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

net/publication/3353330

Part winding resonance: Demerit of interleaved high-voltage transformer

winding

Article  in  IEE Proceedings - Electric Power Applications · June 2000

DOI: 10.1049/ip-epa:20000304 · Source: IEEE Xplore

CITATIONS READS

26 456

2 authors:

Abhinandan De N. Chatterjee
Indian Institute of Engineering Science and Technology, Shibpur 9 PUBLICATIONS   142 CITATIONS   
76 PUBLICATIONS   387 CITATIONS   
SEE PROFILE
SEE PROFILE

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

Smart Grid View project

A Study of the Over Voltage Phenomenon in Grid Connected Transformers under Abnormal System Conditions View project

All content following this page was uploaded by Abhinandan De on 11 July 2017.

The user has requested enhancement of the downloaded file.

 Part winding resonance: Demerit of interleaved
high-voltage transformer winding
A.De and N.Chatterjee

Abstract: Interleaved transformer winding is a new and promising development in the field of design
of high-voltage transformer windings. Although the strengths and weaknesses of the interleaved
winding concept against surge voltages are theoretically known, implementation in a practical
multiwinding EHV transformer may lead to various complications and the winding can show
unpredictable behaviour contrary to existing knowledge. A comparative assessment of interleaved
transformer winding has been made by EMTP simulation of two identical multiwinding EHV
transformer models having interleaved and noninterleaved conventional disc windings. The
observations regarding relative surge performance of interleaved and noninterleaved windings may
throw some light on optimum interleaving of the disc coils in transformer winding. The tap changer
or regulating winding forms a weak point in H V transformers and is found to be responsible for the
majority of failures. Distribution of impulse voltage in the regulating or tap-changer winding has also
been studied. The results of these observations may be useful in finding suitable explanations for the
failure of tap-changer windings in EHV power transformers

1 Introduction of other windings to ground, e.g. solid grounding or

grounding through resistance, and the degree of nonhomo-
The distribution of impulse voltage in transformer winding geneity in the windings caused by local reinforcement of the
has long been a factor of significant importance in the turn insulations, etc. An accurate mathematical analysis
design of high-voltage transformers. With the very large that would cover all these aspects is virtually impossible. In
EHV power transformers now being built, an accurate the present paper, an alternative solution to the problem
assessment of the voltage stresses likely to occur under has been approached which involves the development of a
surge conditions is essential to achieve an optimum insula- suitable model for the concerned transformer winding with
tion design. concentrated inductances, capacitances and resistances that
The behaviour of a surge-stressed transformer winding is only simulate the electrical quantities without any geometri-
studied using analogue models or by numerical methods cal resemblance [3] and finding a suitable numerical
using a digital computer. The analogue model is an impor- method of analysis utilising circuit solution principles [4].
tant tool, since the response of a well-constructed model is The present work deals with the response under lightning
in reasonably good agreement with that of the original [l]. impulse of a three-phase, 31.5MVA, 132kVi33kV star/delta
However, the analogue model method has certain limita- transformer with tap-changer winding, having interleaved
tions. First, it is costly and time consuming. Secondly, it continuous disc winding throughout. To study the effect of
lacks the flexibility, as trying several alternative designs usu- interleaving of turns of the high-voltage windings, transient
ally requires building of several models. On the other hand, studies have been made on another identical transformer,
applications of mathematical analyses for determination of having noninterleaved conventional disc winding, for a
transient processes in transformer windings is mostly con- comparative assessment. Equivalent high-frequency circuit
fined to single-layer cylindrical coils and rectangular inci- models of the transformer windings concerned have been
dent waves [2]. developed taking into consideration various electrostatic
However, transformers, especially the high-voltage high and electromagnetic interactions [5]. The method described
power types, are much more intricate systems than the sin- requires the transformer winding to be replaced by a more
gle-layer coil system. Such systems are too complicated to or less complicated ladder network with finite or infinite
permit mathematical analysis without simplifying assump- number of elements depending on whether its parameters
tions. Transient response depends on several factors, such are considered lumped or distributed. However, reduction
as the type of winding under consideration, the manner in of the continuously distributed self and mutual inductances
which the winding is subjected to surge, the manner in from the original winding to corresponding lumped param-
which other windings are connected, the type of connection eters in the equivalent circuit is dificult and requires exten-
sive mathematical formulation.
0IEE, 2000
2 Equivalent circuit representationof transformer
IEE Proceedings online no. 2oooO304 winding
DO1 IO. 1049/igepa:20000304
Paper fmt received 30th June 1999 and in revised form 21st Januaty 2000 The accuracy of the results obtained from the equivalent
The authors are m i t h the EleCtrical Engineering Department, Jadavpur Univer- circuit method depends entirely on the degree of sophistica-
sity, Calcutta 700 032, Indirt tion introduced in the representation of the windings. The
IEE Pruc.-Elrctr. Power Appl., Vol. 117, No. 3, M u y 2000 161
equivalent circuit must describe the surge behaviour of the have been lumped together to form single equivalent
windings for wide range of variation of the input pulse mutual inductance [15]. The effect of an iron core of the
voltage wave shape and must be responsive to hgher fre- transformer has not been considered in the computation of
quencies [6]. At the same time, the amount of computa- inductances [I].
tional time and cost involved in the solution of such models
should also be taken into consideration [7]. 3.4 Equivalent series resistance
The most detailed modelling will be one in which every Effective resistance of the coils has been considered, taking
turn of the winding is represented and all the turn-to-turn into account nonuniform distribution of current in thc con-
inductances and capacitances are considered. This type of ductors due to skin effect and eddy currents (Section 7.4).
representation is capable of describing the transient behav- However, parallel resistances, representing dissipation
iour of each winding turn in detail [SI. However, such a losses in the capacitor dielectrics, and eddy current losses in
model may be prohibitive from the consideration of com- the conducting cylinder formed by low-voltage winding
puter memory requirement, CPU usage and running cost. have been neglected.
Various degrees of simplifications are possible until the net-
work is reduced by successive lumping of the elements, to a 3.5 Shunt capacitances between different
simple RLC circuit of a few elements [9]. Most of the prac- winding sections
tical transient studies are based on coil-by-coil representa- Shunt capacitances between main and tap windings have
tion of the winding. By such representation, the identity of been considered to take into account electrostatic coupling
the individual turns is no doubt lost but the sections can between the windings.
still be identified. This form of winding representation pro-
vides sufficient detail of the surge behaviour of the winding 3.6 Mutual inductance between distinct
in the normally encountered frequency range of 20 - winding sections
200kHz. Values of the lumped parameters used in the Mutual inductances between main and tap windings has
equivalent circuit are calculated from the dimensions and also been considered to take into account inductive cou-
physical properties of the transformer material and are pling between the windings.
greatly dependent on the coil and terminal connections of All the inductance components have finally been lumped
the transformer. together to an easily representable form of ‘equivalent
inductance per disc coil’ (Section 7.2.2), using the method
3 Development of equivalent circuit of transformer suggested by K.A. Wirgau [ 151.
windings
4 Discussion of results
The equivalent circuit is developed utilising the winding
parameters of a 31.5MVA 132kVl33kV transformer, Results obtained from EMTP simulation of the trans-
impulse tested at the High Voltage Laboratory using the former models are presented here. In Figs. 1-5, VHMI.,,,
neutral current method with the winding terminals con- and VH,,-,,, correspond to voltage with respect to ground
nected as per IEC specification [5]. The present equivalent at the ith disc of HV main winding, andjth disc of HV tap
circuit model contains 78 elementary sections, representing winding, respectively, and ‘disc HMI-HM,,1’ corresponds
46 main and 32 tap winding discs. A schematic diagram of to the ith disc of high-voltage main and ‘disc H7;-HTJ+1’
the equivalent circuit model has been presented in the corresponds to the jth disc of high-voltage tap winding
Appendix (Section 7.6). In the equivalent circuit developed, (Fig. 12). The peak of the incident impulse voltage was
the following basic R,L, C parameters have been consid- chosen to be 550kV as per the test specification [5] for a
ered: 132kV transformer. In all the stages of investigation the
transformers were assumed to be operated under nominal
3. I Series capacitance of each coil tapping condition. Therefore the interconnecting lead of the
In the calculation of series capacitance (Section 7.3), inter- 16th and 17th tap winding discs, i.e. the point HT17 in the
turn as well as interdisc capacitances between adjacent coils equivalent circuit model of Fig. 12 has been grounded.
have been taken into consideration [lo, 111. Small stray It is obvious that accuracy and reliability of the results
capacitances between one coil to other distant coils have obtained from the equivalent circuit model depends entirely
been neglected. on validity of the modelling technique, as well as on how
realistic the mathematical formulations are. Reliability of
3.2 Shunt capacitance between coils and the computational results can be ensured by comparing the
neighbouring earthed bodies results obtained from the model to the actual test results of
The earthed bodies include both the metal tank as well as the transformer concerned. In the present case, comparison
nonimpulsed windings, which have also been earthed either is made of the neutral current obtained during actual
directly or indirectly through resistance at the time of impulse testing and that obtained from the model. The
impulse testing. structure of the neutral current depends on the circuit
parameters of the transformer under test. Any change in
3.3 Self and mutual inductance the magnitude of the circuit parameters like inductance and
Self inductance of the disc coils has been calculated capacitance is reflected on the neutral-current oscillogram
(Section 7.1), using Grover’s extension to Rosa’s formula and is universally considered a reliable checking method.
for computation of self inductance of circular disc coils of The neutral current obtained from the proposed model
rectangular cross-section [121. Mutual inductance between under identical circuit conditions and voltage has been pre-
the disc coils have been calculated (Section 7.2), by the Lyle sented (Section 7.5, Fig. 116) for comparison with the
method of equivalent filament [13], and the tables presented actual neutral-current oscillogram obtained during impulse
by Grover [14]. Since, in a lumped parameter transformer testing (Fig. llu) in support of the capability of the pro-
model, representation of evenly distributed numerous posed model to reproduce true behaviour of the winding.
mutual inductances formed by each coil with the other coils From Figs. la and 6, illustrating time dependence of
is virtually impossible, the distributed mutual inductances potential at different points of the HV main and tap wind-
168 IEE Proc -Elect, Power Appl Vol 147 No 3, Muy 2000
ings for interleaved winding design, it is observed that the 600r
potentials from various points to ground are almost line-
arly distributed over the length of the windings from line
end to ground. This signifies the improvement in the surge
performance of the windings due to interleaving of the
turns when compared with the results for noninterleaved
winding as illustrated in Figs. 2a and b. This is also evident
from the spatial distribution of potential. The spatial poten-
tial distribution curves for noninterleaved design shows a
highly nonlinear initial distribution (Fig. 3b), as the line-end
portion of the main winding is severely stressed compared
with other parts of the winding. Up to a time of 2 ~ ( sfol-
lowing application of the voltage, the initial 50% of the -2001 I

winding from the line terminal withstands 100% of the a

applied potential. Mid-winding portions are stressed at a 150r
delayed time as the applied surge travels through the length
of the winding with finite velocity. From Fig. 3b it is 100
observed that the peak of the voltage wave reaches as high
as 420kV at a point approximately 20 discs away from the > 50
Y
line terminal after 3 0 ~At. a further delayed time of 50w, W
voltage peak occurs at the 32nd disc from the line terminal
with a lesser magnitude of 380kV. This indicates attenua- -8
c

>
0
0

tion of the wave peak owing to various losses as the wave - 50

gradually propagates through the winding. Comparative
figures for the interleaved design show a perfectly linear
spatial distribution till 3Ops following application of the
voltage (Fig. 3u) after which the distribution tends to be - 1 5 0 1
nonlinear, although the degree of nonlinearity is much less 0 50 100 150 200 250
compared with the case of noninterleaved winding. time, v s
b
600r - H M I - G N D a- HM20-GND b
H T I - G N D HT25-GND
-....- HM
HM5-GND - - HM40-GND
IO - GND
HTB-GND HT33-GND
HTl7-GND
Fig. 2 Tine dependence ofpotential to groundfrom d@ermtpoint.s of ~ f m i -
mg for noninterleaved design
a HV main winding
>
Y b Tap winding
W
(5,

-
1

>
Y

0)
m
0
-1001 I I 6 I
c
-
0
a

100r
80
60 I \ a
600
> LO
Y
20 5 0 0 k
(5,
0)

c
0
0
>
-20 Y

0)
-LO 0
0
c
d

- 60 >
0
-801 I 1
0 50 100 150 200 250
time,vs
b
o b -2001
- M I - GND
- - HHM5-GND H T l - GND
HT8 - GND
I

H M I HMI1 HM21 HM31 HMLl HT5 HT15 HT25

I

disc n u m b e r
.... HMIO-GND HT17 -GND
- HM20 -GND HT25 - GND b
-- HM40 -GND HT33 - GND -0.5us ----Ius .....2)Ls - 1ous
Fig. I Tune ~ ( eendence ofpotentud to groundfionz d@ermfpomts of wind- - - - 3oUs 50ws xxxx 1 0 0 ~ s
ing for mterleuvedDdeJzgn Fig.3 Spatial distribution potential along length of winding3
a HV main winding a For interleaved design
b Tap wnding b For noninterleaved design

IEE Proc.-Electr. Power Appl., Vol. 147, No. 3, May 2000 169
 The improvement in linearity of the potential distribution 100
curves for the interleaved design is caused by the large
increase of series capacitance of the winding discs owing to
interleaving, while shunt capacitance values remain practi-
60
cally unaltered for both designs.
Waveforms of the disc voltages for interleaved and non-
interleaved designs are presented in Figs. 4 and 5. respec-
tively. For either design, the line end discs are observed to
be highly stressed. However, the severity of voltage stresses
is greatly reduced in the case of interleaved design. The
peak value of potential across the first line-end disc is
found to be as high as 88kV for the noninterleaved design
(Fig. 5a), which is about 16'Yn of the applied impulse peak -401
and eight times the expected peak value assunling ideal uni- a
form distribution. The corresponding peak value for the
interleaved design is 22.5kV (Fig. 4u), about twice the
average value expected from an ideal uniform distribution. 10
The 5th and 15th discs are observed to be stressed at volt-
age peaks of 44 and 30kV for noninterleaved design. These 7 5
stresses are reduced to 19.5 and 13kV, respectively, by 1
W
interleaving winding sections. Further reduction of stresses B o
on the line-end disc is possible if instead of two-section c
d

>
0
interleave grouping, four or more sections are grouped
-5
together, thereby further increasing the self capacitance of
line-end discs. Other known methods of stress control like
-101
use of line-end shields is also possible.
1 1 1 5 1 -
25 r 0 50 100 150 200 250
time,ps
b
20\
15 b
---
a
H M I - HM2
HM 5 - HM6
-
H T l HT2
HT5 - H T 6
.... HM15-HM16 HT16 - HT17
-
--
HM25 - HM26
HM35-HM36
HT20 - HT21
HT25-HT26
-_ HML5 -HM46 HT32 -HT33
Fig. 5 7 " e cieyauimce of d ~ ~ p t e n i l a l s fironwitet
or leuwd &sign
U F ~nidm
I winding discs
h Foi tap windiiig discs

For either design, the potential across tap-winding discs

exhibit sustained oscillatory behaviour (Figs. 4b and 5b).
For interleaved design, these oscillatory voltages are highly
underdainped and practically unattenuated, giving indica-
6
tion of part winding resonance in the tap winding. Chance
of occurrence of part winding resonance is higher in the
> 4 case of interleaved winding design, as an interleaved wind-
Y

w - 2 ing has a lower resonant frequency, owing to its high series

s o
0,
capacitance, so that input excitation even in the form of a
> standard lightning llnpulse can force the winding to oscil-
-2 late in its natural frequency. This is of serious concern to
-4 the design engineer. To avoid such unwanted resonance in
W v
any part of the high-voltage windings. the resonant fre-
-6v " V quency of the winding sections should always be predeter-
- 8L d --. - - L 1
0 50 100 150 200 250 mined. If possible, the response of the model winding to
time,ps different possible f o r m of input excitation should be stud-
b ied before implementation so that the designed winding can
b
- c -
a
HMI - HM2
HM5 - HM6
HTI -HT2
HT5 - HT6
successfully withstand even nonstandard voltage wave-
shapes having a high-frequency spectrum component of
.... HM15-HMI6 HT16-HT17
large amplitude in operation. In the present case, interleav-
- HM25 -HM26 HT20-HT21
- - HM35 -HM36 HT25 -HT26 ing of tap-winding discs has reduced the estimated resonant
HM45 -HML6 HT32 -HT33
frequency of the tap-winding discs from a value of 80kHz
Fig. 4 ? ' h e &wrr&iice of ciix poiennticdsj&~in~er/euved
ciesigtz
U For niaili winding disc5 for the noninterleaved design to as low as 22kHz in the
h For tap winding discs case of the interleaved deign. Since the new value is quite
close to the input excitation frequency of 20kHz for a light-
These observations are well in agreement with the theory, ning impulse wave of 1 . 2 front ~ time, resonance in the
as well as with the well-known expenniental observations tap-changer winding in response to a standard lightning
pertaining to studies on analogwe models of winding by dif- impulse can be predicted well in advance. The simulation
ferent researchers [I, 151. results support this theoretical proposition, as the frequency
170 IEk Proc -Electr P o w r .4ppl Vol 147, No 3, May 2000
 of oscillation of the tap-winding discs for interleaved design 6 References
is observed to be close to the resonant frequency (Fig. 4b).
A sudden rise in potential is observed across the discs 1 MIKI, A., HSOYA, T., and OKUYAMA, K.: ‘A calculation method
for impulse voltage distribution and transferred voltage in transformer
located at the earthed end of the winding owing to reflec- winding’, IEEE Trans., 1978, PAS-97, (3), pp. 930-939
tion. This is evident from the disc voltage curves (Figs. 4 2 DENT, B.M., HARTILL, E.R., and MILES, J.G.: ‘Method of analy-
and 5). Same phenomena are observable across the discs sis of transformer impulse voltage distribution using a digital conipu-
ter’, Proc. IEE, Part A , 1958, 105, (23), pp. 4 4 4 5 9
located at the interconnecting region of the main and tap 3 DEGENEFF, R.C., and VAKILIAN, M.: ‘Modelling power trans-
windings. This is due to mismatch in the impedance offered formers for transient voltage calculations’. CIGRE, August 1992,
to the propagating wave by the main and tap-winding discs 4 DOMELL, H.W.: ‘Digital computer solution of electromagnetictran-
and reflection of wave thereof. To avoid such unwanted sients in single aiid multi-phase networks’, IEEE Trans.. 1969, P A S
88, pp. 388-399
reflection and voltage multiplication the design engineer 5 ‘Power transformer - insulation levels and dielectric tests’. IEC publi-
should try to reduce nonhomogeneity in the high-voltage cation 763, 1980
windings as far as possible. Gradual building of high 6 MORCHED, A., MARTI, L., and 07TENVANGERS, J.: ‘A high-
frequency transformer model for the EMTP, IEEE T m m , 1993,
potential is observed at the open-ended section of tap- P W R M , (3), pp. 1615-1625
changer winding, away from the grounded neutral point. 7 DEGENEFF, R.C.: ‘Reducing storage and saving computational time
Under operation at the -10% tapping condition when the with a generalization of the Dommel (BPA) solution method’, IEEE
T”.,1978, PAS-97, (2), pp. 319
full portion of tap-changer winding becomes open 8 DE LEON, F., and SEMLYEN, A.: ‘Complete transformer model
circuited, the potential developed at a further point from for electromagnetic transients’, IEEE Truns., 1994, PWRB9, (I), pp.
the grounded neutral may exceed the critical breakdown 231-239
9 CHIMKLAI, S., aiid MARTI, J.R.: ‘Simplified three-phase trans-
potential and lead to failure of the tap-changer winding. former model for electromagnetic transient studies’, IEEE Trans.,
1994. PWRIHO. (3). pp. 1316-1325
5 Conclusion 10 PEDERSEN, A.: ‘On‘ the response of interleaved transformer wind-
ings to surge voltages’, Trans. AIEE, Part IIL 1963, 82, pp. 349-356
A new lumped-parameter modelling technique for multi- 11 VAN N W S . R.: ‘Interleaved higli-voltage transfoniier windings’,
IEEE Trrms.. 1978, PAS-97: ( 5 )
winding power transformers has been proposed and diag- 12 ROSA, E.B.: ‘Calculation of the self-inductance of single-layer coils‘:
nostic studies on certain types of winding techniques have B. S. Bull., 1906, 2, pp. 161-187
been made. Indirect proof in favour of validity of the mod- 13 LYLE, : ”, Phil. Mug., 1902, 3, () pp. 310
14 GROVER, F.W.: ‘Inductance calculation: Working fonnulas and
elling technique and reliability of the results presented have tables‘ (Dover, 1962)
also been furnished in the form of comparison of the neu- 15 WIRGAU, K.A.: ‘Inductance calculation of an air-core disk winding’:
tral currents obtained from the model and actual test. The IEEE Trans.. 1976, PAS-95, (1)
EMTP simulation results show significant improvement in 16 STEIN, G.M.: ‘A study of initial surge distribution in concentric
transformer windings’, IEEE Trans., 1964, PAS83, pp. 877-893
linearity of potential distribution along the length of the
windings from line terminal to ground. A reduction in the 7 Appendix
oscillatory behaviour of the main winding is also observa-
ble. Reduction in the magnitude of potential developed 7.I Self inductance of disc coils
across the winding discs is also significant. All these The formula used here applies to thick coils of disc shapc,
improvements in the surge performance signify the merit of for which the radial dimension is considerably greater than
the interleaved disc winding over conventional disc wind- the axial dimension [12, 141. The equivalent self inductance
ing. is given by
It has been shown that high series capacitance, which is
the cardinal characteristic of interleaved winding, is not a L = L, - 0.004 x R x N x u ( G ~ Hi)+ (1)
guarantee of low impulse stresses everywhere in a sectional- where
ised winding. Rather, it has been observed that the degree
of homogeneity in the high-voltage windings has a very L , = 0.001N2aP ILH(ain cni) (2)
vital role to play. In a multiwinding transformer the inter- and
connecting region of two discrete winding sections may
experience an abrupt voltage multiplication due to mis-
match in the impedance offered by two dissimilar winding
sections to the propagating surge wave. The phenomenon
is more prominent in an interleaved winding than in a con-
ventional disc winding. Two correction terms are
Interleaved winding has a lower resonant frequency GI = l n ( B + C ) / p In e + (4)
owing to high series capacitance. Therefore part winding
resonance is more likely to occur in response to input pulse and

+2 [( N );
excitation in an interleaved winding compared to a conven- 1.
tional disc winding. Resonance phenomena were shown in HI = H - 111 k + -12
((p2 -72)
the interleaved tap winding due to incident standard light-
ning impulse at the line terminal. This is a very critical situ- (111 $ + 1.270)
ation where the voltage across the resonating discs becomes N
an almost undamped high-frequency oscillation of consid-
erable magnitude, the peak value of the wave being practi-
cally limited by the winding resistance. High voltage
stresses were also observable at the earthed end and open
part of the tap winding. The condition will be of greater
concern when the transformer is operated at -10% tapping, where H depends on the number of turns N in the disc coil
thereby making the full portion of tap winding open- and its value is obtained from [14]. /3 = Blp, y = C/p, value
circuited. In such a case the voltage stresses developed may of In k is also obtained from [14]. All other nomenclatures
be sufficiently high to cause failure of the tap winding. are given along with Fig. 6.
IEE Pioc -Eleclr Poiiei Appl Vol 147 No 3 Mal 2000 171
 If on the other hand the second coil has its radial dimen-
L
I
O !
I
-JCL sion c2 greater than axial b2 the coil is to be replaced by
I two coplanar circular filaments 33' and 44' located at the
i median plane but having equivalent radii r2 + 6 and r2 - 6,
! respectively, where

I
N turns I
I
t
I 1
7-2 = A (1 + g) (9)

I and
Fig.6 Schrmutic diugrum of disc coil
U = mean radius of disc coil (10)
P = winding pitch
C = thickness of insulation
B = height of insulated conductor The mutual inductance between the two coils is then given
N = tums in disc coil
c = Np = approx. radial width of disc coil by the formula

7.2 Mutual inductance between elements

The mutual inductance between two coaxial circular
fdaments of negligible cross-sectional area, and radii a and The mutual inductance between the fdaments is calculated
A , respectively, separated by distance d between their by eqn. 6.
planes is found to be dependent on the two parameters: alA
Table:
and dlA, and is given by
Filaments Product of turns Radii Axial spacing

11'&33' (NIxN2)/4 . rl and (r, + 6) D + /3

in whichfis obtained from [13] as a function of the varia-
ble 1 1 ' & 44' (NI x N2V4 rl and (r2- 6) D +p
I 22' & 33' (NI x N2)/4 rl and (r, + 6) 0-p
22' & 44' (NI x N2)/4 r, and ( r 2 - 6 ) 0-p

7.2.2 Determination of equivalent inductance of

The formulas given apply only to circular filaments of neg- coil: In all lumped-parameter transformer models the rep-
ligible cross-section. resentation of numerous mutual inductances formed by
I 0 ! one coil with other coils on the same limb is virtually
impossible and therefore distributed mutual inductances are
lumped together to form a single equivalent mutual induct-
ance, which when added to the self inductance of a disc coil
gives the equivalent lumped inductance of the disc coil con-
cerned. Assuming a section of the winding having n identi-
-L- ' I I
cal coils each of self inductance L and mutual inductances
- k4.a
tl,*;
b2

-+
.e!+
I

j 4 ;
F--
L
I
!
! A
L' i 3'
!
r!
M ibetween pairs of coils, where i indicates the number of
coils away from the reference coil, the equivalent induct-
ance of the whole winding section is given by the formula
n-1
Fig.7 Redctwn of coils to eyuivulentjlmnents

7.2.I Lyle method of equivalent filaments: This is

very accurate method for coaxial coils of dimensions such Moreover, for two distinct winding sections which are not
that the fourth and higher-order differential coefficient in too far away from each other (e.g. main and tap-changer
Taylor's series expansion is negligible. The dimension of windings of a transformer) mutual inductances among the
equivalent filaments in any case is illustrated by Fig. 7, elements of two windings need to be taken into considera-
which shows two circular coils of rectangular cross-sections tion. The method proposed by K.A. Wirgau [14] has been
of mean radii a and A , axial dimension bl and b2, radial used in the present case. This method replaces all the ele-
dimensions cI and c2, having number of turns N1 and N2 ments within one winding section by an equivalent lumped
and spacing of median planes D.The Lyle method replaces element. The proposed method has been illustrated in
the two coils by four equivalent filaments. Each filament is Fig. 8. Here two winding sections have been considered,
assumed to have half the number of turn of its coil. If the the first having Xelements each of N1 turns, and the sec-
axial cross-sectional dimension bl is greater than radial c1, ond having Y sections each of N2 turns have been replaced
filaments 11' and 22' will have an equivalent radius r1 by equivalent lumped elements of XNl and YN2 number
slightly larger than the mean radius a, and the two fda- of turns, respectively. Distance D between the elements is
ments are located at an axial distance p on either side of same as the distance of separation of the central elements
the median plane. The defining equations for rI and @ are of the two winding sections. Mutual inductance calculated

+ &)
between two such lumped elements gives a result with 4%
7-1 = a (1 (7) accuracy in most of the practical cases.

and 7.3 Series capacitance of interleaved disc coils

The series capacitance of a disc coil is composed of two
parts, being the resultant of interdisc capacitance and inter-
turn capacitance.
172 IEE Proc.-Eiectr. Power Appl.. Vol. 147, No. 3, May 2000
 -c,,vz
1
2
+
= [;ct(;)2x21] [+(;. V)2x21]

(15)
C, is calculated at mid-turn of a disc coil, applying the for-

I (X.3)
1 mula for parallel-plate capacitors with electrodes separated
by paper dielectric.

I -7---
I
I
I
I

hl
I
I
I
I
A
-J--- I I
I I M M F distribution
h 1 1

Fig.8 Mutual d c t m c e between two distinct winding sections 2- W

! I
CI
I I

7.3. I Determination of average interdisc capaci- Fig.9 MMF dutrrbutwn M lmninated conductorsplaced m slot
tance: For calculating the capacitance, the voltage is
assumed to be evenly distributed among the turns w i t h a 7.4 Equivalent high-frequencyresistance of disc
disc coil. The calculation of resultant interdisc capacitance coils
is based on the principle that sum of the energies accumu- Fig. 9 shows a case of subdivided (laminated) conductors
lated in all the part capacitances within a section is equal to placed in iron slot. It is assumed that the conductor is
the entire energy of the section. Let Cdr = resultant inter- divided into N layers, each of height h l , width b and length
disc capacitance between two interleaved members and C, L. Total height of all the layers is h = N hl. The average
= resultant interdisc capacitance between two adjacent loss ratio for N layers is given by
interleaved groups. Since there are 22 turns in a main wind-
ing disc coil, each interleaved disc pair consist of four turns.
Assuming a linear distribution voltage among the turns,
and applying the energy principle where RDcis the DC resistance, and R A C is the equivalent
1 1 resistance at high frequency, considering the skin effect,
-CdrV2 = -c@ and
2 2

PW
where f is frequency and p is the resistivity of the conduc-
tor. The generalised expression can be extended for a trans-
former winding if each winding of axial height Lc can be
considered to be located in a slot of width H,, which is the
height of the transformer window. Comparing Fig. 10 for a
transformer with that of laminated conductors placed in
Sunilarly slots, as in Fig. 9, eqn. 17 can be modlfied as

+ 32 + 52 + . . . + 212)+
(12
(452 + 472 + 49’ + . . . + 65’)
(14)
1
c d and C, are calculated at mid-turn of the disc coils, lbt-
applying the formula for capacitance between parallel-plate
electrodes, assuming that pressboard covers 35% of the
total horizontal surface area of a disc coil and oil covers the
remaining 65%.
HW

7.3.2 Determination of interturn capacitances:

For the main winding discs there are 22 turns per disc.
Therefore there will be 42 such capacitors, each of average
value C, which hold the voltage V. Assuming uniform dis- -
tribution of voltage among the turns it is observed that 21
such capacitors have a voltage V (22/44) i.e. V/2 across
L N layers 1
them, whereas rest of the 21 capacitors have a voltage I/ Fig. 10 Trmfitmer wkhgplaced in iron window
(2 1/44) across them. Applying the same energy principle, b = radial thickness of one conductor
H = height of transformer window
and assuming C, is the resultant interturn capacitance L, = axial height og winding

IEE Proc.-Electr. Power Appl., Vol. 147. No. 3, May 2000 173
 neutral-current waves

-'"I
- 20

-301 I I I I

0 20 40 60 80 100
time,s XIO-6
b
Fig. 11 Compurison of actual rmdsimulated neutrularrent waves
a Neutral current wave of original transformer obtained during impulse testing
b Simulated neutral current wave obtained from the developed model

HV main winding
parallel path 2 parallel path 2 parallel path 1 parallel path 1
disc 11-46 disc 1-10 disc 1-10 disc 11-46

Fig. 12 Schematic diugrum of developed e q u i d n t circuit model

-1 55
z.m
E
-IO*/. tapping

174 IEE Proc.-Electr. Power Appl., Vol. 147, No. 3, May 2000

View publication stats