This is the author's version of an article that has been published in this journal.

Changes were made to this version by the publisher prior to publication.

The final version of record is available at http://dx.doi.org/10.1109/TMAG.2015.2481004

CMP-296 1

Transformer Windings’ RLC Parameters Calculation and

Lightning Impulse Voltage Distribution Simulation
Tomislav Župan1, Bojan Trkulja1, Roman Obrist2, Thomas Franz2, Bogdan Cranganu-Cretu3, and Jasmin Smajić2
1
University of Zagreb, Faculty of Electrical Engineering and Computing, Unska 3, 10000 Zagreb, Croatia
2
University of Applied Sciences of Eastern Switzerland (HSR), Oberseestrasse 10, Rapperswil, Switzerland
3
ABB Switzerland Ltd., Badenerstrasse 780, 8048 Zürich, Switzerland

This paper presents a numerical method for calculating resistance, inductance and capacitance matrices of transformer windings.
Importance of their precise calculation is shown in the simulation of voltage distribution over the windings for lightning-impulse test.
The results obtained in frequency domain analysis are in a good agreement with the measurement data. All the parameters are
calculated using the self-developed solvers, the theory and novelty of which are described in this paper. The presented approach allows
fast and accurate high-frequency modeling of transformer windings.

Index Terms—coils, electromagnetic transients, numerical simulation, transformers.

frequency signal, a detailed equivalent circuit of the windings

I. INTRODUCTION has to be made. Lumped circuit model of winding is most

S INCE POWER AND DISTRIBUTION TRANSFORMERS are

electromagnetic devices built to operate for a couple of
decades, it is of paramount importance to design them to
withstand the high-voltage and high-frequency (HF) stresses
that are bound to happen during their lifetime. The lightning-
impulse (LI) test is one of the factory tests devised to proof the
HF (impulse) behavior of a transformer [1]-[2]. It shows the
influence of the winding capacitances, which are usually
negligible during the steady-state (low frequency) operation.
The LI-test can be simulated numerically in order to accelerate
the design process of the transformer and thus to reduce its
development costs [3].
This paper focuses on presenting the methods for
calculating the RLC parameters that are self-developed and
have acceptable accuracy compared to professional software.
The frequency domain analysis using the obtained RLC
parameters is used for LI-test simulation. The proposed
approach is verified using the ABB RESIBLOC® dry-type
transformer. Compared to the existing time-domain methods
and techniques for HF modeling and simulation of power and
distribution transformers recently reviewed in [3], the original
contribution of this paper is manifold: (a) application of the
self-developed BEM based field solver for evaluating the
capacitance matrix in order to avoid surface mesh of a
complicated winding cross-section, (b) evaluation of the
frequency dependent winding resistances by using a self-
developed integral equation based field solver, and (c) LI-
simulations in frequency domain.
II. RLC PARAMETERS CALCULATION
To accurately model the transformer’s response to a high-
Manuscript received July 6, 2015; revised September 1, 2015; accepted
September 18, 2015. Date of publication July 10, 2015; date of current
version July 31, 2015. Corresponding author: T. Zupan (e-mail:
tomislav.zupan@fer.hr).
Color versions of one or more of the figures in this paper are available
online at http://ieeexplore.ieee.org.
Digital Object Identifier (inserted by IEEE).
frequently applied in calculation of voltage distribution [3]-
[5]. Parts of windings are represented by equivalent
capacitances, inductances and resistances. Depending on the
way of the connection of the conductors, an equivalent circuit
is made for each single turn of the transformer coil. The
resistance, inductance and capacitance matrices are obtained
using the methods briefly described in the subsequent sections.
A typical transformer window is visible (simplified) in Fig.
1. Low and high voltage windings are circularly wound
around closed ferromagnetic core. Conductors are typically of
rectangular cross section insulated by paper insulation and
grouped in layers with radial and/or axial canals for heat
transfer purposes.
FIG. 1 HERE
A. Capacitance calculation
As opposed to the FEM based approach described in [3],
the method used for calculating the capacitance matrix is
based on the boundary element method (BEM) and the 2D
axially symmetric approximation [6]. This yields a more
efficient simulation procedure since the BEM discretization is
done only at material interfaces and boundaries thus
effectively reducing the order of the mathematical model by
one dimension. The tiny conductors’ dielectric insulation is
taken into account and every turn of the coil is regarded as a
separate modeling entity. The transformer core is grounded.
Electromagnetic field in transformers is transverse and the
capacitances may be computed on the basis of electrostatic
analysis. The electric field in a space consisting of conducting
regions at known potentials and regions filled with different
dielectrics can be solved by a pair of coupled integral
equations [7]:

0018-9464 © 2015 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. (Inserted by IEEE.)

Copyright (c) 2015 IEEE. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing pubs-permissions@ieee.org.
 This is the author's version of an article that has been published in this journal. Changes were made to this version by the publisher prior to publication.
The final version of record is available at http://dx.doi.org/10.1109/TMAG.2015.2481004

CMP-296 2

N






for the unknown magnetic field H k at the center of each
ϕ0 ( r ) = ∑ ∫
k =1
σk P(r , r ′)dC ; r ∈ C0
element k of the prisms within the ferromagnetic core.
∆C
(1) The self- and the mutual inductances are then determined
N

ε −ε



σ (r ) = 2 out in
ε out + ε in ∑σ ∫
k =1
k DN (r , r ′)dC ; r ∈ Ci . by computing an integral of the vector magnetic potential over
the path of a conductor. The contribution of the ferromagnetic
∆C
Here, C0 is the boundary at known potential φ0 and Ci is the core is calculated using the mentioned analytic formulation [8]
whereas the contribution of the conducting ring in air is more
interface between two dielectrics, ε out and εin are the outer
conveniently calculated from [9]. Fig. 2 shows the obtained
and inner region’s permittivity, N is the total number of capacitance and inductance matrices of the test transformer.



discretization elements and ∆C ∈ C0 ∪ Ci . Terms P (r , r ′) and As opposed to the FEM technique presented in [3], the



DN (r , r′) represent the Green’s function of the Laplace suggested method accurately takes into account the magnetic
electrostatic problem and its normal derivative, respectively influence of the transformer’s yoke, which is a 3-D feature of
[6]. A linear equation system for unknown coefficients σi is the transformer.
obtained by enforcing an exact solution at midpoints of each
finite segment (point-matching). The integrals in (1) are
computed numerically. If the point of interest is on ∆C, the FIG. 2 HERE
integrals become singular. In such a case their vicinity is
treated separately and this contribution is calculated
analytically (logarithmic singularities). C. Resistance calculation
Conductor boundaries and interfaces between two
Contrarily to the previous parameters, the winding’s
dielectrics are discretized into finite segments and the
resistance is considerably frequency dependent due to the
unknown distribution of surface charge density σ on the i-th
skin- and proximity effect. Since the core is positioned
segment is considered constant [6]. After the surface charge
relatively far away from the conductors, its influence on the
distribution on conductors is obtained, the capacitance matrix
resistance calculation is negligible.
is calculated by:
The resistances of the conductors are calculated with the
Nj
Qij vector magnetic potential formulation and integral equations
Cij =
ϕi − ϕ j
; Qij = ∫∫ σ dS = ∑σ
Sj
j j
k =1
kj S kj , i ≠ j. (2) approach. A 2D axially symmetric approximation is used.



Vector magnetic potential A at the point r from n current
Here, ϕi > 0 and ϕ j = 0 are potentials of i-th and j-th densities in elements Ωi , i = 1,..., n is [10]:
conductor, Qij is the total charge on the j-th conductor




µ0 n J i ( r ', t )


influenced by the charge on the i-th conductor, Nj is the
number of finite segments on j-th conductor, σkj and Skj are the
A(r ,t ) = ∑∫
4π i =1


dVi + A0 ( r , t ) .
r −r '
(3)
Ωi
surface charge density and the surface of the k-th segment of

The current density J i of element i is determined from the
j-th conductor, respectively.

The numerical results have been thoroughly compared with uniform current density of J e,i produced by the external
professional software Infolytica ElecNet and Ansys Maxwell

current source and eddy current density J eddy ,i :
used as benchmarks. A good agreement has been found

between results with the maximum overall difference less than


dA
3%. J eddy ,i = J i − J e,i = −κ , i = 1,..., n (4)
dt
B. Inductance calculation where κ is the electric conductivity. A combination of (3)
The inductance matrix is calculated using the 3D and (4) by taking into account the axial symmetry of harmonic
magnetostatic moment method. When performing the currents yields:


lightning impulse test on a transformer, the secondary winding
µ0 n
J i ( r ')

∑∫
d
J j ( r ') + κj

dVi + J j 0 ( r ' ) = 0, j = 1,..., n (5)
is shorted. This, as a consequence, weakens the influence of 4π dt i =1 r − r '
Ωi
the core’s nonlinearity. Thus, the frequency dependence for the
inductance calculation can be omitted and the problem can be dA j ,0
considered linear. where function J j 0 = − J e, j + κ can be determined from
dt
The iron core is divided into elements with uniform



the total current in element j, taking into account A0 ( r , t ) = 0 :
magnetization vector M . For a uniformly magnetized prism,


the equation for magnetic field is reduced to analytic
formulation [8]. ∫
i j = J j (r ')dS j , j = 1,..., n (6)

Sj
To solve for the unknown magnetization M over elements Each conductor is non-uniformly discretized in rectangular
of the ferromagnetic core, a system of equations is assembled elements (Fig. 3). The frequency and amplitude set are
previously determined from the known input wave form. The

Copyright (c) 2015 IEEE. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing pubs-permissions@ieee.org.
 This is the author's version of an article that has been published in this journal. Changes were made to this version by the publisher prior to publication.
The final version of record is available at http://dx.doi.org/10.1109/TMAG.2015.2481004
CMP-296
system of equations is then solved for the unknown surface
current densities on rectangular elements for each frequency.
The point matching technique is applied and both skin and
proximity effects are taken into account. Finally, resistances of
each of the m conductors are calculated from the Joule losses
in conductor:
1
2
Reff , j = 2
ij κ Ω∫J ( r ') dV , j = 1,..., m. (7)
j {
FIG. 3 HERE
D. Calculation time
Calculation of the RLC parameters was done on a standard
computer with dual-core processor. Resulting overall time was
15h using Infolytica ElecNet and MagNet, 33h using Ansys
Maxwell (using only one core because of the license
limitations) and 7h using the presented method.
To give an objective calculation time comparison of the
presented approach with the professional software is not
straightforward. Firstly, all the models done in Infolytica and
Ansys are 2D and the presented inductance calculation method
is 3D. Secondly, being self-developed solvers for scientific
purposes, implementation of the shown methods is not
sufficiently time-optimized. On the other hand, all the solvers
can be easily parallelized on multi-core computers. The
parallelization possibility along with the additional solver
optimization could result in the significant computation time
reduction.
III. MEASUREMENT AND SIMULATION RESULTS
ABB RESIBLOC® dry-type transformer with additional tap
connections that allow voltage measurement along the
winding is used to verify the presented approach. The
measurement results of the lightning-impulse test are shown in
Fig. 4.
FIG. 4 HERE
FIG. 5 HERE
A. Simulation method
The numerical method used for simulation results takes into
account all capacitive and inductive couplings between every
single turn of the winding. An example of such equivalent
circuit for a coil with four turns is depicted in Fig. 5.
Therefore, L and C matrices are dense. A given input signal is
transformed to frequency domain and the resistance matrix is
calculated for every dominating frequency.
After obtaining RLC matrices, a system of algebraic circuit
equations is solved for every frequency point using Wolfram
Mathematica:
Copyright (c) 2015 IEEE. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing pubs-permissions@ieee.org.
3
   R (ω )  [ K ]   I   a 
 jω [ L ] [ 0]  −      
{}
 { }
 
  [ 0 ] [ C ]  −  K T    =  , (8)

    [ 0]    U  {b} 
 
  { }
where K is a topological matrix dependent on the conductors’
{} {1
connecting scheme, I = I ,..., I N } { } {
and U = U ,...,U 2 N }
are the vectors of the unknown turn currents and voltages,
}
{a} = U1 , 0,...., 0 and {b} = jωU1 {C21 , C31 ,..., CN 1} are the
source terms and N is the number of turns. Results are
transformed back to time-domain and shown in Fig. 4.
B. Results
As one can see in Fig. 4, the suggested numerical
simulation procedure gives satisfactory results. The main
oscillation of the winding has a frequency of 560 kHz
according to the measurements and 567 kHz obtained using
the simulation, resulting in 2% difference.
In order to get more results for verification of the method, a
comparison with measurements has to be done across the
entire winding over the entire LI-time. The main focus is on
determining the voltage peak of the first layer of the winding,
between input lead-out “11” and lead-out “F”. Obtaining the
deepest voltage dips along the winding is helpful in seeing the
nonlinear voltage distribution throughout the turns of the
winding. This is necessary for determining the insulation
stresses between turns and layers of the winding. Fig. 6 shows
the measured and simulated voltage distributions for the entire
winding at the time of 4.2 µs.
FIG. 6 HERE
The obtained measured and simulated results for the three
deepest voltage dips for the first layer are shown in Table 1.
As can be seen, both the value of the voltage dip and the time
at which it happened can be simulated with an acceptable
accuracy.
TABLE 1 HERE
IV. CONCLUSION
It has been demonstrated that the presented method for
obtaining the RLC parameters of the transformer winding
using the self-developed solvers can be effectively used as a
fast way of simulating the high-frequency behavior of power
and distribution transformers. The presented comparison of the
obtained numerical results against the measurements revealed
a good agreement acceptable for daily design.
ACKNOWLEDGMENT
This work has been supported in part by the Croatian
 This is the author's version of an article that has been published in this journal. Changes were made to this version by the publisher prior to publication.
The final version of record is available at http://dx.doi.org/10.1109/TMAG.2015.2481004

CMP-296 4

Science Foundation under the project number IP-2013-1118.
REFERENCES
[1] G. Liang, H. Sun, X. Zhang, X. Cui, “Modeling of transformer windings
under very fast transient overvoltages” IEEE Trans. Electromagn.
Compat., vol. 48, no 4, Nov. 2006.
Fig. 1. Winding layout in a typical transformer.
Fig. 2. Calculated C and L matrices for the ABB RESIBLOC® transformer.
[2] M. Popov, L. van der Sluis, R. P. P. Smeets and J. L. Roldan, "Analysis
of very fast transients in layer-type transformer windings", IEEE Trans.
Power Del., vol. 22, no. 1, pp. 238-247, Jan. 2007.
[3] J. Smajic, T. Steinmetz, M. Rüegg, Z. Tanasic, R. Obrist, J. Tepper, B.
Weber, M. Carlen, “Simulation and measurement of lightning-Impulse
voltage distributions over transformer windings,” IEEE Trans. Magn.,
vol. 50, no. 2, 7013604, Feb. 2014.
[4] Y. Shibuya, T. Matsumoto and T. Teranishi, "Modelling and analysis of
transformer winding at high frequencies", International Conference on
Power Systems Transients (IPST'05), Montreal, Canada, Jun. 2005.
[5] K. Pedersen, M. E. Lunow, J. Holboell and M. Henriksen, "Detailed
high frequency models of various winding types in power transformers",
International Conference on Power Systems Transients (IPST'05),
Montreal, Canada, Jun. 2005.
[6] T. Župan, Ž. Štih, B. Trkulja, "Fast computation of inductances and
capacitances of high voltage power transformer windings," Proceedings
of the 15th International IGTE Symposium on Numerical Field
Calculation in Electrical Engineering, pp.137-143, Graz, Austria, Sep.
2012.
[7] Ž. Štih, “High voltage insulating system design by application of
electrode and insulator contour optimization”, IEEE Trans. Electr.
Insul., vol. EI-21, no.4, Aug. 1986.
[8] O. Chubar, C. Benabderrahmane, O. Marcouille, F. Marteau, F.J.
Chavanne, P. Elleaume, "Application of Finite Volume Integral
Approach to Computing of 3D Magnetic Fields Created by Distributed
Iron-dominated Electromagnet Structures," Proceedings of EPAC 2004,
Lucerne, Switzerland, pp. 1675-1677, 2004.
[9] T. Župan, Ž. Štih, B. Trkulja, "Fast and Precise Method for Inductance
Calculation of Coaxial Circular Coils with Rectangular Cross Section
using the One-Dimensional Integration of Elementary Functions
Applicable to Superconducting Magnets", IEEE Trans. App. Supercond.,
vol. 24, no. 2, 4901309, Apr. 2014.
[10] I. Doležel, "Self-inductance of an air cylindrical coil", Acta Technica
ČSAV, vol. 34, no. 4, pp. 443-473, 1989.

Fig. 3. Non-uniformly distributed discretization of m=4 conductors.

Fig. 5. Equivalent circuit of a four-turn transformer coil.

Fig. 6. Voltage distribution over the high voltage winding of the ABB
RESIBLOC® dry-type transformer at the time of 4.2 µs.
TABLE 1
MEASURED AND SIMULATED THREE DEEPEST VOLTAGE DIPS FOR THE FIRST
LAYER OF ABB RESIBLOC® DRY-TYPE TRANSFORMER

Voltage Time Voltage Time

No. Method
(V) (µs) difference difference
Meas. 0.317 2.56
1
Fig. 4. Simulated (top) and measured (bottom) voltage at different high Simul. 0.298 2.59 -6% 1%
voltage winding positions during the lightning-impulse test for the ABB Meas. 0.309 4.27
2
RESIBLOC® dry-type transformer. Lead-out 11 is the beginning of the Simul. 0.285 3.80 -8% -11%
winding, numbered lead-outs are distributed along layer 1, lettered lead-outs Meas. 0.284 8.03
3
are distributed along other layers Simul. 0.270 7.84 -5% -2%

Copyright (c) 2015 IEEE. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing pubs-permissions@ieee.org.