thermal impact
94
95
96
97
98
99
100
ro
et al. [19] have predicted for the OD cooling modes with high
flow velocities, a stagnant flow or even reverse flow can be
observed in the first horizontal channels of the pass. A compre-
hensive review of the thermal modeling in power transformers,
state of the art as well as a proper literature review has become
available in [9].
In this contribution, the oil speed in horizontal ducts of an OD
of electrical insulation.
enlargement shows the horizontal cooling channels are created

by the opening between two consecutive spacers and discs. The
vertical cooling channels are bounded by the cardboard cylin-
ders and by the sticks aligning the spacers. Consequently, the
entire winding geometry can be subdivided into symmetrical
130
131
132
133
134
EP
101 cooled winding is investigated by experimental and numerical
sections. For reasons of an improved visibility concerning the 135
102 methods.
cooling system inside the winding, the outer cardboard cylinder 136
103 Representatively for often used winding types, a so-called
is displayed in a partly transparent color. 137
104 “zigzag” arrangement of a disc-type winding was investigated.
The investigated operating conditions are defined by the tem- 138
105 In such a case, winding discs are layered above each other,
perature of the oil entering the model, the oil flow rate passing 139
106 while so-called spacers keep the axial distance between the
the winding model and the losses supplied to each conductor 140
107 discs and determine the height of the horizontal ducts for the
model. However, since the latter two properties refer strictly to 141
108 oil. The sticks ensure the proper fixation of the spacers and
a single symmetrical section, they are not easy to put into per- 142
109 keep the radial distance between the discs and the outer cover.
spective of common transformer design rules. Since common 143
110 The vertical ducts are formed by the space between discs and
transformer design rules usually refer to entire windings, the 144
111 outer cover. The oil is led from the bottom into the vertical
properties given in this contribution, to reference investigated
IEE
145
112 ducts and flows from there upwards, while it distributes into
operating conditions, are transferred accordingly. For example, 146
113 the particular horizontal ducts. To ensure a proper distribution
instead of specifying the heating power per conductor model 147
114 of the oil to the horizontal ducts, which should be as equal
representing a section of 8° in the circumferential direction of 148
115 as possible, the vertical duct is intermitted alternating between
on entire winding turn, the corresponding losses per complete 149
116 the inner and the outer duct after a specified number of discs
winding turn are referenced. Analogous considerations are ap- 150
117 by so-called washers. This leads to an oil flow in a “zigzag”
plied for the oil flow rate. For that reason, the actually applied 151
118 manner.
heating power per conductor model and the oil flow rates pass- 152
119 In order to confirm the suitability of a numerical 3D winding
ing the winding models are 8°/360° = 1/45 times the referenced 153
120 model to predict the realities in transformer windings accurately,
values pointed out in the figure captions. 154
121 a comparison between experimentally determined oil flow ve-
A top view of a symmetric section of this chosen winding de- 155
122 locities with corresponding numerical data is presented in the
sign including all relevant geometric details to assess this design 156
123 paper.
is shown in Fig. 2. Since the exact reproduction of this geome- 157
try in Fig. 2(a) imposes enormous challenges for the connected 158
124 II. DEVELOPMENT OF AN EXPERIMENTAL MODEL FOR A
manufacturing processes, the respective design of the experi- 159
125 DISC WINDING mental model was modified. For that purpose, the curvature of 160
126 The design of disc type transformer windings shows a strong the conductors in circumferential direction is neglected, as dis- 161
127 symmetry in the circumferential direction and can be subdivided played in Fig. 2(b). To eliminate the detrimental influences of 162
128 into symmetrical sections. Fig. 1 shows a three dimensional neglecting the curvature in 8°, all the dimensions are selected 163
129 view on a disc type winding with four turns per disc. The given according to the symmetrical section. 164
TENBOHLEN et al.: INVESTIGATION OF THERMAL BEHAVIOR OF AN OIL-DIRECTED COOLED TRANSFORMER WINDING 3
Fig. 4. Piping and instrumentation diagram of the laboratory setup designed

to create strictly controlled boundary conditions inside the winding model [20].
of
Fig. 4 depicts schematically the designed experimental setup 202
in which the flow meter, the flow heater, the gear pump and a 203
Fig. 3. Components of a conductor model for the experimental winding model control unit for supplying power to each heating cartridges are 204
[20]. a) Schematic layout of experimental winding model. b) Experimental employed for obtaining the boundary conditions precisely. It 205
winding model in tank during operation. allows an accurate control over the mass flow rate ṁoil , the oil 206
165
166
167
168
169
170
171
ro
The paper wrapping typically applied around conductors in
transformer windings for purposes of an electrical insulation is
substituted by solid blocks of plastic (Polyvinylidenefluoride)
located only in between the conductors of a disc. Prior experi-
mental investigations showed a great sensitivity to the evolving
temperature levels on the exact thickness of the applied paper
wrapping. Unfortunately, this property, especially with regard to
temperature at the model inlet ϑin and the heating power of each
conductor Plosses, i . 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 flowmeter. Furthermore, two temperature
sensors are employed at inlet of the winding in order to guarantee
that there is no deviation between the oil temperature provided
by the flow heater sensor and that entering the winding model.
207
208
209
210
211
212
213
214
EP
172 the comparably short experimental conductor models, is difficult To minimize the heat leakage from the winding model to its 215
173 to keep within small tolerances. To eliminate respective detri- surroundings, the entire model is submerged in an open experi- 216
174 mental influences but still allow a realistic thermal decoupling mental tank filled with oil of the same temperature as provided 217
175 of the conductor models in one disc, the paper is substituted, by the flow heater. Due to the chosen laboratory setup, a total 218
176 as illustrated in detail in Fig. 2(c). In this contribution the inlet measuring tolerance concerning the temperatures of approxi- 219
177 temperature is set to ϑin = 80 °C while the oil flow rates are set mately 0.5 K is achieved. Consequently, temperature gradients 220
178 to ṁoil = 3 kg/s, ṁoil = 9 kg/s and ṁoil = 18.0 kg/s. The can be determined with an accuracy of around 1 K. 221
179 losses per winding turn are kept at Plosses, i = 360 W/turn. The occurring deviations concerning the mass flow measure- 222
180 To provide the heating power normally dissipated in the ment and heating power control are below 1 %. Because oil is a 223
181 winding turns of a power transformer during operation, heating transparent liquid, the investigation of the oil flow requires addi- 224
cartridges are injected into each conductor model through a tional measures. For that purpose, tracing particles were added
IEE
182 225
183 borehole. An accurate measurement of the conductors’ temper- to the oil of the experimental setup. 226
184 ature is accomplished with temperature sensors located inside These particles were chosen according to the material prop- 227
185 each conductor. The cooling system of OD cooled disc windings erties of the mineral oil and the application conditions of the 228
186 is usually subdivided into so-called passes in the axial direction. investigation. To make the particles visible, a strong source of 229
187 By periodical placement of washers inside the vertical chan- light has to be applied. High power LEDs were mounted di- 230
188 nels alternating between the inner and outer winding diameter, rectly in front of the horizontal channels in experimental setup, 231
189 two different types of passes result. Therefore, the experimental as indicated in Fig. 5. 232
190 winding model must consist of at least two consecutive passes. A digital camera was then mounted in a defined alignment of 233
191 In addition, representative entering and exiting conditions are the horizontal channels and focused on the investigated area in 234
192 required before and after the investigated passes. This ensures the middle of the horizontal channels. Due to the chosen optical 235
193 thermal and hydraulic conditions that are representative for ev- equipment and the connected setup, the depth of field defin- 236
194 ery pass throughout the complete cooling system of the modelled ing the investigated areas in the z-Dimension shown in Fig. 5 237
195 winding. Fig. 3(a) shows the chosen winding model layout in a are limited to approx. 6 mm. The particle velocities within the 238
196 schematic view as well as a clear reference for the pass, disc and investigated area were then determined by taking photographs 239
197 conductor numbering upon which the presented post analysis re- applying a defined exposure time. Firstly, a reference picture of 240
198 lies. Next to a representative winding design, also the operating a glass spacer is taken to create common scale for one measure- 241
199 conditions of the modelled winding are chosen accordingly with ment set, then the camera is aligned in front of the first channel 242
200 respect to the winding type and dimensions. Fig. 3(b) contains of pass and the optical focus is set between the solid spacer 243
201 a photo of the experimental winding model during operation. and the glass spacer. Depending on the oil flow velocity in the 244
of
Fig. 7. Modelling domains of the numerical winding models created for
CFD-analysis [20].
III. CFD MODEL 267
Fig. 5. Experimental setup of the optical analysis [23]. The development of numerical winding models suitable for 268
CFD analysis first of all comprises the definition and creation 269
ro of respective modelling domains [3]. In case of a 3D winding

model, the common design of disc windings allows a significant
reduction of the winding geometry in circumferential direction,
as carried out for the experimental winding model described
above. In addition, the symmetric alignment of the spacers in-
side a symmetrical winding section, offers potential for further
reduction by application of symmetric boundary conditions, as
indicated in Fig. 7.
270
271
272
273
274
275
276
277
EP
The chosen modelling domain of the 2D winding model, al- 278
Fig. 6. Example of a photograph of particle tracks with contrast enhance-
ment and enlargement of the investigated area in the taken pictures and further
lowing a significant reduction of the connected computational 279
enlargement of one particle track [23]. efforts, requires further explanation. Since ANSYS CFX does 280
not feature a dedicated 2D solver, even the analysed 2D ge- 281
245 respective channel, an appropriate setup for the LED power level ometry must consist of 3D finite volumes. However, since only 282
246 and the chosen shutter are determined. It should be noted that for one layer of 3D cells is necessary, the detrimental consequences 283
247 each pass during the measurement, distance between the camera resulting from this requirement are negligible. On the contrary, 284
248 and the model are kept constant. After measuring the particle modelling the winding geometry with a 2D domain can even 285
249 track lengths in the taken pictures, the particle velocities can be take advantage of this restriction by creating a 2D modelling 286
250 calculated with the relation as follows: domain featuring a non-uniform cell depth in third dimension 287
[21]. Therefore, the widening of the horizontal cooling channels
IEE
288
Local velocity of particle vi = (Length of path line of from inner to outer winding diameter, as displayed in Fig. 7, can 289
particle i li )/ (Selected shutter speed Δt) (1) also be acknowledged in the 2D winding model. The chosen 2D 290
cell depths represent a defined fraction of the existing opening 291
251 As demonstrated in Fig. 6, the analysed area is also limited widths of the horizontal cooling channels. 292
252 in the y-Dimension and in the x-Dimension to the exit location Nevertheless, due to the restriction of the 2D approach in 293
253 of the oil inside the horizontal channels. CFX, it does not account for influence of side surface of spac- 294
254 For that reason, only particles inside a volume of approxi- ers. Since the spacers are covering a certain area of the heat- 295
255 mately 20 mm × 2 mm × 6 mm (x × y × z) were analysed transferring surface on the conductors, the numerical model 296
256 within each channel. This restriction is necessary since the assumes no heat transfer at these areas. 297
257 oil flow velocities are not expected to be homogeneously dis- After the described choice of the respective modelling do- 298
258 tributed. Consequently, the investigated area should be as small mains, the creation of the CFD winding models comprises the 299
259 as possible. However, since the velocities might also fluctuate discretization process including the determination of an appro- 300
260 over time due to effects of turbulences, a certain number of priate meshing density. Next to the global mesh size, special 301
261 particle tracks need to be considered for each investigated area. attention has to be paid towards the discretization of the bound- 302
262 Since the number of particles captured on photograph inside a ary layers. The total number of hexahedral meshes employed for 303
263 certain area is directly dependent on the size of the investigated discretization of the 2D and 3D domains are approx. 7 × 106 304
264 area, a compromise between restricting the investigated area cells and approx. 200 × 106 , respectively. Due to the connected 305
265 and the resulting number of available particle tracks recorded computational efforts, only the analysis of the 2D domain should 306
266 for post analysis had to be made. be considered for state of the art desktop workstations. For a 307
TABLE I
INVESTIGATED OPERATING CONDITIONS AND RESPECTIVE REYNOLDS
Operating Mass flow inlet temperature Reynolds numbers

points rate [kg/s] [°C] [1]
1 3 80 667
2 9 80 2000
3 18 80 4000
of
Fig. 9. Distribution of the determined oil flow velocities in pass 1 at ṁ o il =
18.0 kg/s, ϑ in = 80 °C and P lo sse s, i = 360 W/turn [23].
ro
Fig. 8. Distribution of measured velocities in channel number 6 of pass 1 at
ṁ o il = 9.0 kg/s, ϑ in = 40 °C and P lo sse s, i = 360 W/turn with illustration
of an idealized distribution according to a Gaussian distribution function [23].
EP
308 more efficient analysis following the described approach, high
309 performance computing centers become advisable.
310 Table I gives three different operation points corresponding to
311 the mass flow rates equal to ṁoil = 3 kg/s, ṁoil = 9 kg/s and
Fig. 10. Comparison of measured with 2D and 3D numerical calculated
312 ṁoil = 18.0 kg/s with reference inlet temperature of 80 °C. mid channel velocities in pass 1, ṁ o il = 9.0 kg/s, ϑ in = 80 °C, P lo sse s, i =
313 For the determination of the Reynolds numbers, the reference 360 W/turn.
314 flow rate is calculated at the smallest hydraulic diameter through
315 which the entire oil should pass, inside the vertical channel on
316 the inner winding diameter side located across the oil washer. defined operating conditions in the context of an idealized Gaus- 336
317 Table I gives the calculated Reynolds numbers for different sian distribution function. Since this distribution function ap- 337
pears suitable for the collected measurement data, it is applied
IEE
338
318 operation conditions.
319 Operating point 1 indicates the analysis of the heat transfer for the post-processing of the experimental results. Fig. 9 gives 339
320 at a low Reynolds number. The behavior of the fluid flow in the the results at specified operating conditions in pass 1. Next 340
321 low Reynolds number is laminar, whereas operating points 2 and to the averaged, minimum and maximum values the displayed 341
322 3 indicate the analysis at higher Reynolds numbers, where the boxes illustrate the amount of scattering experienced during 342
323 application of turbulence model is required to achieve accurate post-processing the measurements. For that purpose, the boxes 343
324 results with CFD simulations since the laminar character of the enclose the range of measured values comprising 80% of all 344
325 oil is affected by high oil flow rates with high inlet temperature. conducted measurements according to a Gaussian distribution. 345
326 Turbulent numerical models are defined by an identical solver It can be noted, that especially the first and last channel show 346
327 setup, which applies the SST (Shear Stress Transport) turbu- a wider distributed velocity distribution within the investigated 347
328 lence model [22] with its transitioning onset model at different area. While the experimental determination of the oil flow dis- 348
329 operating conditions. tribution is extensive, the corresponding numerical data can be 349
extracted from the CFD result files. Fig. 10 shows the compar- 350
ison of the experimental results and the 2D and 3D numerical 351
330 IV. RESULTS OF THE OPTICAL ANALYSIS REGARDING
post analyses at various operating conditions in pass 1 of the 352
331 OIL VELOCITY winding model. For comparing both approaches, measured data 353
332 After determining the particle velocities in a certain chan- are compared with both numerical models. 354
333 nel at specific operating condition, the gathered data is fur- The respective numerical data at middle cross-section in the 355
334 ther post-processed. To illustrate this process, Fig. 8 shows the 3D winding model of horizontal channels are given as well. 356
335 distribution of measured particle velocities in channel 6 at the Especially in the first channels the analysed location is of great 357
of
Fig. 11. Oil flow distribution onto horizontal channels in pass 1, ϑ in = 80 °C,
P lo sse s, i = 360 W/turn.
ro Fig. 13. Comparison of the oil flow velocities within the 2D and 3D CFD wind-
ing model at ṁ o il = 18.0 kg/s, ϑ in = 80 ◦ C and P lo sse s = 360 W/turn
[20]. a) 2D model (d in on left side). b) 3D model (d o u t in front) inner view.
c) 3D model (d in in front) outer view.
region of the passes. It gets obvious that separation eddies are

blocking the entry of the horizontal channels.
380
381
EP
V. LOCATION OF HOT SPOT 382
One central objective of this contribution is the assessment 383

of the pursued numerical modelling approaches for transformer 384
windings. Since 3D winding models require significantly higher 385
computational efforts in comparison to the 2D approach, it is of 386
keen interest if those additional efforts are justified. To start the 387
Fig. 12. Streamlines at a flow rate of ṁ o il = 9.0 kg/s.
respective comparison between the 2D and 3D winding model, 388
Fig. 13 shows the numerical results for the oil flow distribution 389
358 influence on the determined velocity. This can be attributed to inside the winding models for an oil flow rate of 18.0 kg/s 390
359 separation eddies at the duct entrance (see Fig. 12). qualitatively. 391
The measured mean values agree impressively well with the Although no quantitative comparison of the oil amount pass-
IEE
360 392
361 numerical results in middle position at all operating conditions. ing the different horizontal channels is carried out, the chosen 393
362 Since the thermal investigation shows great agreement between visualization already gives a clear insight into the respective dif- 394
363 experimental and numerical results, the respective results for ferences resulting from the chosen modelling approach. While 395
364 the oil flow velocities are also in good agreement. The numer- the 2D winding model by definition suggests a homogeneous 396
365 ical results for the oil flow distribution are consequently also oil flow distribution in the circumferential direction, the 3D re- 397
366 assumed to be valid. The velocity is proportional to the oil mass sults clearly contradict this assessment. Especially the visible oil 398
367 in the channels. By calculation of the average of the velocities in flow in the vertical channels of the 3D winding model displays 399
368 each channel, the shares of oil flow on the horizontal channels a strong heterogeneous character. 400
369 are determined at three different flow rates as shown in Fig. 11. However, because the oil flow distribution is only of sec- 401
370 The high velocity of the oil after passing the washer (see ondary interest in comparison to the temperature distribution, 402
371 also Fig. 13(a)) is the reason for the formation of the separation Fig. 14 gives the results of the respective temperatures within 403
372 eddies. The oil flow in the lower ducts is becoming very small the investigated winding geometry for pass1 of the 3D winding 404
373 and can even turn to backflow. In order to investigate the reason model for an oil flow rate of 9.0 kg/s. In the investigated arrange- 405
374 for the uneven fluid flow distribution in Fig. 11, Fig. 12 shows ment the location of the hot spot is in the middle two discs (3 and 406
375 the streamlines of the oil flow for a flow rate of 9 kg/s. The results 4), which is in contradiction with the common thinking that the 407
376 of the CFD simulation indicate that the flow distributions on the hot spot can be found on top of the winding. The reason for this 408
377 horizontal oil channels are not equally. For low flow rate of behavior is the non-uniform distribution of the oil flow as seen 409
378 3 kg/s, the maximum velocity occurs in the lower part of the in Fig. 11. The oil velocity in the two last horizontal channels 410
379 pass, whereas for high winding flow rates shifts to the upper is much higher as in the four lower channels which results in a 411
of
ro
EP
Fig. 14. Temperature distribution at ṁ o il = 9.0 kg/ s, ϑ in = 80 °C and
412 better cooling of these discs. Even the turns within one disc do
413 not have the same temperature due to the thermal resistance of
414 the insulation material between the turns. To better quantify the
415 deviations connected to the numerical modelling setup, Fig. 15
416 presents the results for the temperature gradients between oil
and conductors at all investigated operating conditions in the
IEE
417
418 second pass of the winding model.
419 Next to the CFD results, the shown data also contains the
420 measurements, allowing a proper assessment of the respective
421 modelling approach. To calculate the local temperature gradi-
422 ent at a specific conductor, the oil temperature at the inlet of
423 the pass holding that conductor is applied. In case of the mea-
424 surements, oil temperatures are not measured but determined
425 thermodynamically, taking the mass flow rate andthe specific
426 thermal capacity of the oil together with all losses below the
427 investigated pass into account.
428 The main advantage of the temperature gradients in com-
429 parison to absolute temperatures is the improved comparabil-
430 ity of related results at different inlet temperatures. According
431 to the 3D CFD results, a very good agreement with the ex-
432 perimental findings throughout the entire spectrum of investi-
433 gated operating conditions can be observed. Deviations are in Fig. 15. Comparison of the measured temperature gradients with numerically
determined 2D and 3D CFD results in pass 2 at various oil flow rates and
434 most of the cases below 1 K. Consequently, the 3D numerical with specific losses per winding turn of P lo sse s, i = 360 W/turn. a) ṁ o il =
435 modelling approach allows an accurate prediction of all rel- 18.0 kg/s,ϑ in = 80 °C, pass 2. b) ṁ o il = 9.0 kg/s, ϑ in = 80 °C, pass 2. c) ṁ o il
436 evant conductor temperatures at operating conditions that are = 3.0 kg/s, ϑ in , = 80 °C pass 2.
437 representative for OD cooled disc windings. In addition, the [14] E. J. Kranenborg, C. O. Olsson, B. R. Samuelson, L./A. Lundin, and R. 500
438 application of 2D winding models might lead towards a con- M. Missing, “Numerical study on mixed convection and thermal streaking 501
in power transformer windings,” in Proc. 5th Eur. Thermal-Sci. Conf., the 502
439 servative thermal design leaving room for a more cost effective Netherlands, 2008. 503
440 winding layout. [15] X. Zhang, Z. Wang, and Q. Liu, “Prediction of pressure drop and flow 504
distribution in disc type transformer winding in an OD cooling mode,” 505
IEEE Trans. Power Del., vol. 32, no. 4, pp. 1655–1664, Aug. 2016. 506
441 VI. CONCLUSION [16] F. Torriano, M. Chaaban, and P. Picher, “Numerical study of parameters 507
affecting the temperature distribution in a disc-type transformer winding,” 508
442 In this contribution, a 2D and 3D simulation model was pre- Int. J. Appl. Thermal Eng., vol. 30, pp. 234–244, 2010. 509
443 sented to determine the oil speed of an OD cooled winding [17] F. Torriano, P. Picher, and M. Chaaban, “Numerical investigation of 3d 510
flow and thermal effects in a disc-type transformer winding,” Int. J. Appl. 511
444 followed by an analysis of the hot-spot temperature and its po- Thermal Eng., vol. 40, pp. 121–131, 2012. 512
445 sition. The presented winding model, which has been validated [18] J. Smolka and A. J. Nowak, “Experimental validation of the coupled fluid 513
446 by experimental results, provides useful insight into the hori- flow, heat transfer and electromagnetic numerical model of the medium 514
power dry-type electrical transformer,” Int. J. Thermal Sci., vol. 47, no. 10, 515
of
447 zontal cooling channels perpendicular to the main oil flow di- pp. 1393–1410, 2008. 516
448 rection. The experimentally determined oil flow velocities and [19] H. Campelo, M. Lopez- Fernandez, P. Picher, and F. Torriano, “Advanced 517
449 their optical investigation showed a good agreement with the thermal modelling techniques in power transformers,” in Proc. Adv. Res. 518
Workshop Transformers, Spain, 2013. 519
450 corresponding numerical data and, thereby, confirmed the suit- [20] S. Tenbohlen, C. Breuer, F. Devaux, R. Lebreton, N. Schmidt, and 520
451 ability of 3D numerical models to calculate the winding thermal T. Stirl, “Evaluation of the thermal performance of transformer wind- 521
452 behavior. The investigation indicated a strong non-uniform oil ings by numerical investigations and measurements,” in Proc. CIGRE, 522
Paris, France, 2016, pp. A2–309. 523
453 flow distribution on the horizontal channels influencing the po- [21] N. Schmidt, S. Tenbohlen, S. Chen, and C. Breuer, “Numerical and ex- 524
454 sition of the hot-spot. The study also showed that the oil flow perimental investigation of the temperature distribution inside oil-cooled 525
455
456
457
458
459
460
velocity in the upper channels is greater than the lower ones.
Correspondingly, the cooling in the lower channels is less and,
therefore, the hot-spot occurs in the bottom area of each pass.
The presented model can be utilized by transformer designers
to optimize the cooling of power transformer windings.
REFERENCES
ro transformer windings,” in Proc. 18th Int. Symp. High Voltage Eng., Seoul,
South Korea, 2013, pp. 25–30.
[22] F. R. Menter, “Two-equation eddy-viscosity turbulence models for engi-
neering applications,” AIAA J. vol. 32, no. 8, pp. 1598–1605, 1994.
[23] N. Schmidt, “Simulation and experimental validation of forced oil flow in
power transformers,” Ph.D. dissertation, University of Stuttgart, Stuttgart,
Germany, 2018, to be published.
Stefan Tenbohlen (M’04–SM’14) received the

526
527 Q5
528
529
530
531
532
533
EP
diploma and Dr.-Ing. degrees from the Tech- 534
461 [1] E. I. Amoiralis, M. A. Tsili, and A. G. Kladas, “Transformer design and nical University of Aachen, Aachen, Germany, 535
462 optimization: A literature survey,” IEEE Trans. Power Del., vol. 24, no. 4, in 1992 and 1997, respectively. In 1997, he 536
463 pp. 1999–2024, Oct. 2009. joined ALSTOM Schorch Transformatoren GmbH, 537
464 [2] P. Picher, F. Torriano, M. Chaaban, and B. Girard, “Optimization of trans- Mönchengladbach, Germany, where he was respon- 538
465 former overload using advanced thermal modelling,” in Proc. CIGRE, sible for basic research and product development. 539
466 Paris, France, 2010, pp. A2–305. From 2002 to 2004, he was the Head of the Electrical 540
467 [3] J. Zhang and X. Li, “Oil cooling for disk-type transformer windings-part I: and Mechanical Design Department. In 2004, he was 541
468 Theory and model development,” IEEE Trans. Power Del., vol. 21, no. 3, appointed to a professorship and Head of the institute 542
469 1318–1325, Jul. 2006. of Power Transmission and High Voltage Technology 543
470 [4] J. Zhang and X. Li, “Oil cooling for disk-type transformer windings-part of the University of Stuttgart, Stuttgart, Germany. He holds several patents and 544
471 II: Parametric studies of design parameters,” IEEE Trans. Power Del., published more than 400 papers. His main research fields are high voltage tech- 545
472 vol. 21, no. 3, 1326–1332, Jul. 2006. nique, power transmission, and electromagnetic compatibility. Dr. Tenbohlen 546
473 [5] Z. Radakovic and M. Sorgic, “Basics of detailed thermal-hydraulic model is a member of the CIGRE SC A2 (Power Transformers), several international 547
IEE
474 for thermal design of oil power transformers,” IEEE Trans. Power Del., working groups, and Chairman of the German Power Engineering Society VDE- 548
475 vol. 25, no. 2, pp. 790–802, Apr. 2003. ETG FB Q2 (materials, electrical insulations, and diagnostics). 549
476 [6] Z. Radakovic and K. Feser, “A new method for the calculation of the 550
477 hotspot temperature in power transformers with ONAN cooling,” IEEE
Nicolas Schmidt received the diploma degree in me- 551
478 Trans. Power Del., vol. 18, no. 4, 1284–1292, Oct. 2003.
chanical engineering from the University of Stuttgart, 552
479 [7] H. M. R. Campelo, C. M. Fonte, R. C. Lopes, M. M. Dias, and J. C.
Stuttgart, Germany, in 2009. After his graduation, 553
480 B. Lopes, “Network modelling applied to core power transformers and
he joined the Institute of Power Transmission and 554
481 validation with cfd simulations,” in Proc. Int. Colloq. Transformer Res.
High Voltage Technology, University of Stuttgart. In 555
Q3 482 Asset Manage., 2012.
2015, he joined the Robert Bosch GmbH, Stuttgart, 556
483 [8] J. Codde, W. Van der Veken, and M. Baelmans, “Assessment of a hydraulic
where he is responsible for thermal modelling of elec- 557
484 network model for zig-zag cooled power transformer windings,” Int. J.
tronic control units. 558
485 Appl. Thermal Eng., vol. 80, pp. 220–228, 2015.
559
486 [9] Transformer thermal modelling, CIGRE Working Group, A2.38, Tech.
Q4 487 Rep., 2016.
488 [10] N. El Wakil, N. C. Chereches, and J. Padet, “Numerical study of heat
489 transfer and fluid flow in a power transformer,” Int. J. Thermal Sci., vol. 45,
Christian Breuer received the diploma degree in 560
490 pp. 615–626. 2006.
electrical engineering from the University of Applied 561
491 [11] A. Skillen, A. Revell, H. Lacovides, and W. Wu, “Numerical prediction of
Science – Georg Agricola, Bochum, Germany. After 562
492 local hot-spot phenomena in transformer windings,” Int. J. Appl. Thermal the graduation in 2009, he joined GE Grid Solutions 563
493 Eng., vol. 36, pp. 96–105, 2012.
(former Areva and Alstom) as a member of the Re- 564
494 [12] J. Mufuta and E. Van, “Modeling of the mixed convection in the windings
search and Development Department. His main area 565
495 of a disc-type power transformer,” Int. J. Appl. Thermal Eng., vol. 20,
of expertise is related to medium power transform- 566
496 pp. 417–437, 2000.
ers, for example, creation of design rules and finite- 567
497 [13] J. Gastelurrutia, J. C. Ramos, G. S. Larraona, A. Rivas, and J. Izagirre,
elements analysis. 568
498 “Numerical models of natural convection of distribution transformers,”
569
499 Int. J. Appl. Thermal Eng., vol. 31, pp. 493–505, 2011.
570 Saeed Khandan received the M.Sc. degrees in heat Raphael Lebreton received the Ing. diploma from the National Superior School 580
571 transfer and fluid mechanical engineering in Iran, of Electricity and Mechanics, Nancy, France, in 2006. In 2008, he joined Areva 581
572 and is currently pursuing the Ph.D. degree in me- T&D, where he was responsible for partial discharge monitoring research and 582
573 chanical engineering. He has been a Researcher and development. Since 2014, he has been working on thermal optimization of 583
574 an Assistant with the Institute of Power Transmis- transformers. 584
575 sion and High Voltage Technology, University of 585
576 Stuttgart, Stuttgart, Germany, since 2016. His princi-
577 pal research interest is in cooling system behaviors
578 of power transformers.
579
of
ro
EP
IEE
586 QUERIES
587 Q1. Author: Please provide expansion for acronym “CFD”. If required.
588 Q2. Author: Please check whether the affiliations for all the authors are ok as set.
589 Q3. Author: Please provide page range for Refs. [7], [14], and [19].
590 Q4. Author: Please provide the report number and organization name and location in Ref. [9].
591 Q5. Author: Please check whether Refs. [21] and [23] are ok as set.
of
ro
EP
IEE
IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 00, NO. 00, 2017 1
1 Investigation of Thermal Behavior of an Oil-Directed

2 Cooled Transformer Winding
3 Stefan Tenbohlen, Senior Member, IEEE, Nicolas Schmidt, Christian Breuer, Saeed Khandan, and Raphael Lebreton
4 Abstract—In this contribution, the oil speed in horizontal chan- determine the hot-spot value. A Literature survey based on IEEE 39
5 nels of an oil-directed cooled winding is investigated by experi- Transactions in [1] revealed the continued interest in the appli- 40
6 mental and numerical methods. The presented winding model of- cation of advanced techniques for transformers design optimiza- 41
7 fers insight into the horizontal cooling channels perpendicular to
8 the main oil-flow direction. To visualize oil flow, tracing particles tion. For the investigation of the temperature distribution within 42
of
9 were added to the cooling oil and illuminated by high-power light- the windings of an OD cooled transformer, different approaches 43
10 emitting diodes. The particle velocities were then determined by can be pursued [2]. 44
11 taking photographs with a defined exposure time. The design of The importance of winding hotspot determination is reflected 45
12 the sophisticated winding model is described in the contribution. in published efforts to predict the hot-spot temperature in the 46
13 In addition to the experimental results, this contribution presents
14 a comparison with respective numerical results from 2-D and 3-D windings. The thermal-hydraulic networks (THN) can be used 47
15 computational fluid dynamics calculations. Finally, numerical re- for describing the entire oil circuit in order to investigate natural 48
16 sults from the winding model concerning the oil-flow distribution circulations [3]–[5]. However, it has been shown in the literature 49
Q1
17
18
19
20
21
22
23
24
inside the winding at various operating conditions are presented.
enabling the designer to optimize the cooling of the power trans-

former windings.
Index Terms—Power transformer, oil-directed cooling, optical
investigation, thermal modelling, computational fluid dynamics,
oil-flow rate.
ro
The investigation indicated a strong nonuniform oil-flow distribu-
tion on the horizontal channels. The presented results give a deep
insight into the oil flow and temperature behavior of windings,
that comparing with the THN, the CFD method provides better
accuracy in prediction of the temperature distributions and fluid
behaviors [6]–[8].
Thermal modeling offers a detailed view of temperature dis-
tribution over the whole winding and temperature values at the
characteristic points. The originally developed algorithm ther-
mal modeling network for temperature calculations is based on
characteristic temperatures.
50
51
52
53
54
55
56
57
EP
25
An efficient thermal design offers the opportunity to signifi- 58
26 I. INTRODUCTION cantly reduce costs in the production process and to improve the 59
OWER transformers are key components applied in trans- thermal performance [9]. Due to the numerical simulations with 60
27
28
29
P mission and distribution systems. Their power rating and
life cycle characteristics are strongly dependent on thermal as-
CFD techniques, a better comprehension of the cooling mecha-
nism of core type transformers has developed to high levels.
61
62
30 pects. The higher the evolving temperature levels at a given In recent years, most of the numerical models are considered 63
31 loading rate are settling, the faster the insulation materials will in 2D approach. El Wakil et al. [10] have performed a 2D sim- 64
32 age at that loading rate. Therefore, a profound knowledge about ulation of a step-down 3-phase power transformer that included 65
33 the temperature distribution inside a transformer is crucial for the core, primary and secondary windings using a commercial 66
34 an appropriate assessment of the component’s power rating and finite volume code. Skillen et al. have [11] investigated on a 2D 67
IEE
35 for that reason offers the opportunity to cut costs in the pro- axisymmetric model the local hot-spot in detail within the low- 68
36 duction process. In the past, and still today to some extent, voltage winding and predicted with CFD simulations the hot 69
37 analytical and empirical methods have been used to predict the streak locations. Mufuta et al. [12] have presented 2D results of 70
38 temperature distribution in the transformer components and to the flow pattern as a function of the Reynolds and Grashof num- 71
bers as well as of geometric parameters. Gastelurrutia et al. [13] 72
have performed the CFD solutions for an ONAN distribution 73
Manuscript received December 19, 2016; revised March 14, 2017; ac-
cepted May 20, 2017. Date of publication; date of current version. Paper no. layer-type transformer. Their model simulated the natural con- 74
TPWRD-01484-2016. (Corresponding author: Stefan Tenbohlen.) vection behavior within the power transformers and displayed 75
S. Tenbohlen, N. Schmidt, and S. Khandan are with the Institute of the oil flow behavior inside distribution layer-type transformers. 76
Power Transmission and High Voltage Technology, University of Stuttgart,
Stuttgart 70569, Germany (e-mail: stefan.tenbohlen@ieh.uni-stuttgart.de; A 2D CFD model is used also to recognize the significant ef- 77
nicolas.schmidt@ieh.uni-stuttgart.de; saeed.khandan@ieh.uni-stuttgart.de). fects of buoyancy term and hot oil streak formations, the latter of 78
C. Breuer is with the R&D Engineer Power Transformers Department, GE which could more or less worsen downstream oil temperatures 79
Grid Solution, 41065 Mönchengladbach, Germany (e-mail: christian.breuer@
ge.com). [14]. Dimensional analyses are conducted in 2D OD cooled 80
Q2
R. Lebreton is with the R&D Engineer Power Transformers Department, GE winding model to identify the independent dimensionless vari- 81
Grid Solution, Puteaux 92900, France (e-mail: raphael.lebreton@ge.com). ables that effect pressure drops in the windings [15]. Torriano 82
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org. et al. [16] have focused on the parameters affecting on the tem- 83
Digital Object Identifier 10.1109/TPWRD.2017.2711786 perature distributions and fluid flow in the winding. 84
0885-8977 © 2017 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.
Fig. 1. Basic winding design and alignment of spacers creating symmetrical

sections with periodic cooling channel system [20].
85 Besides the 2D approach for simulating thermal behaviors of

86 windings, 3D approach is also employed to determine the tem-
of
87 perature distributions and fluid flow in recent CFD calculations.
88 A low-voltage winding has been simulated and the results show
89 that the performances of the 3D numerical computations are sig-
90 nificantly better than that of the 2D models [17]. Smolka et al.
Fig. 2. Design characteristics of the modelled winding, geometry and the
91 [18] have presented full geometry CFD analyses coupled to elec- winding model [20]. a) Top view on winding segment. b) Cross section of wind-
92 tromagnetic simulations to examine the specific power losses ing model. c) Cross section of winding model with given channel dimensions
93 within the coil and core of a dry-type transformer. Campelo and insulation thickness applied between conductors resembling thermal impact
94
95
96
97
98
99
100
ro
et al. [19] have predicted for the OD cooling modes with high
flow velocities, a stagnant flow or even reverse flow can be
observed in the first horizontal channels of the pass. A compre-
hensive review of the thermal modeling in power transformers,
state of the art as well as a proper literature review has become
available in [9].
In this contribution, the oil speed in horizontal ducts of an OD
of electrical insulation.
enlargement shows the horizontal cooling channels are created

by the opening between two consecutive spacers and discs. The
vertical cooling channels are bounded by the cardboard cylin-
ders and by the sticks aligning the spacers. Consequently, the
entire winding geometry can be subdivided into symmetrical
130
131
132
133
134
EP
101 cooled winding is investigated by experimental and numerical
sections. For reasons of an improved visibility concerning the 135
102 methods.
cooling system inside the winding, the outer cardboard cylinder 136
103 Representatively for often used winding types, a so-called
is displayed in a partly transparent color. 137
104 “zigzag” arrangement of a disc-type winding was investigated.
The investigated operating conditions are defined by the tem- 138
105 In such a case, winding discs are layered above each other,
perature of the oil entering the model, the oil flow rate passing 139
106 while so-called spacers keep the axial distance between the
the winding model and the losses supplied to each conductor 140
107 discs and determine the height of the horizontal ducts for the
model. However, since the latter two properties refer strictly to 141
108 oil. The sticks ensure the proper fixation of the spacers and
a single symmetrical section, they are not easy to put into per- 142
109 keep the radial distance between the discs and the outer cover.
spective of common transformer design rules. Since common 143
110 The vertical ducts are formed by the space between discs and
transformer design rules usually refer to entire windings, the 144
111 outer cover. The oil is led from the bottom into the vertical
properties given in this contribution, to reference investigated
IEE
145
112 ducts and flows from there upwards, while it distributes into
operating conditions, are transferred accordingly. For example, 146
113 the particular horizontal ducts. To ensure a proper distribution
instead of specifying the heating power per conductor model 147
114 of the oil to the horizontal ducts, which should be as equal
representing a section of 8° in the circumferential direction of 148
115 as possible, the vertical duct is intermitted alternating between
on entire winding turn, the corresponding losses per complete 149
116 the inner and the outer duct after a specified number of discs
winding turn are referenced. Analogous considerations are ap- 150
117 by so-called washers. This leads to an oil flow in a “zigzag”
plied for the oil flow rate. For that reason, the actually applied 151
118 manner.
heating power per conductor model and the oil flow rates pass- 152
119 In order to confirm the suitability of a numerical 3D winding
ing the winding models are 8°/360° = 1/45 times the referenced 153
120 model to predict the realities in transformer windings accurately,
values pointed out in the figure captions. 154
121 a comparison between experimentally determined oil flow ve-
A top view of a symmetric section of this chosen winding de- 155
122 locities with corresponding numerical data is presented in the
sign including all relevant geometric details to assess this design 156
123 paper.
is shown in Fig. 2. Since the exact reproduction of this geome- 157
try in Fig. 2(a) imposes enormous challenges for the connected 158
124 II. DEVELOPMENT OF AN EXPERIMENTAL MODEL FOR A
manufacturing processes, the respective design of the experi- 159
125 DISC WINDING mental model was modified. For that purpose, the curvature of 160
126 The design of disc type transformer windings shows a strong the conductors in circumferential direction is neglected, as dis- 161
127 symmetry in the circumferential direction and can be subdivided played in Fig. 2(b). To eliminate the detrimental influences of 162
128 into symmetrical sections. Fig. 1 shows a three dimensional neglecting the curvature in 8°, all the dimensions are selected 163
129 view on a disc type winding with four turns per disc. The given according to the symmetrical section. 164
Fig. 4. Piping and instrumentation diagram of the laboratory setup designed

to create strictly controlled boundary conditions inside the winding model [20].
of
Fig. 4 depicts schematically the designed experimental setup 202
in which the flow meter, the flow heater, the gear pump and a 203
Fig. 3. Components of a conductor model for the experimental winding model control unit for supplying power to each heating cartridges are 204
[20]. a) Schematic layout of experimental winding model. b) Experimental employed for obtaining the boundary conditions precisely. It 205
winding model in tank during operation. allows an accurate control over the mass flow rate ṁoil , the oil 206
165
166
167
168
169
170
171
ro
The paper wrapping typically applied around conductors in
transformer windings for purposes of an electrical insulation is
substituted by solid blocks of plastic (Polyvinylidenefluoride)
located only in between the conductors of a disc. Prior experi-
mental investigations showed a great sensitivity to the evolving
temperature levels on the exact thickness of the applied paper
wrapping. Unfortunately, this property, especially with regard to
temperature at the model inlet ϑin and the heating power of each
conductor Plosses, i . 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 flowmeter. Furthermore, two temperature
sensors are employed at inlet of the winding in order to guarantee
that there is no deviation between the oil temperature provided
by the flow heater sensor and that entering the winding model.
207
208
209
210
211
212
213
214
EP
172 the comparably short experimental conductor models, is difficult To minimize the heat leakage from the winding model to its 215
173 to keep within small tolerances. To eliminate respective detri- surroundings, the entire model is submerged in an open experi- 216
174 mental influences but still allow a realistic thermal decoupling mental tank filled with oil of the same temperature as provided 217
175 of the conductor models in one disc, the paper is substituted, by the flow heater. Due to the chosen laboratory setup, a total 218
176 as illustrated in detail in Fig. 2(c). In this contribution the inlet measuring tolerance concerning the temperatures of approxi- 219
177 temperature is set to ϑin = 80 °C while the oil flow rates are set mately 0.5 K is achieved. Consequently, temperature gradients 220
178 to ṁoil = 3 kg/s, ṁoil = 9 kg/s and ṁoil = 18.0 kg/s. The can be determined with an accuracy of around 1 K. 221
179 losses per winding turn are kept at Plosses, i = 360 W/turn. The occurring deviations concerning the mass flow measure- 222
180 To provide the heating power normally dissipated in the ment and heating power control are below 1 %. Because oil is a 223
181 winding turns of a power transformer during operation, heating transparent liquid, the investigation of the oil flow requires addi- 224
cartridges are injected into each conductor model through a tional measures. For that purpose, tracing particles were added
IEE
182 225
183 borehole. An accurate measurement of the conductors’ temper- to the oil of the experimental setup. 226
184 ature is accomplished with temperature sensors located inside These particles were chosen according to the material prop- 227
185 each conductor. The cooling system of OD cooled disc windings erties of the mineral oil and the application conditions of the 228
186 is usually subdivided into so-called passes in the axial direction. investigation. To make the particles visible, a strong source of 229
187 By periodical placement of washers inside the vertical chan- light has to be applied. High power LEDs were mounted di- 230
188 nels alternating between the inner and outer winding diameter, rectly in front of the horizontal channels in experimental setup, 231
189 two different types of passes result. Therefore, the experimental as indicated in Fig. 5. 232
190 winding model must consist of at least two consecutive passes. A digital camera was then mounted in a defined alignment of 233
191 In addition, representative entering and exiting conditions are the horizontal channels and focused on the investigated area in 234
192 required before and after the investigated passes. This ensures the middle of the horizontal channels. Due to the chosen optical 235
193 thermal and hydraulic conditions that are representative for ev- equipment and the connected setup, the depth of field defin- 236
194 ery pass throughout the complete cooling system of the modelled ing the investigated areas in the z-Dimension shown in Fig. 5 237
195 winding. Fig. 3(a) shows the chosen winding model layout in a are limited to approx. 6 mm. The particle velocities within the 238
196 schematic view as well as a clear reference for the pass, disc and investigated area were then determined by taking photographs 239
197 conductor numbering upon which the presented post analysis re- applying a defined exposure time. Firstly, a reference picture of 240
198 lies. Next to a representative winding design, also the operating a glass spacer is taken to create common scale for one measure- 241
199 conditions of the modelled winding are chosen accordingly with ment set, then the camera is aligned in front of the first channel 242
200 respect to the winding type and dimensions. Fig. 3(b) contains of pass and the optical focus is set between the solid spacer 243
201 a photo of the experimental winding model during operation. and the glass spacer. Depending on the oil flow velocity in the 244
of
Fig. 7. Modelling domains of the numerical winding models created for
CFD-analysis [20].
III. CFD MODEL 267
Fig. 5. Experimental setup of the optical analysis [23]. The development of numerical winding models suitable for 268
CFD analysis first of all comprises the definition and creation 269
ro of respective modelling domains [3]. In case of a 3D winding

model, the common design of disc windings allows a significant
reduction of the winding geometry in circumferential direction,
as carried out for the experimental winding model described
above. In addition, the symmetric alignment of the spacers in-
side a symmetrical winding section, offers potential for further
reduction by application of symmetric boundary conditions, as
indicated in Fig. 7.
270
271
272
273
274
275
276
277
EP
The chosen modelling domain of the 2D winding model, al- 278
Fig. 6. Example of a photograph of particle tracks with contrast enhance-
ment and enlargement of the investigated area in the taken pictures and further
lowing a significant reduction of the connected computational 279
enlargement of one particle track [23]. efforts, requires further explanation. Since ANSYS CFX does 280
not feature a dedicated 2D solver, even the analysed 2D ge- 281
245 respective channel, an appropriate setup for the LED power level ometry must consist of 3D finite volumes. However, since only 282
246 and the chosen shutter are determined. It should be noted that for one layer of 3D cells is necessary, the detrimental consequences 283
247 each pass during the measurement, distance between the camera resulting from this requirement are negligible. On the contrary, 284
248 and the model are kept constant. After measuring the particle modelling the winding geometry with a 2D domain can even 285
249 track lengths in the taken pictures, the particle velocities can be take advantage of this restriction by creating a 2D modelling 286
250 calculated with the relation as follows: domain featuring a non-uniform cell depth in third dimension 287
[21]. Therefore, the widening of the horizontal cooling channels
IEE
288
Local velocity of particle vi = (Length of path line of from inner to outer winding diameter, as displayed in Fig. 7, can 289
particle i li )/ (Selected shutter speed Δt) (1) also be acknowledged in the 2D winding model. The chosen 2D 290
cell depths represent a defined fraction of the existing opening 291
251 As demonstrated in Fig. 6, the analysed area is also limited widths of the horizontal cooling channels. 292
252 in the y-Dimension and in the x-Dimension to the exit location Nevertheless, due to the restriction of the 2D approach in 293
253 of the oil inside the horizontal channels. CFX, it does not account for influence of side surface of spac- 294
254 For that reason, only particles inside a volume of approxi- ers. Since the spacers are covering a certain area of the heat- 295
255 mately 20 mm × 2 mm × 6 mm (x × y × z) were analysed transferring surface on the conductors, the numerical model 296
256 within each channel. This restriction is necessary since the assumes no heat transfer at these areas. 297
257 oil flow velocities are not expected to be homogeneously dis- After the described choice of the respective modelling do- 298
258 tributed. Consequently, the investigated area should be as small mains, the creation of the CFD winding models comprises the 299
259 as possible. However, since the velocities might also fluctuate discretization process including the determination of an appro- 300
260 over time due to effects of turbulences, a certain number of priate meshing density. Next to the global mesh size, special 301
261 particle tracks need to be considered for each investigated area. attention has to be paid towards the discretization of the bound- 302
262 Since the number of particles captured on photograph inside a ary layers. The total number of hexahedral meshes employed for 303
263 certain area is directly dependent on the size of the investigated discretization of the 2D and 3D domains are approx. 7 × 106 304
264 area, a compromise between restricting the investigated area cells and approx. 200 × 106 , respectively. Due to the connected 305
265 and the resulting number of available particle tracks recorded computational efforts, only the analysis of the 2D domain should 306
266 for post analysis had to be made. be considered for state of the art desktop workstations. For a 307
TABLE I
INVESTIGATED OPERATING CONDITIONS AND RESPECTIVE REYNOLDS
Operating Mass flow inlet temperature Reynolds numbers

points rate [kg/s] [°C] [1]
1 3 80 667
2 9 80 2000
3 18 80 4000
of
Fig. 9. Distribution of the determined oil flow velocities in pass 1 at ṁ o il =
18.0 kg/s, ϑ in = 80 °C and P lo sse s, i = 360 W/turn [23].
ro
Fig. 8. Distribution of measured velocities in channel number 6 of pass 1 at
ṁ o il = 9.0 kg/s, ϑ in = 40 °C and P lo sse s, i = 360 W/turn with illustration
of an idealized distribution according to a Gaussian distribution function [23].
EP
308 more efficient analysis following the described approach, high
309 performance computing centers become advisable.
310 Table I gives three different operation points corresponding to
311 the mass flow rates equal to ṁoil = 3 kg/s, ṁoil = 9 kg/s and
Fig. 10. Comparison of measured with 2D and 3D numerical calculated
312 ṁoil = 18.0 kg/s with reference inlet temperature of 80 °C. mid channel velocities in pass 1, ṁ o il = 9.0 kg/s, ϑ in = 80 °C, P lo sse s, i =
313 For the determination of the Reynolds numbers, the reference 360 W/turn.
314 flow rate is calculated at the smallest hydraulic diameter through
315 which the entire oil should pass, inside the vertical channel on
316 the inner winding diameter side located across the oil washer. defined operating conditions in the context of an idealized Gaus- 336
317 Table I gives the calculated Reynolds numbers for different sian distribution function. Since this distribution function ap- 337
pears suitable for the collected measurement data, it is applied
IEE
338
318 operation conditions.
319 Operating point 1 indicates the analysis of the heat transfer for the post-processing of the experimental results. Fig. 9 gives 339
320 at a low Reynolds number. The behavior of the fluid flow in the the results at specified operating conditions in pass 1. Next 340
321 low Reynolds number is laminar, whereas operating points 2 and to the averaged, minimum and maximum values the displayed 341
322 3 indicate the analysis at higher Reynolds numbers, where the boxes illustrate the amount of scattering experienced during 342
323 application of turbulence model is required to achieve accurate post-processing the measurements. For that purpose, the boxes 343
324 results with CFD simulations since the laminar character of the enclose the range of measured values comprising 80% of all 344
325 oil is affected by high oil flow rates with high inlet temperature. conducted measurements according to a Gaussian distribution. 345
326 Turbulent numerical models are defined by an identical solver It can be noted, that especially the first and last channel show 346
327 setup, which applies the SST (Shear Stress Transport) turbu- a wider distributed velocity distribution within the investigated 347
328 lence model [22] with its transitioning onset model at different area. While the experimental determination of the oil flow dis- 348
329 operating conditions. tribution is extensive, the corresponding numerical data can be 349
extracted from the CFD result files. Fig. 10 shows the compar- 350
ison of the experimental results and the 2D and 3D numerical 351
330 IV. RESULTS OF THE OPTICAL ANALYSIS REGARDING
post analyses at various operating conditions in pass 1 of the 352
331 OIL VELOCITY winding model. For comparing both approaches, measured data 353
332 After determining the particle velocities in a certain chan- are compared with both numerical models. 354
333 nel at specific operating condition, the gathered data is fur- The respective numerical data at middle cross-section in the 355
334 ther post-processed. To illustrate this process, Fig. 8 shows the 3D winding model of horizontal channels are given as well. 356
335 distribution of measured particle velocities in channel 6 at the Especially in the first channels the analysed location is of great 357
of
Fig. 11. Oil flow distribution onto horizontal channels in pass 1, ϑ in = 80 °C,
ro Fig. 13. Comparison of the oil flow velocities within the 2D and 3D CFD wind-
ing model at ṁ o il = 18.0 kg/s, ϑ in = 80 ◦ C and P lo sse s = 360 W/turn
[20]. a) 2D model (d in on left side). b) 3D model (d o u t in front) inner view.
c) 3D model (d in in front) outer view.
region of the passes. It gets obvious that separation eddies are

blocking the entry of the horizontal channels.
380
381
EP
V. LOCATION OF HOT SPOT 382
One central objective of this contribution is the assessment 383

of the pursued numerical modelling approaches for transformer 384
windings. Since 3D winding models require significantly higher 385
computational efforts in comparison to the 2D approach, it is of 386
keen interest if those additional efforts are justified. To start the 387
Fig. 12. Streamlines at a flow rate of ṁ o il = 9.0 kg/s.
respective comparison between the 2D and 3D winding model, 388
Fig. 13 shows the numerical results for the oil flow distribution 389
358 influence on the determined velocity. This can be attributed to inside the winding models for an oil flow rate of 18.0 kg/s 390
359 separation eddies at the duct entrance (see Fig. 12). qualitatively. 391
The measured mean values agree impressively well with the Although no quantitative comparison of the oil amount pass-
IEE
360 392
361 numerical results in middle position at all operating conditions. ing the different horizontal channels is carried out, the chosen 393
362 Since the thermal investigation shows great agreement between visualization already gives a clear insight into the respective dif- 394
363 experimental and numerical results, the respective results for ferences resulting from the chosen modelling approach. While 395
364 the oil flow velocities are also in good agreement. The numer- the 2D winding model by definition suggests a homogeneous 396
365 ical results for the oil flow distribution are consequently also oil flow distribution in the circumferential direction, the 3D re- 397
366 assumed to be valid. The velocity is proportional to the oil mass sults clearly contradict this assessment. Especially the visible oil 398
367 in the channels. By calculation of the average of the velocities in flow in the vertical channels of the 3D winding model displays 399
368 each channel, the shares of oil flow on the horizontal channels a strong heterogeneous character. 400
369 are determined at three different flow rates as shown in Fig. 11. However, because the oil flow distribution is only of sec- 401
370 The high velocity of the oil after passing the washer (see ondary interest in comparison to the temperature distribution, 402
371 also Fig. 13(a)) is the reason for the formation of the separation Fig. 14 gives the results of the respective temperatures within 403
372 eddies. The oil flow in the lower ducts is becoming very small the investigated winding geometry for pass1 of the 3D winding 404
373 and can even turn to backflow. In order to investigate the reason model for an oil flow rate of 9.0 kg/s. In the investigated arrange- 405
374 for the uneven fluid flow distribution in Fig. 11, Fig. 12 shows ment the location of the hot spot is in the middle two discs (3 and 406
375 the streamlines of the oil flow for a flow rate of 9 kg/s. The results 4), which is in contradiction with the common thinking that the 407
376 of the CFD simulation indicate that the flow distributions on the hot spot can be found on top of the winding. The reason for this 408
377 horizontal oil channels are not equally. For low flow rate of behavior is the non-uniform distribution of the oil flow as seen 409
378 3 kg/s, the maximum velocity occurs in the lower part of the in Fig. 11. The oil velocity in the two last horizontal channels 410
379 pass, whereas for high winding flow rates shifts to the upper is much higher as in the four lower channels which results in a 411
of
ro
EP
Fig. 14. Temperature distribution at ṁ o il = 9.0 kg/ s, ϑ in = 80 °C and
412 better cooling of these discs. Even the turns within one disc do
413 not have the same temperature due to the thermal resistance of
414 the insulation material between the turns. To better quantify the
415 deviations connected to the numerical modelling setup, Fig. 15
416 presents the results for the temperature gradients between oil
and conductors at all investigated operating conditions in the
IEE
417
418 second pass of the winding model.
419 Next to the CFD results, the shown data also contains the
420 measurements, allowing a proper assessment of the respective
421 modelling approach. To calculate the local temperature gradi-
422 ent at a specific conductor, the oil temperature at the inlet of
423 the pass holding that conductor is applied. In case of the mea-
424 surements, oil temperatures are not measured but determined
425 thermodynamically, taking the mass flow rate andthe specific
426 thermal capacity of the oil together with all losses below the
427 investigated pass into account.
428 The main advantage of the temperature gradients in com-
429 parison to absolute temperatures is the improved comparabil-
430 ity of related results at different inlet temperatures. According
431 to the 3D CFD results, a very good agreement with the ex-
432 perimental findings throughout the entire spectrum of investi-
433 gated operating conditions can be observed. Deviations are in Fig. 15. Comparison of the measured temperature gradients with numerically
determined 2D and 3D CFD results in pass 2 at various oil flow rates and
434 most of the cases below 1 K. Consequently, the 3D numerical with specific losses per winding turn of P lo sse s, i = 360 W/turn. a) ṁ o il =
435 modelling approach allows an accurate prediction of all rel- 18.0 kg/s,ϑ in = 80 °C, pass 2. b) ṁ o il = 9.0 kg/s, ϑ in = 80 °C, pass 2. c) ṁ o il
436 evant conductor temperatures at operating conditions that are = 3.0 kg/s, ϑ in , = 80 °C pass 2.
437 representative for OD cooled disc windings. In addition, the [14] E. J. Kranenborg, C. O. Olsson, B. R. Samuelson, L./A. Lundin, and R. 500
438 application of 2D winding models might lead towards a con- M. Missing, “Numerical study on mixed convection and thermal streaking 501
in power transformer windings,” in Proc. 5th Eur. Thermal-Sci. Conf., the 502
439 servative thermal design leaving room for a more cost effective Netherlands, 2008. 503
440 winding layout. [15] X. Zhang, Z. Wang, and Q. Liu, “Prediction of pressure drop and flow 504
distribution in disc type transformer winding in an OD cooling mode,” 505
IEEE Trans. Power Del., vol. 32, no. 4, pp. 1655–1664, Aug. 2016. 506
441 VI. CONCLUSION [16] F. Torriano, M. Chaaban, and P. Picher, “Numerical study of parameters 507
affecting the temperature distribution in a disc-type transformer winding,” 508
442 In this contribution, a 2D and 3D simulation model was pre- Int. J. Appl. Thermal Eng., vol. 30, pp. 234–244, 2010. 509
443 sented to determine the oil speed of an OD cooled winding [17] F. Torriano, P. Picher, and M. Chaaban, “Numerical investigation of 3d 510
flow and thermal effects in a disc-type transformer winding,” Int. J. Appl. 511
444 followed by an analysis of the hot-spot temperature and its po- Thermal Eng., vol. 40, pp. 121–131, 2012. 512
445 sition. The presented winding model, which has been validated [18] J. Smolka and A. J. Nowak, “Experimental validation of the coupled fluid 513
446 by experimental results, provides useful insight into the hori- flow, heat transfer and electromagnetic numerical model of the medium 514
power dry-type electrical transformer,” Int. J. Thermal Sci., vol. 47, no. 10, 515
of
447 zontal cooling channels perpendicular to the main oil flow di- pp. 1393–1410, 2008. 516
448 rection. The experimentally determined oil flow velocities and [19] H. Campelo, M. Lopez- Fernandez, P. Picher, and F. Torriano, “Advanced 517
449 their optical investigation showed a good agreement with the thermal modelling techniques in power transformers,” in Proc. Adv. Res. 518
Workshop Transformers, Spain, 2013. 519
450 corresponding numerical data and, thereby, confirmed the suit- [20] S. Tenbohlen, C. Breuer, F. Devaux, R. Lebreton, N. Schmidt, and 520
451 ability of 3D numerical models to calculate the winding thermal T. Stirl, “Evaluation of the thermal performance of transformer wind- 521
452 behavior. The investigation indicated a strong non-uniform oil ings by numerical investigations and measurements,” in Proc. CIGRE, 522
Paris, France, 2016, pp. A2–309. 523
453 flow distribution on the horizontal channels influencing the po- [21] N. Schmidt, S. Tenbohlen, S. Chen, and C. Breuer, “Numerical and ex- 524
454 sition of the hot-spot. The study also showed that the oil flow perimental investigation of the temperature distribution inside oil-cooled 525
455
456
457
458
459
460
velocity in the upper channels is greater than the lower ones.
Correspondingly, the cooling in the lower channels is less and,
therefore, the hot-spot occurs in the bottom area of each pass.
The presented model can be utilized by transformer designers
to optimize the cooling of power transformer windings.
REFERENCES
ro transformer windings,” in Proc. 18th Int. Symp. High Voltage Eng., Seoul,
South Korea, 2013, pp. 25–30.
[22] F. R. Menter, “Two-equation eddy-viscosity turbulence models for engi-
neering applications,” AIAA J. vol. 32, no. 8, pp. 1598–1605, 1994.
[23] N. Schmidt, “Simulation and experimental validation of forced oil flow in
power transformers,” Ph.D. dissertation, University of Stuttgart, Stuttgart,
Germany, 2018, to be published.
Stefan Tenbohlen (M’04–SM’14) received the

526
527 Q5
528
529
530
531
532
533
EP
diploma and Dr.-Ing. degrees from the Tech- 534
461 [1] E. I. Amoiralis, M. A. Tsili, and A. G. Kladas, “Transformer design and nical University of Aachen, Aachen, Germany, 535
462 optimization: A literature survey,” IEEE Trans. Power Del., vol. 24, no. 4, in 1992 and 1997, respectively. In 1997, he 536
463 pp. 1999–2024, Oct. 2009. joined ALSTOM Schorch Transformatoren GmbH, 537
464 [2] P. Picher, F. Torriano, M. Chaaban, and B. Girard, “Optimization of trans- Mönchengladbach, Germany, where he was respon- 538
465 former overload using advanced thermal modelling,” in Proc. CIGRE, sible for basic research and product development. 539
466 Paris, France, 2010, pp. A2–305. From 2002 to 2004, he was the Head of the Electrical 540
467 [3] J. Zhang and X. Li, “Oil cooling for disk-type transformer windings-part I: and Mechanical Design Department. In 2004, he was 541
468 Theory and model development,” IEEE Trans. Power Del., vol. 21, no. 3, appointed to a professorship and Head of the institute 542
469 1318–1325, Jul. 2006. of Power Transmission and High Voltage Technology 543
470 [4] J. Zhang and X. Li, “Oil cooling for disk-type transformer windings-part of the University of Stuttgart, Stuttgart, Germany. He holds several patents and 544
471 II: Parametric studies of design parameters,” IEEE Trans. Power Del., published more than 400 papers. His main research fields are high voltage tech- 545
472 vol. 21, no. 3, 1326–1332, Jul. 2006. nique, power transmission, and electromagnetic compatibility. Dr. Tenbohlen 546
473 [5] Z. Radakovic and M. Sorgic, “Basics of detailed thermal-hydraulic model is a member of the CIGRE SC A2 (Power Transformers), several international 547
IEE
474 for thermal design of oil power transformers,” IEEE Trans. Power Del., working groups, and Chairman of the German Power Engineering Society VDE- 548
475 vol. 25, no. 2, pp. 790–802, Apr. 2003. ETG FB Q2 (materials, electrical insulations, and diagnostics). 549
476 [6] Z. Radakovic and K. Feser, “A new method for the calculation of the 550
477 hotspot temperature in power transformers with ONAN cooling,” IEEE
Nicolas Schmidt received the diploma degree in me- 551
478 Trans. Power Del., vol. 18, no. 4, 1284–1292, Oct. 2003.
chanical engineering from the University of Stuttgart, 552
479 [7] H. M. R. Campelo, C. M. Fonte, R. C. Lopes, M. M. Dias, and J. C.
Stuttgart, Germany, in 2009. After his graduation, 553
480 B. Lopes, “Network modelling applied to core power transformers and
he joined the Institute of Power Transmission and 554
481 validation with cfd simulations,” in Proc. Int. Colloq. Transformer Res.
High Voltage Technology, University of Stuttgart. In 555
Q3 482 Asset Manage., 2012.
2015, he joined the Robert Bosch GmbH, Stuttgart, 556
483 [8] J. Codde, W. Van der Veken, and M. Baelmans, “Assessment of a hydraulic
where he is responsible for thermal modelling of elec- 557
484 network model for zig-zag cooled power transformer windings,” Int. J.
tronic control units. 558
485 Appl. Thermal Eng., vol. 80, pp. 220–228, 2015.
559
486 [9] Transformer thermal modelling, CIGRE Working Group, A2.38, Tech.
Q4 487 Rep., 2016.
488 [10] N. El Wakil, N. C. Chereches, and J. Padet, “Numerical study of heat
489 transfer and fluid flow in a power transformer,” Int. J. Thermal Sci., vol. 45,
Christian Breuer received the diploma degree in 560
490 pp. 615–626. 2006.
electrical engineering from the University of Applied 561
491 [11] A. Skillen, A. Revell, H. Lacovides, and W. Wu, “Numerical prediction of
Science – Georg Agricola, Bochum, Germany. After 562
492 local hot-spot phenomena in transformer windings,” Int. J. Appl. Thermal the graduation in 2009, he joined GE Grid Solutions 563
493 Eng., vol. 36, pp. 96–105, 2012.
(former Areva and Alstom) as a member of the Re- 564
494 [12] J. Mufuta and E. Van, “Modeling of the mixed convection in the windings
search and Development Department. His main area 565
495 of a disc-type power transformer,” Int. J. Appl. Thermal Eng., vol. 20,
of expertise is related to medium power transform- 566
496 pp. 417–437, 2000.
ers, for example, creation of design rules and finite- 567
497 [13] J. Gastelurrutia, J. C. Ramos, G. S. Larraona, A. Rivas, and J. Izagirre,
elements analysis. 568
498 “Numerical models of natural convection of distribution transformers,”
569
499 Int. J. Appl. Thermal Eng., vol. 31, pp. 493–505, 2011.
570 Saeed Khandan received the M.Sc. degrees in heat Raphael Lebreton received the Ing. diploma from the National Superior School 580
571 transfer and fluid mechanical engineering in Iran, of Electricity and Mechanics, Nancy, France, in 2006. In 2008, he joined Areva 581
572 and is currently pursuing the Ph.D. degree in me- T&D, where he was responsible for partial discharge monitoring research and 582
573 chanical engineering. He has been a Researcher and development. Since 2014, he has been working on thermal optimization of 583
574 an Assistant with the Institute of Power Transmis- transformers. 584
575 sion and High Voltage Technology, University of 585
576 Stuttgart, Stuttgart, Germany, since 2016. His princi-
577 pal research interest is in cooling system behaviors
578 of power transformers.
579
of
ro
EP
IEE
586 QUERIES
587 Q1. Author: Please provide expansion for acronym “CFD”. If required.
588 Q2. Author: Please check whether the affiliations for all the authors are ok as set.
589 Q3. Author: Please provide page range for Refs. [7], [14], and [19].
590 Q4. Author: Please provide the report number and organization name and location in Ref. [9].
591 Q5. Author: Please check whether Refs. [21] and [23] are ok as set.
of
ro
EP
IEE
View publication stats

Investigation of Thermal Behavior of An Oil Directed Cooled Transformer Winding

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Investigation of Thermal Behavior of An Oil Directed Cooled Transformer Winding

Uploaded by

Copyright:

Available Formats

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

Investigation of Thermal Behavior of an Oil Directed Cooled Transformer

Article in IEEE Transactions on Power Delivery · June 2017

Stefan Tenbohlen Raphael Lebreton

SEE PROFILE SEE PROFILE

Particle triggered PD in HVDC GIS View project

Frequency Response Analysis View project

The user has requested enhancement of the downloaded file.

1 Investigation of Thermal Behavior of an Oil-Directed

enabling the designer to optimize the cooling of the power trans-

Fig. 1. Basic winding design and alignment of spacers creating symmetrical

85 Besides the 2D approach for simulating thermal behaviors of

enlargement shows the horizontal cooling channels are created

Fig. 4. Piping and instrumentation diagram of the laboratory setup designed

III. CFD MODEL 267

ro of respective modelling domains [3]. In case of a 3D winding

Operating Mass flow inlet temperature Reynolds numbers

region of the passes. It gets obvious that separation eddies are

One central objective of this contribution is the assessment 383

Stefan Tenbohlen (M’04–SM’14) received the

1 Investigation of Thermal Behavior of an Oil-Directed

enabling the designer to optimize the cooling of the power trans-

Fig. 1. Basic winding design and alignment of spacers creating symmetrical

85 Besides the 2D approach for simulating thermal behaviors of

enlargement shows the horizontal cooling channels are created

Fig. 4. Piping and instrumentation diagram of the laboratory setup designed

III. CFD MODEL 267

ro of respective modelling domains [3]. In case of a 3D winding

Operating Mass flow inlet temperature Reynolds numbers

region of the passes. It gets obvious that separation eddies are

One central objective of this contribution is the assessment 383

Stefan Tenbohlen (M’04–SM’14) received the

View publication stats

You might also like