2042 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 23, NO.

4, OCTOBER 2008

High-Frequency Modeling of Power Transformers for

Use in Frequency Response Analysis (FRA)
Nilanga Abeywickrama, Student Member, IEEE, Yuriy V. Serdyuk, and Stanislaw M. Gubanski, Fellow, IEEE

Abstract—This paper presents an advanced model of the fre- the FRA spectrum are seen. One should realize however that it
quency response of a three-phase power transformer for use is not only the change in physical dimension of the transformer,
in conjunction with diagnostic measurements by the method of but also the characteristics of the insulation system (pressboard,
frequency response analysis (FRA). The model includes high-
frequency behavior of the laminated core and the insulation paper and oil) and the magnetic circuit components (steels of
through taking into account the frequency dependencies of the the core) that influence the distributed inductances and capac-
complex permittivity of insulation materials (paper, pressboard, itances, resulting in deviations of the frequency response eeee,
and oil) and of the anisotropic complex permeability of magnetic [4].
core and conductors. A lumped parameter circuit model is used Today, the FRA technique is essentially a comparative
to simulate the frequency response of open-circuit impedance,
short-circuit impedance, and impedance between primary and method, in which one performs the frequency response
secondary windings, in which the characteristics of circuit ele- measurements over a wide frequency range (typically
ments are calculated by means of the finite-element method. The Hz MHz) and compares the result with a fingerprint
effect of correct representation of each circuit element on the FRA measurement of the investigated unit or with other reference
response is analyzed and discussed in comparison to measurement responses (from a sister unit or another phase of the same
results on a real transformer.
transformer) [5]. A comparison of frequency response data is
Index Terms—Finite-element method, frequency response anal- usually done by graphical inspection or by using statistical
ysis, high-frequency modeling, power transformer diagnosis.
indicators, such as, for example, the correlation coefficient [6].
The interpretation of FRA results can, however, be improved
I. INTRODUCTION by introducing precise modeling of the high-frequency be-
OWER transformers are so designed that they can with- havior of transformer windings, which definitely may assist in
P stand mechanical forces appearing under system faults,
transportation and installation. However, lack of clamping pres-
identifying salient features of the winding responses, including
the effects of localized faults on the circuit parameters. This
sure or mechanical strength of insulating parts (e.g., pressboard requires detailed models of the windings, and attempts to make
spacers), resulting from natural aging, may lead to mechan- use of the lumped parameter circuit approach have already been
ical deformations, displacements of winding structure or even presented [7]–[9]. However, in most of these models, assump-
to electrical short-circuiting between turns and winding sec- tions are made that the effect of the core at frequencies above
tions. Measurement of leakage inductance had been the tradi- 10 kHz can be neglected in the calculations of inductance of the
tional method for detecting changes in winding geometry. How- windings and that the core surface remains equipotential when
ever, practical experience showed that only radial deformations calculating winding capacitances. Developing a transformer
can be detected when using this method, while localized axial model for small excitation signals, such as the ones used in
deformations are hardly observable since the latter introduces case of frequency sweep FRA measurements, that can take
only a small change in the leakage flux [1]. As an alternative, into account all the frequency-dependent phenomena in the
the frequency response analysis (FRA) technique, pioneered by insulation, in the laminated magnetic core and in the winding
Dick and Erven [2], has become increasingly popular for the as- conductors, has yet remained a challenge. The work pre-
sessment of mechanical integrity in power transformers. In this sented in this paper has therefore concentrated on modeling a
three-phase power transformer by means of the lumped
method, use is made of the fact that the shape of the response
parameter circuit approach including all of the aforementioned
at higher frequencies is uniquely determined by the geometrical
frequency-dependent material characteristics. Analytical for-
construction of the transformer. Since mechanical deformations
mulae [10], based on simplified geometrical representations or
change the capacitive and inductive parameters, deviations in
on numerical calculations, can be used for estimating the circuit
Manuscript received August 7, 2007. First published June 27, 2008; cur-
parameters. In this paper, the frequency-dependent diamagnetic
rent version published September 24, 2008. This work was supported in part effect of windings (represented in the form of a tensor) [11] and
by SIDA-SAREC of Sweden and in part by ABB/SECRC. Paper no. TPWRD- the frequency-dependent complex permeability tensor of the
00492-2007.
The authors are with the Division of High Voltage Engineering, Chalmers
core material that accounts for the characteristics of the joints
University of Technology, Gothenburg 412 96, Sweden (e-mail: nilanga. between yokes and limbs were both included into the induc-
[email protected]; [email protected]; stanislaw.gubanski tance and resistance calculations [12], which was implemented
@chalmers.se). in finite-element (FEM) software. The frequency-dependent
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org. complex permittivity of paper/pressboard insulation at different
Digital Object Identifier 10.1109/TPWRD.2008.917896 conditions (e.g., temperature and moisture content) was used
0885-8977/$25.00 © 2008 IEEE
ABEYWICKRAMA et al.: HIGH-FREQUENCY MODELING OF POWER TRANSFORMERS 2043

As shown in Fig. 1, the lumped parameter circuit model of a

transformer consists of capacitive, inductive, conductive, and
resistive elements. Here, the resistors in the inductive branches
represent dc losses, ac losses (skin and proximity effect), and
equivalent eddy current losses in the core, and all the circuit
parameters are frequency dependent. Although the implemen-
tation of such a passive network with constant circuit parame-
ters is straightforward in an ordinary circuit simulation package,
such as, for example, PSpice, especially the implementation
of frequency-dependent mutual inductance becomes a tedious
task. To overcome this difficulty, the lumped parameter circuit
model was instead implemented in Matlab as a set of coupled
Fig. 1. Lumped parameter circuit model of a two-winding transformer. differential equations that describe the physical behavior of the
circuit elements.
The circuit model of the two winding transformer (Fig. 1)
when calculating the capacitances and conductances. The im- consists of pairs of winding sections related to high-voltage
provements in the simulated frequency response are presented (HV) and low-voltage (LV) windings, resulting in
and compared with results of measurements of the winding nodes and inductive branches and associated capacitances,
impedances and impedances between primary and secondary conductances, and resistances. If one considers the current bal-
windings of a distribution transformer (400 kVA, 15/0.4 kV, ance at each node and voltage-current relationship across in-
). The measurements were performed by means of a net- ductive elements, it will yield a minimum set of variables [in-
work analyzer (MS 4360B, 10 Hz–300 MHz) and a reflection ductor current and nodal voltages ] in order to represent
bridge (MA 2401A, 10 Hz–30 MHz). the lumped parameter circuit shown in Fig. 1 as a multiport state
space model, in which the state vector consists of the nodal
II. MODELING A TRANSFORMER BY A LUMPED voltages and the inductor currents . Complete derivation
PARAMETER CIRCUIT APPROACH of the state-space model and the definition of the corresponding
In the lumped parameter circuit approach, sections of a matrices can be found in [3] and [10]. The resulting final matrix
winding, represented by their inductances, resistances, and ca- equation, representing the frequency response of all the nodal
pacitances, are lumped together. The size of the section should voltages and inductor currents, is expressed as
be small enough to assume that the current flowing through
the section is constant [13] and the lower limit of the size of a (1)
section is usually determined by the desired bandwidth of the
model and the geometry of the windings. At 50 Hz and up to where
a few hundreds of Hertz, the capacitive displacement current
is not significant, then only the induced electromotive force
(emf) and the ohmic voltage drop are taken into account so that
a winding can merely be modelled by means of its self-induc- and
tance, corresponding mutual inductances, and resistance. At
higher frequencies, this approximation is no longer valid and
the displacement current becomes significant, which ought to
additionally include capacitive couplings. All of the significant Here, and are the time derivatives of matrices and ,
displacement currents from a section to other sections or to respectively. and square matrices stand for effective induc-
conductive bodies have to be represented. A complete trans- tance (self and mutual) and resistance of the windings. and
former model can then be obtained by connecting a desired matrices are the capacitance and conductance matrices repre-
number of circuit blocks ( ) as depicted in Fig. 1, of which the senting capacitance and conductance between various parts of
circuit parameters are defined as follows: the windings and grounded bodies. The matrix connects the
current and voltage equations. It consists of and 0. and
earth capacitance of HV and LV windings; matrices consist of columns removed from and ma-
conductance to earth of HV and LV windings; trices upon applying external voltage to a node.
capacitance and conductance between HV and It is possible to obtain the transfer function of all the state vari-
LV windings; ables with respect to applied voltage by taking the Laplace
series capacitance of HV and LV windings; transformation of (1)
series conductance of HV and LV windings;
(2)
inductances of HV and LV windings;
resistances of HV and LV windings; where is the Laplace operator and is the identity matrix,
mutual inductance between th and th winding the latter has number of columns and rows equal to the size of
elements. the state matrix .
2044 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 23, NO. 4, OCTOBER 2008

Fig. 2. Real (a) and imaginary (b) components of the complex permittivity of
pressboard containing different amounts of moisture at 50 C.
Fig. 3. Single laminate of thickness 2b with anisotropic magnetic properties in
an external magnetic field.

In case of transformers, it can be considered as three

sets of single-phase units connected in either delta or star con- rical capacitances and corresponding effective relative permit-
figuration. The matrices described in modeling should be tivity values are in more detail elaborated in [3].
modified in order to account for the mutual coupling among the
windings on different limbs of transformer and connection of B. Effective Inductance and Resistance
winding ends. For a transformer, the number of columns
and rows in these matrices generally become three times larger When considering the large differences in typical dimensions
than that of case [3]. of power transformer elements, where the size of the whole core
is in meters and the conductors in the windings are usually in the
millimeter range, performing detailed and precise modeling of
III. CALCULATION OF CIRCUIT PARAMETERS
the electromagnetic processes in a transformer appears to be an
The capacitances and conductances of the lumped circuit extremely difficult task. Calculations that would simultaneously
model can be approximated reasonably well by means of include both elements in 3-D require very powerful computa-
analytical formulae, based on simplified geometrical considera- tional facilities and time. This disadvantage can, however, be
tions. The same approach is not possible to apply in case of the overcome by separately performing the analyses in two spatial
complex and anisotropic nature of a laminated core. Therefore, levels (i.e., in small and large scales). First, the effective com-
FEM-based software COMSOL Multiphysics was used to find plex permeabilities of the core components and windings are
the inductances and resistances of the model. calculated in the small-scale level (2-D) and then these effec-
tive parameters are used in a large-scale 3-D model to calculate
A. Capacitance and Conductance inductance and resistances, as will be described.
1) Effective Relative Permeability of Laminated Core: A
The capacitance of a physical system is defined by its geomet-
transformer core is made up of a number of electrical steel
rical capacitance and the relative permittivity of the dielectric
laminations stacked together. Each lamination is coated with
material used. In dielectrics, in addition to the ordinary displace-
an insulation layer in order to avoid interlaminar eddy current.
ment current, losses due to polarization and conduction phe-
If a single lamination (Fig. 3) is subjected to a time-varying
nomena appear. In the lumped circuit model, a parallel circuit
magnetic field (the symbol “ ” denotes the
representation was used for representing these losses. Deriva-
complex values and bold style is used for vector quantities),
tion of the parallel circuit admittance by means of its com-
there are three eddy currents , and generated in each
plex permittivity and geometrical capacitance
plane of the lamination because of the finite conductivity of
is as follows:
the silicon steel, due to Faraday’s induction phenomenon (see
Fig. 3). By assuming linear anisotropic magnetic permeability
and isotropic electric conductivity of a single steel lamination,
the constitutive relations for the material can be written as
(3)
and
where and represent the capaci-
tance and conductance of the parallel circuit model (e.g., where is the permeability tensor and stands for the elec-
and in Fig. 1). For transformers in which several dielec- trical conductivity (scalar) of the electrical steel. For the single
tric materials are used, the geometrical arrangement of the in- lamination shown in Fig. 3, the effective complex permeability
sulating components has to be taken into account when calcu- can be found using well-known formulae (see [15]) and its
lating its effective relative permittivity, including the relevant and components are defined as
complex permittivities of paper , pressboard , and oil
, which can be found from frequency response measure- (4)
ments on material samples [14]. Examples of such frequency
responses for pressboard containing different amounts of mois-
(5)
ture are shown in Fig. 2(a) and (b). The calculations of geomet-
ABEYWICKRAMA et al.: HIGH-FREQUENCY MODELING OF POWER TRANSFORMERS 2045

Fig. 5. Schematic view of the cross section of a mitered overlapping region cut
at an angle of 45 to the rolling direction of a limb and a yoke of the laminated
core, showing the calculation domain with boundary conditions and geometrical
dimensions.

Fig. 4. Effective complex relative permeabilities 

^ and  ^ of the laminated
magnetic core (real and imaginary parts) as a function of frequency according
to classical formulae (4) and (5).

where

and

In expressions (4) and (5), is skin depth and is the an-

gular frequency of the magnetic field. Since the core consists
of a stack of laminations, the effective area that carries mag-
netic flux becomes reduced, which can be accounted for by in-
troducing a so-called stacking factor ( means
Fig. 6. Effective complex relative permeability 
^ of a mitered overlapping
the thickness of a lamination with insulation layer included), by region.
which the permeabilities given in (4) and (5) should be mul-
tiplied to obtain the effective permeability of the bulk of the
core. The -component of the permeability is given by permeability of the mitered overlapping region can be estimated
[16]. Thus, knowing all components of by averaging the magnetic flux density over length upon ap-
the effective complex permeability, the permeability tensor can plying an external magnetic field as given in (7),
be expressed as [18]. Fig. 6 shows the resulting real and imaginary parts of

(6) (7)

2) Diamagnetic Effect of Conductors: Several effects be-

Computations were performed using the following parameters: come apparent when high-frequency alternating current flows
S/m, through a multiturn winding of a power transformer: skin effect
and the results are shown in Fig. 4. that confines the current to the conductor surface, proximity
Yet another problem related to the transformer core structure effect that influences current density distribution due to the
must be taken into account (i.e., different regions of the mag- magnetic fields generated from the adjacent conductors, and
netic core are characterized by different effective permeabili- geometric effects associated with the action of the transverse
ties). These include core limbs, yokes, and joints that are called magnetic field from the rest of magnetic system. All three listed
mitered overlapping areas. The approach discussed before can effects are interlinked together and they are dependent upon the
only be used for describing the bulk of the core. The mitered re- geometry of the windings and the structure of the magnetic core
gions formed by overlapping steel sheets require special consid- that dictate the magnetic-field distribution. Aforementioned
eration. A particular type of joint, called single-step-lap (SSL), phenomena produce a diamagnetic effect by repelling flux
is considered in this study (Fig. 5) [17]. The magnetic proper- from the volume of conductors. This can be represented by an
ties at the regions differ from the regions in yokes and limbs effective complex permeability of which the real part may be
due to the sudden change of the rolling direction. Since it is less than unity, reflecting the diamagnetic effect, and the imag-
too complex to solve this problem analytically, a cross section, inary part reflects the losses associated with skin and proximity
parallel to the direction of the magnetic flux path at the joint, effects. A portion of multiturn LV and HV windings can be
was considered and modelled by means of FEM, as depicted in represented as a system of regularly located solid conductors of
Fig. 5 where is the magnetic vector potential. The effective rectangular and circular cross-sections, respectively, together
2046 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 23, NO. 4, OCTOBER 2008

The governing equation in terms of can be derived from

Maxwell’s first equation (Ampere’s law) as

in
(8)
in

which can be solved numerically. Here, is the complex fre-

quency-dependent permeability tensor; is the unit vector in the
azimuthal direction of the winding; is the number of turns in
the winding; is the total cross-sectional area of the winding
section; and is the current per turn. The tensors of ferromag-
netic core permeability in (8) are

in
in (9)
in
in

Fig. 7. 3-D transformer model showing computational domains (one half of Definitions of permeability tensors are given in [11]. Having
the core is modelled). solved the 3-D model, one can estimate effective resistance and
inductance of the winding either from energy balance method or
by calculating the induced voltage in the winding. It was found
with air gaps and insulating layers. It is then assumed that the in the simulations that the both methods yielded the same re-
system of the conductors is infinitely long in the transverse di- sults. In the former case, the complex magnetic energy in all of
rection to the 2-D cross-sectional plane of the conductors so as the domains is calculated as
one needs to solve the electromagnetic problem in 2-D, which
can be represented as a composite heterogeneous medium with
the periodic structure of an elementary cell containing a unit
conductor. The effective complex permeability tensor of such
a medium can be evaluated by solving the field problem in
Here, “ ” denotes the complex conjugate. The parameters of the
this cell by applying an external unit magnetic field in ,
winding can be found from the complex impedance defined as
and directions. Since this analyzed 2-D unit cell is located
on a radial plane of a cylindrical coordinate system, the com-
plex permeability tensor should be transformed into Cartesian
(10)
coordinates of the large scale 3-D model. Formulation and
implementation of the models and calculation of effective (11)
parameters for the conductors are described in [11] and [12].
3) Large Scale 3-D FEM Model: In this study, a distribution where and are self inductance and resistance of the th
transformer (15/0.4 kV, 400 kVA, ) having a three-limb section, respectively, and is the magnetic energy when cur-
core was considered. Exact dimensions of the core and the wind- rent is applied to the th section of a winding. Note that the value
ings are given in [18]. Only one half of the core was modelled of in (10) represents the contribution of the losses in the core
by taking advantage of symmetry of the system. The problem and in the volume of the conductors. Mutual inductance
is defined within four domains as shown in Fig. 7: and and resistance between th and th winding sections can
are the sections of HV or LV windings, volume corresponds be calculated by using the parameters calculated in (10) and (11)
to medium surrounding the core, and is the volume of the and an additional solution for energy stored when current is ap-
laminated core. plied to both winding sections [19]
Since the frequency-dependent effects are already incorpo-
rated in the calculated effective permeability tensors for the (12)
core and the volume of the conductors, the problem can be for-
mulated in magnetostatic mode. This is an artificial approach, (13)
which avoids a high resolution mesh required to solve the
problem (in quasistatic mode) due to decreasing skin depth at
high frequencies. At the same time, this approach reflects the Here, is the total magnetic energy when currents and
main features of real physical picture of the electromagnetic are applied to winding sections and , respectively. Calcula-
processes. Then, the final solution is obtained by separately tions of self and mutual inductances and resistances of both HV
solving the magnetostatic problem for each frequency with and LV winding sections [according to (10)–(13)] on all three
corresponding values of effective complex permeabilities. limbs are implemented as a Script in COMSOL.
ABEYWICKRAMA et al.: HIGH-FREQUENCY MODELING OF POWER TRANSFORMERS 2047

Fig. 8. Mutual inductance between a disk of HV winding and a turn of LV Fig. 9. Measured and simulated frequency responses (with and without core
winding placed at the top of the middle limb. joint representation) of open-circuit impedance of an outer limb.

IV. RESULTS AND DISCUSSION

The developed model was utilized for simulating the fre-
quency response of open-circuit impedance and impedance
between primary and secondary windings. It appears that the
inclusion of material properties for steel laminations, insula-
tion, and conductors into the lumped parameter circuit model
improves and affects the simulated frequency response in dif-
ferent ways (i.e., changing damping and resonant frequencies,
etc.). The three mostly used spectral indicators (correlation
coefficient , absolute sum of logarithmic error (ASLE), and
spectrum deviation ) [6], [20], [21] in FRA characterize the
changes in frequency response and are defined as follows:

(14)

Fig. 10. Measured and simulated frequency responses (with and without dia-
(15) magnetic effect of conductors) of open-circuit impedance of an outer limb.

of diamagnetic properties of conductors starts showing up (see

Fig. 8), though this effect is very weak (maximum reduction of
inductance at 1 MHz is %). Figs. 9 and 10 show the compar-
(16) ison between measured and simulated frequency responses of
the open-circuit impedance of an outer limb.
In (14)–(16), and refer to the An improvement of the result of the simulation can clearly
data points of two frequency responses with number of fre- be noticed upon the inclusion of the core joint representation
quencies in the frequency range 100 Hz–1 MHz. (Fig. 9), which results in higher impedance below 1 kHz and im-
proved response at frequencies between 10 kHz and 100 kHz.
A. Effect of Characteristics of Core Joints and Diamagnetic As expected, less effect is seen upon the inclusion of the diamag-
Properties of Conductors netic characteristics (Fig. 10), which does not show any apparent
The introduction of the core joint representation results in an changes in the simulation. On the other hand, when the primary
increase of inductance at low frequencies, as depicted in Fig. 8, and secondary terminals are tied together separately in order to
showing that mutual inductance between a disk (HV winding) simulate the impedance between primary and secondary wind-
and a turn (LV winding) placed on the middle limb increased ings, the influence of the core joint does not show up because
by 7.8 % at frequencies below 1 kHz. On the contrary, the intro- only the leakage field is present outside the core. Furthermore,
duction of diamagnetic property of the conductors seems to have the diamagnetic effect of the windings is also not visible up to
almost no influence on the inductance at frequencies below 100 1 MHz. These influences are quantified by means of the spectral
kHz. However, at higher frequencies ( kHz), the influence indicators in Table I.
2048 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 23, NO. 4, OCTOBER 2008

TABLE I
RELATIVE CHANGES OF SPECTRAL INDICATORS UPON INTRODUCING JOINT
AND DIAMAGNETIC CHARACTERISTICS W.R.T. MEASURED RESPONSE

Fig. 12. Simulated frequency responses of open-circuit impedance of an outer

limb for two different moisture content at 80 C.

TABLE II
RELATIVE CHANGES OF SPECTRAL INDICATORS UPON INTRODUCING
DIFFERENT CONDITIONS OF INSULATIONS

Fig. 11. Simulated frequency responses of open-circuit impedance of an outer

limb for new and aged insulation at 80 C with 4% moisture content in paper/
pressboard.

B. Effect of Characteristics of Insulation Materials

Moisture content, aging, and temperature affect the relative imposed by the laminated structure and the presence of joints
complex permittivity of the materials used for insulation of in the core, by the diamagnetic property of the winding conduc-
power transformers [14] which, in turn, result in changing tors, and by the state of the insulation are presented. The asso-
the capacitive and conductive elements of the lumped circuit ciated losses are represented by respective changes of the com-
model. An example is shown in Fig. 11, where the effect of plex permeability of core material and the complex permittivity
aging influences the winding impedance by stronger damping of pressboard and paper. It was found that the correct represen-
of the resonances below 100 kHz. This effect appears because tation of the core characteristics (joints, laminated nature, etc.)
of the increased losses (higher imaginary part of complex per- is of paramount importance when calculating circuit inductance
meability) in the aged pressboard insulation [14]. The damping and resistances. It is the resistance (including winding resistance
can also be affected due to different amounts of moisture in and eddy current losses in the core) in the circuit that contributes
the pressboard, since higher moisture content (m.c.) results to most of the damping in the whole frequency range. However,
in not only in higher losses but also in increased real part of the damping of the resonant peaks and valleys at lower frequen-
permittivity [see Fig. 2(a)]. The effect is shown in Fig. 12, and cies ( kHz) is also affected by the condition of insulation
Table II presents the changes of the spectral indicators due to (aging, moisture content, etc.), whereas introduction of diamag-
aging, moisture content, and variation of the temperature. The netic properties of the conductors becomes noticeable above
indicators reveal that the change of moisture content affects the 1 MHz.
frequency response the most. Correct representation of the frequency-dependent parame-
ters of the core and the insulation materials leads to a better
fit of simulated responses to the measured ones. Especially the
V. CONCLUSION calculations of inductances could further be improved if cor-
A development of lumped parameter circuit model for use in rect values of local permeability of steel laminations were
FRA analysis is presented. The model is capable of simulating known, which was chosen to be equal to 500 in this study. Fur-
different types of frequency responses of and trans- ther improvement of the simulation should therefore rely on
formers. Calculation of circuit parameters accounting for all of concentrating efforts to measure permeability in a broad fre-
the frequency-dependent phenomena related to the influences quency range.
ABEYWICKRAMA et al.: HIGH-FREQUENCY MODELING OF POWER TRANSFORMERS 2049

REFERENCES [19] E. Bjerkan, “High frequency modeling of power transformers,” Ph.D.

[1] M. D. Nigris, R. Passaglia, R. Berti, L. Bergonzi, and R. Maggi, “Ap- dissertation, Norwegian Inst. Technol., Trondheim, Norway, Nov.
plication of modern techniques for the condition assessment of power 2005.
transformers,” in Cigre Session 2004, Paper A2-207, Paris, France, [20] J.-W. Kim, B. Park, S. C. Jeong, S. W. Kim, and P. Park, “Fault di-
2004. agnosis of a power transformer using an improved frequency-response
[2] E. P. Dick and C. C. Erven, “Transformer diagnostic testing by fre- analysis,” IEEE Trans. Power Del., vol. 20, no. 1, pp. 167–178, Jan.
quency response analysis,” IEEE Trans. Power App. Syst., vol. PAS-97, 2005.
no. 6, pp. 2144–2150, Nov./Dec. 1978. [21] J. Back-Jensen, B. Back-Jensen, and S. D. Mikkelsen, “Detection of
[3] N. Abeywickrama, “Effect of dielectric and magnetic material charac- faults and aging phenomina in transformers by transfer functions,”
teristics on frequency response of power transformers,” Ph.D. disserta- IEEE Trans. Power Del., vol. 10, no. 1, pp. 308–314, Jan. 1995.
tion, Chalmers Univ. Technol., Gothenburg, Sweden, Dec. 2007.
[4] K. G. N. B. Abeywickrama, Y. V. Serdyuk, and S. M. Gubanski,
“Effect of core magnetization on frequency response analysis (ERA)
of power transformers,” IEEE Trans. Power Del., vol. 23, no. 3, pp. Nilanga Abeywickrama (S’04) received the B.Sc.
1432–1438, Jul. 2008. degree in electrical and electronic engineering and
[5] S. Ryder, “Diagnosing a wider range of transformer faults using the M.Sc. degree in electrical power and energy from
frequency response analysis,” presented at the 13th Int. Symp. High the University of Peradeniya, Sri Lanka, in 2001 and
Voltage Engineering, 2003. 2003, respectively, and the Lic. Eng. degree from the
[6] S. A. Ryder, “Methods for comparing frequency response analysis School of High Voltage Engineering, Chalmers Uni-
measurements,” in Proc. IEEE Int. Symp. Electrical Insulation, versity of Technology (CTH), Göteborg, Sweden,
Boston, MA, Apr. 7–10, 2002, pp. 187–190. in 2005, where he is currently pursuing the Ph.D.
[7] E. Rahimpour, J. Christian, K. Feser, and H. Mohseni, “Transfer func- degree.
tion method to diagnose axial displacement and radial deformation of His research interests include high-frequency
transformer windings,” IEEE Trans. Power Del., vol. 18, no. 2, pp. modeling, high-frequency measurements, and diag-
493–505, Apr. 2003. nosis of power transformers.
[8] R. Ragavan and L. Satish, “An efficient method to compute transfer
function of a transformer from its equivalent circuit,” IEEE Trans.
Power Del., vol. 20, no. 2, pt. 1, pp. 780–788, Apr. 2005.
[9] M. Florkowski and J. Furgal, “Detection of transformer winding de- Yuriy V. Serdyuk received the M.Sc. and Ph.D.
formations based on the transfer function—Measurements and simula- degrees in high-voltage engineering from the
tions,” Meas. Sci. Technol., vol. 14, pp. 1986–92, Sep. 2003. National Technical University of Ukraine “Kiev
[10] K. G. N. B. Abeywickrama, Y. V. Serdyuk, and S. M. Gubanski, “Ex- Polytechnic Institute’ Kiev, Ukraine, in 1986 and
ploring possibilities for characterization of power transformer insula- 1995, respectively.
tion by frequency response analysis (FRA),” IEEE Trans. Power Del., In 1986, he was an Engineer, and became a
vol. 21, no. 3, pp. 1375–1382, Jul. 2006. Researcher at The Institute of Electrodynamics of
[11] A. D. Podoltsev, K. G. N. B. Abeywickrama, Y. V. Serdyuk, and S. the National Academy of Science of Ukraine Kiev,
M. Gubanski, “Multi-scale computations of parameters of power trans- Ukraine, in 1991. In 1996–1997, he was a Research
former windings at high frequencies, part I: Small-scale level,” IEEE Engineer in the R&D HV Laboratory, ABB High
Trans. Magn., vol. 43, no. 11, pp. 3991–3998, Nov. 2007. Voltage Technologies Ltd., Zurich, Switzerland.
[12] A. D. Podoltsev, K. G. N. B. Abeywickrama, Y. V. Serdyuk, and S. Currently, he is an Assistant Professor in the Division of High Voltage Engi-
M. Gubanski, “Multi-scale computations of parameters of power trans- neering, Department of Materials and Manufacturing Technology, Chalmers
former windings at high frequencies, part II: Large-scale level,” IEEE University of Technology, Göteborg, Sweden.
Trans. Magn., vol. 43, no. 12, pp. 4076–4082, Dec. 2007.
[13] P. Holmberg, “Modelling the transient response of windings, laminated
steel cores and electromagnetic power devices by means of lumped
circuits,” Ph.D. dissertation, Uppsala Univ., Uppsala, Sweden, 2002. Stanislaw M. Gubanski (M’89–SM’90–F’01) re-
[14] C. Ekanayake, “Diagnosis of moisture in transformer insulation—ap- ceived the M.Sc. degree in high voltage engineering
plication of frequency domain spectroscopy,” Ph.D. dissertation, and the Ph.D. degree in material science from the
Chalmers Univ. Technol., Gothenburg, Sweden, Jan. 2006. Technical University of Wroclaw, Wroclaw, Poland,
[15] J. Lammeraner and M. Stafl, Eddy Current.. London, U.K.: ILIFFE in 1973 and 1976, respectively.
Books, 1966. He was a Research Fellow at the University
[16] J. Gyselinck and P. Dular, “A time-domain homogenization technique College of North Wales, Bangor, U.K., from 1976
for laminated iron cores in 3-D finite-elements models,” IEEE Trans. to 1977 and a Senior Lecturer at the Technical Uni-
Magn., vol. 40, no. 2, pt. 2, pp. 856–859, Mar. 2004. versity of Wroclaw, Wroclaw, from 1977 to 1988.
[17] F. Loffler, T. Booth, H. Pfutzner, C. Bengtsson, and K. Gramm, He was Associate Professor at the Royal Institute
“Relevece of step-lap joints for magnetic characteristics of trans- of Technology, Stockholm, Sweden. Currently,
former cores,” in Proc. Inst. Elect. Eng., Electr. Power Appl., 1995, he is Professor in High Voltage Engineering, Department of Materials and
vol. 142, no. 6, pp. 371–378. Manufacturing Technology, Chalmers University of Technology, Göteborg,
[18] K. G. N. B. Abeywickrama, A. D. Podoltsev, Y. V. Serdyuk, and S. M. Sweden, and Research Leader of High Voltage Valley, Ludvika, Sweden. He is
Gubanski, “Computation of parameters of power transformer windings Chair of the IEEE-DEIS Nominations Committee. He is also Convener of the
for use in frequency response analysis,” IEEE Trans. Magn., vol. 43, no. CIGRE Task Force D1.01.14 “Dielectric Response Methods for Diagnostics of
5, pp. 1983–1990, May 2007. Power Transformers.”