Vous êtes sur la page 1sur 5

Electric Power Systems Research 78 (2008) 18141818

Contents lists available at ScienceDirect

Electric Power Systems Research


journal homepage: www.elsevier.com/locate/epsr

3D nite-element determination of stray losses in power transformer


Livio Susnjic a, , Zijad Haznadar b , Zvonimir Valkovic c
a

Faculty of Engineering, Vukovarska 58, 51000 Rijeka, Croatia


Faculty of Electrical Engineering and Computing, Unska 3, 10000 Zagreb, Croatia
c
Polytechnic of Zagreb, Konavoska 2, 10000 Zagreb, Croatia
b

a r t i c l e

i n f o

Article history:
Received 16 March 2007
Received in revised form 3 August 2007
Accepted 10 March 2008
Available online 22 April 2008
Keywords:
Stray losses
Finite-element analyses
Eddy currents
Power transformer

a b s t r a c t
This paper presents a three-dimensional (3D) nite-element (FE) analysis of eddy current losses generated
in the tank walls and yoke clamps of a three-phase 40 MV A power transformer. The time harmonic FE
model is used to compute the magnetic leakage eld in the case of a short circuit condition of the power
transformer. Three cases are analyzed to study the impact of modeling tank walls and yoke clamp plates in
FE context in estimation of their losses. The load loss test was carried out on an experimental transformer
to validate the simulation.
2008 Elsevier B.V. All rights reserved.

1. Introduction
Accurate predictions of stray losses of a power transformer
and their reduction mechanisms are necessary for improving
transformer design. The stray losses arise from eddy current and
hysteresis effects inside the yoke clamps and tank walls. The main
portion of stray losses in carbon steel plates are the eddy current
losses [1], whereas hysteresis losses comprise between 25 and
30% of total stray losses. Several authors have considered the stray
losses of the power transformer, not taking into account hysteresis
losses [2,3]. To determine stray losses, a full three-dimensional
(3D) nite-element (FE) analysis of the whole structure is required.
3D geometry discretization of a power transformer requires a
huge number of nodes and elements because overall transformer
dimensions are measured in meters and there exist regions such as
tank walls and yoke clamps where the penetration depth of eddy
current is measured in millimeters. In order to reduce the number
of nodes and elements and to avoid using very demanding computational resources, the computation of eddy current losses by
3D FE is made using a different model of the tank and yoke clamps
in the context of the nite element [46]. The objective of this
research is to investigate the impact of different modeling for the
tank walls and yoke clamps on the computed eddy current losses.

Corresponding author at: Faculty of Engineering, Department of Electrical Engineering, Vukovarska 58, 51000 Rijeka, Croatia. Tel.: +385 51 651435;
fax: +385 51 651416.
E-mail address: livio.susnjic@riteh.hr (L. Susnjic).
0378-7796/$ see front matter 2008 Elsevier B.V. All rights reserved.
doi:10.1016/j.epsr.2008.03.009

This paper summarizes previously works [7,8] and in addition gives


results for simulation by use of the non-linear surface impedance
method. First, yoke clamp plates and unshielded tank walls are
modeled with skin depth independent shell elements, rather than
with the linear surface impedance method, and in the end with
the non-linear surface impedance method. The electromagnetic
leakage eld based on magnetic scalar potential has been calculated for a transformer short circuit condition with a rated current.
The magnetic non-linearity of the transformer core material is
considered. The inuence of regulating coil tapping position on
computed losses is also presented. A commercial FE software (Flux
3D V9.3 [9]) was used to perform the simulation shown in this
paper. The computed results are discussed and compared with
experimental ones.
2. Numerical analysis
The transformer FE model is shown in Fig. 1, and the relevant
data are given in Table 1. Fig. 2 shows a sketch of the transformer
cross-section. The tank walls and clamps thickness are 10 mm
and 25 mm, respectively. BH data of carbon steel material used
for the tank and clamps are given in Table 2. The rated ampereturns are prescribed for the appropriate coils of each phase. The
ampere-turns balanced equation corresponding to the short circuit
condition, in phasor form is
IH (NH + NR ) + IL NL = 0

(1)

Balance of the ampere-turns can be assumed for the coils


wound on the same leg. With the exception of the coils region,

L. Susnjic et al. / Electric Power Systems Research 78 (2008) 18141818

1815

Table 2
BH data of carbon steel material used for the tank and clamps
H (A/m)

B (T)

H (kA/m)

B (T)

16
33
84
119
219
288
483
619
788

0.051
0.1
0.238
0.324
0.527
0.642
0.89
1.01
1.13

1.0
1.26
1.59
2.0
2.52
3.97
6.25
7.82
9.80

1.250
1.350
1.436
1.504
1.551
1.631
1.720
1.763
1.8

2.1. Skin depth independent shell elements

Fig. 1. The transformer FE model (tank is not shown).

Table 1
Transformer data
Symbol

Quantity

Value

S
f
VH /VL
IH /IL
NL /NH /NR


Rated power
Frequency
Rated voltages
Rated currents
Number of turns
Clamp plate and tank conductivity

40 MV A
50 Hz
110 15%/21 kV
209.9/1100 A
152/677/120+120
5 106 S/m

The thin steel plates can be modeled by means of surface region,


given frequency, permeability and conductivity of the material.
Shell elements independent in terms of skin depth have been used
for calculating eddy current losses. The tangential component of the
magnetic eld in a thin plate of thickness e through the depth of
the plate in the z-direction is described analytically by the following
expression:
Ht (z) =

 ae
2

+ az

+ H 2t sh

 ae
2

az



(2)

where a = (1 + j)/, H1t and H2t are the eld values on both sides
of the plate and is the skin depth.The volume current density
variation has a tangential component only, and is described by
J(z) =

the sub-regions of the calculation domain are dened with the


total scalar magnetic potential formulation. Reduced potential
described the coils region. The calculation of the magnetic eld
from the Biot-Savarts law allows for the exclusion of the coils from
the nite element mesh.
The transformer is reconectable on the high voltage (HV) side.
The current in the coils and the number of turns corresponding
to a tapping position are given in Table 3. Estimation methods for computing the eddy current losses are briey presented
as follows.

1
H 1t sh
sh(ae)

a
H 1t sh
sh(ae)

 ae
2

+ az

H 2t sh

 ae
2

az



(3)

Eddy current loss per surface unit in the plate is


e/2


P=

2
1 
J(z) dz
2

(4)

e/2

where  is material conductivity.


2.2. Linear surface impedance method

Surface impedance links the component of the magnetic eld H


tangential on the thin steel surface to the tangential component of

the electric eld E:

xE = Zs n
x(n
xH)
n

(5)

For linear material it is a ratio of the tangential electric eld Es and


the tangential magnetic eld Hs :
Zs =

Es
1+j
=
Hs


(6)

The surface current density is dened as


s
xH
K = n

(7)

The steel plate power loss density (surface density) in W/m2 is given
by

 2

P = 0.5Re(Zs )Hs 

(8)

Table 3
Coils data

Fig. 2. The transformer cross-section.

Tapping position

Turns number (LV/HV/RC)

Current (A) (LV/HV/RC)

15%
0
+15%

152/677/0
152/677/120
152/677/120 + 120

1100/247.1/0
1100/209.9/209.9
1100/182.4/182.4

1816

L. Susnjic et al. / Electric Power Systems Research 78 (2008) 18141818

2.3. Non-linear surface impedance method


According to [5], the surface impedance for non-linear material
(permeability depends on magnetic ux density) over a large range
of elds (from low to high elds), and is given by the following
formula:
Zs = kw (Hs )Zsl + (1 kw (Hs ))Zsnl

(9)

where Zsl and Zsnl are the surface impedances for the linear and
non-linear material, respectively, and kw is the weighting function.
Weighting function kw (Hs ) is:
kw (Hs ) =

1
1 + k(Hs /Hk )

(10)

Hk corresponds to the value of the magnetic eld at the knee of the


BH curve, and k is the coefcient to be chosen. The magnetic eld
reaches the tank wall and clamp plates in a mostly normal direction. In this case, the electric eld is mostly tangential, remaining
unchanged through an interface and is assumed to be mostly sinusoidal. The value of coefcient k equals 1 in this case of sinusoidal
electric eld [5].
The steel plate power loss density in W/m2 is calculated by (8).
3. Results
Calculations on a three-phase, three-limb transformer rated
at 40 MV A and 110/21 kV have been performed. The computed
results by skin depth independent shell elements and by linear surface impedance method are shown in Figs. 3 and 4, respectively.
These results correspond to the regulating coil tapping position 0.
Figs. 3a and 4a shows the relative tank permeability (rt ) depen-

Fig. 4. Permeability dependence of the: (a) tank loss value for both xed (rcl = 500)
and varied relative permeability of the yoke clamps and (b) yoke clamps loss value
for both xed (rt = 500) and varied relative permeability of the tank (linear surface
impedance method).

dence of the tank loss values for both parametrically given clamps
permeability (rcl = 500) and varied relative permeability of the
yoke clamps (rcl = rt ). Figs. 3b and 4b shows the relative clamps
permeability (rcl ) dependence of yoke clamps loss values for both
parametrically given (rt = 500) and varied relative permeability
of the tank (rt = rcl ). The variation of the relative permeability
for the tank and yoke clamps is simultaneous in the range from
100 to 1000, with an incremental value of 100. The eddy current losses computed by modeling tank walls and yoke clamps
with skin depth independent shell elements and the linear surface
impedance method are in close agreement. From Figs. 3 and 4 it
is obvious that the losses depend on the chosen steel permeability. The clamp plates loss depends on the permeability of the plate
as well as on the permeability of the tank. It has been shown that
a higher prescribed permeability of the tank results in a reduced
leakage eld in the clamp plate area, and as a consequence reduced
yoke clamps losses. A similar conclusion applies to the tank loss.
The skin depths variation for linear analyses is from 3.18 mm to
1 mm, for the chosen relative permeability of the steel from 100
to 1000. The eddy current losses computed by simulation with the
non-linear surface impedance method, for different regulating coil
tapping positions, are given in Table 4. It could be seen that the
losses obtained with non-linear surface impedance analyses are at
Table 4
Computed losses

Fig. 3. Permeability dependence of the: (a) tank loss value for both xed (rcl = 500)
and varied relative permeability of the yoke clamps and (b) yoke clamps loss value
for both xed (rt = 500) and varied relative permeability of the tank (skin depth
independent shell elements).

Tapping position

Clamp plates (W)

Tank (W)

Total losses (W)

15%
0
+15%

3808
5522
7930

10,700
19,680
36,476

14,508
25,202
44,406

L. Susnjic et al. / Electric Power Systems Research 78 (2008) 18141818

Fig. 5. Distribution of eddy current density on the tank wall (non-linear surface
impedance method).

1817

Fig. 7. Power loss distribution on the tank inner surfaces (non-linear surface
impedance method).

Fig. 8. Distribution of magnetic induction (leakage eld) on symmetry plain in the


coils region (max. value is 0.22 T).

Fig. 6. Surface power density on the clamp plate (non-linear surface impedance
method).

least 30% higher than those with linear surface impedance or independent skin depth shell elements. Fig. 5 shows the distribution of
the surface eddy current density on the tank wall. Distribution of
power loss density on the clamp plate surface is shown in Fig. 6.
Power loss density distribution on the inner tank surface is shown
in Fig. 7. Distribution of the magnetic induction or leakage eld on
the symmetry plain outside the core (in the coils regions) is shown
in Fig. 8. Maximum value of the leakage eld is 0.22 T.
4. Experimental validation
According to IEEE Std. C57.12.90, power transformer load losses
should be measured at a load current equal to the rated current

Fig. 9. The transformer during its manufacturing (Koncar Power Transformers Ltd.).

1818

L. Susnjic et al. / Electric Power Systems Research 78 (2008) 18141818

Table 5
Discrepancy between computed eddy current and measured stray losses
Tapping position

Total losses

I2 R

Winding eddy
current losses

Tank and clamps


stray losses

Tank and clamps eddy


current losses

Discrepancy eddy
current vs. stray losses

15%
0
+15%

209,500
210,100
228,900

169,500
155,600
146,200

19,200
21,100
27,600

20,800
33,400
55,100

14,508
25,202
44,406

30.2%
24.5%
19.4%

for the corresponding regulating coils tapping position. The load


loss test is accomplished by short-circuiting the secondary winding and applying a reduced voltage to the primary winding, i.e. the
voltage necessary to cause a rated load current to ow. A 40 MV A
transformer was used to investigate stray losses (Fig. 9). The power
absorbed in the short-circuit test consists of the I2 R and eddy current losses in the winding, and stray losses in constructive steel
parts. Stray losses in the constructive steel parts are obtained by
subtracting I2 R losses and eddy current losses in the winding from
the power obtained in the load test [10]:
Pstray = Pload Pi2 R Pec

(11)

where Pload is the load losses (W); Pi2 R the I2 R losses in winding (W)
and Pec the winding eddy current losses (W).
The stray losses obtained by (11) are treated as measured losses.
The winding eddy current losses are calculated analytically by
known distribution of the magnetic leakage eld, calculated previously. The magnetic leakage eld inside windings are calculated as
a 2D axisymetric eld [11]. The method described in Ref. [12] is used
to estimate winding eddy current losses. The winding eddy current
losses depend on the tapping position of the regulating coils, e.g.
for a 0 tapping position there are 21.1 kW. Table 5 shows the measurement values of total load losses and I2 R in windings, calculated
winding eddy current losses, stray losses and calculated eddy current losses in the tank walls and clamp plates. Also, stray losses
for different tapping positions are compared with the eddy current
losses computed by modeling clamp plates and tank walls with the
non-linear surface impedance method. The comparison between
the computed and the experimental results shows discrepancy. The
discrepancy results due to approximation in the modeling clamp
plates geometry and in not taking hysteresis losses into account.
For the three methods mentioned above approximation in modeling clamp plates geometry is needed, so brackets (elements for
tight the coils) are not included. Made of carbon steel, brackets are
high in permeability and liable to invite leakage ux concentration
causing eddy current losses.
5. Conclusion
In this paper, power transformer eddy current losses in the yoke
clamps and unshielded tank walls are computed by 3D FE analyses.

The permeability of the tank and clamps has a signicant inuence


on eddy current losses. The losses obtained with non-linear surface
impedance analyses are at least 30% higher than those with linear
surface impedance or independent skin depth shell elements. Stray
losses are a function of many factors including the physical geometry of the cores and coils, the voltage class of the transformer, and
the material used in the tank and clamps construction. The computed values of losses by 3D FE analysis do not match closely the test
values. The comparison between the computed and the experimental results shows the discrepancy which arise due to approximation
in the modeling clamp plates geometry (where the brackets were
not included) and in not taking hysteresis losses into account. The
tapping position is shown to have a strong inuence on the obtained
eddy current losses values.
References
[1] N. Takahashi, Z. Cheng, Team Problem 21, in: Team-Based Benchmark Family (Problem 21/21a /21b /21c ), R&D Center, Reading Tianwei Group Co. LTD.,
2005.
[2] J.C. Olivares, et al., 2D nite-element determination of tank wall losses in pad
transformers Electric Power Research, vol. 71, ELSEVIER, 2004, pp. 179185.
[3] M. Rizzo, et al., Inuence of ux collectors on stray losses in transformers, IEEE
Trans. Magn. 36 (2000) 19151918.
[4] C. Guerin, et al., 3D Eddy current losses calculation in transformer tanks using
the nite element method, IEEE Trans. Magn. 29 (1993) 14191422.
[5] C. Guerin, et al., Surface impedance for 3D non-linear Eddy current
problemsapplication to loss computation in transformers, IEEE Trans. Magn.
32 (1996) 808811.
[6] C. Guerin, et al., Shell element for computing 3D Eddy currentsapplication to
transformers, IEEE Trans. Magn. 31 (1995) 13601363.
[7] L. Susnjic, et al., Electromagnetic analysis applied to the prediction of stray
losses in power transformer, ICEM 2004, Book of Digests, Krakow, 2004, pp.
907908.
[8] L. Susnjic et al., Prediction of Stray Losses in Power Transformer, ISEF 2005, International Symposium on Electromagnetic Fields in Mechatronics, Proceedings,
Baiona, CFSA-2.2., 2005.
[9] FLUX V.9.3, CAD package for electromagnetic and thermal analysis using nite
elements, CEDRAT, 2006.
[10] J.C. Olivares, et al., Experimental study to reduce the distribution-transformers
stray losses using electromagnetic shields, Electr. Power Syst. Res. 63 (2002)
17.
[11] M. Christoffel, A. Kuster, Methode zur Berechnung von Kurzschlusskraeften
in Transformatorwicklungen mit Hilfe von digitalen Rechenmaschinen, Brown
Boveri Mitt. 47 (1960) 321328.
[12] W. Dietrich, Berechnung der Wirkverluste von Transformatorenwicklungen
unter Beruecksichtigung des tatsaechlichen Streufeldverlaufes, Arch. Elektrotech. 46 (1961) 209222.

Vous aimerez peut-être aussi