F E A T U R E A R T I C L E

Impulse Voltage Distribution and

Frequency Response of Intershield
Windings
Key words: intershield winding, impulse voltage distribution, frequency response

Introduction Mehdi Bagheri

Impulse voltage distributions in continuous disk windings
are nonlinear, due to the small series capacitance of the wind-
Electrical Engineering Department,
ing. Consequently, transformer manufacturers often design large Nazarbayev University, Astana, Kazakhstan
power transformers with interleaved windings to increase series
capacitance, but in doing so reduce the value of the initial im- B. T. Phung
pulse voltage distribution coefficient α given by School of Electrical Engineering and
α = √(Cg/Cs), (1)
Telecommunications, University of New
South Wales, Sydney, Australia
where Cg is the shunt capacitance and Cs is the series capacitance
of the winding. Mohammad Salay Naderi
Sometimes fully interleaved windings are employed because Niroo Research Institute, Tehran, Iran
they are more reliable, ensuring that the transformer will oper-
ate properly under lightning strike. However, fully interleaved
windings require longer assembly times, and manual welding In transformers with intershield
is necessary. Manufacturers have therefore been using series
combinations of continuous and interleaved windings, known as windings, inserting shield turns be-
multiple-α type windings [1]. However, this approach also has tween winding conductors increases
disadvantages, e.g., the design and analysis of performance un-
der normal and abnormal transient conditions are complicated,
the series capacitance, and results in
and the impedances of continuous disks differ from those of in- a more linear impulse voltage distri-
terleaved disks [2], causing reflection of surge traveling waves at bution and fewer oscillations in the
connection points. Transformer designers may therefore have to
adjust inter-disk distances and repeat transient simulation stud- frequency response.
ies several times, to find an optimum arrangement.
Intershield winding (IW), sometimes called electrostatic-
shield disk type winding, is another option for designers and the winding. This increase will decrease the value of the initial
manufacturers of high voltage transformers. IW is based on sev- impulse voltage distribution coefficient, and hence provide more
eral United States–registered patents [3]–[7]; however, in our uniform initial voltage distribution along the transformer wind-
opinion IW as practiced in industry more closely resembles the ing. A uniform initial voltage distribution will lead to less stress
shield pattern introduced by Degeneff et al. [3]. Some manufac- on the upper disks of the winding. It has been claimed that, us-
turers have replaced IW with continuous and interleaved disk ing IW, the series capacitance of the winding will increase sig-
windings to reduce manufacturing time and increase the value nificantly. However, the calculation of the series capacitance and
of α. IW increases the series capacitance between conductors of the analysis of IW performance over different frequency ranges

32 0883-7554/16/©2016/IEEE IEEE Electrical Insulation Magazine

have not been widely discussed. This might be due to protection
of trade secrets associated with the IW construction process. A
valuable study of IW series capacitance was published by Del
Vecchio et al. [8]. It seems that transformer researchers subse-
quently paid less attention to this winding type. Transformer ser-
vice companies are therefore facing difficulties with IW repair
or overhaul. Technical information on IW is not readily avail-
able, contrasting with continuous disk or interleaved winding.
In this study, some challenges associated with IW design and
manufacturing are discussed, and the series capacitance is cal-
culated based on a practical approach. Four windings were fab-
ricated: one interleaved, one continuous disk, one IW with one
shield turn, and one IW with six shield turns in each disk. Initial
impulse voltage distributions along the winding bodies were
measured and compared, to guide the IW series capacitance cal-
culation. In addition, frequency response analysis (FRA) studies
of the four windings were carried out to observe and interpret
differences between the windings.

Voltage Distribution on Windings

A lightning strike may result in an impulse voltage stress on
a transformer winding, especially if protection devices do not
respond rapidly. To analyze this situation simply, a transformer
winding may be represented by a uniformly distributed circuit
of series inductance and shunt/series capacitance. When a unit
step function voltage is applied to this equivalent circuit, the ini-
tial voltage distribution is determined entirely by the capacitive
network. The initial voltage at any point of the winding may be
calculated by solving the differential equations for the network
and inserting the necessary boundary conditions. The final volt-
age distribution is determined by series resistance, or if there is a
superimposed power frequency excitation, by inductance. Mak-
ing a distinction between the two (series resistance and induc-
tance) is superfluous, since both quantities are usually linear and
the distribution therefore becomes a straight line corresponding
to α = 0.
To make the most efficient use of the insulation system within
the transformer, it is imperative to design windings in such a
way as to eliminate, or at least substantially reduce, the initial
voltage concentration at the line end, and subsequent high oscil-
latory voltages in the body of the winding. Any improvement,
i.e., greater linearity of the initial distribution, will decrease the
divergence between the initial and final responses of the wind-
ing. It will also reduce the magnitude of the subsequent oscil-
lations. It may therefore be sufficient to concentrate solely on
improving the initial voltage distribution in the windings. Since
this distribution is controlled entirely by the ratio of shunt and
series capacitances, reducing the shunt capacitance or increasing
the series capacitance can help to achieve a better distribution.
To analyze initial impulse voltage distribution, a transformer
winding can be modeled as a capacitance network. The initial
impulse voltage distribution for this network is given by [9]
U(x) = A1eαx + A2e−αx, (2)
where x denotes the distance relative to the winding end (neu-
tral). The constants A1 and A2 depend on the boundary conditions
Figure 1. Initial and final voltage distributions in a transformer
winding.
at the two ends of the winding. To avoid insulation breakdown,
it is necessary to improve the initial voltage distribution across
the winding, i.e., to reduce the maximum disk-to-disk voltage
gradient. In this regard, it is necessary to make α as small as pos-
sible [9]. A lower coefficient α may be obtained by reducing Cg
or by increasing Cs.
Figure 1 illustrates the influence of α on the initial voltage
distribution along a winding. It may also indicate the amount
of stress on a winding, especially on upper disks. Each winding
disk experiences very nearly the same stress when α approaches
zero.
Numerous studies have been conducted over the past three
decades, with the aim of decreasing α in various types of trans-
former windings [10], [11]. These studies have resulted in a
number of methods and practical techniques, some of which
have been applied by transformer manufacturers. Sequential
continuous, sequential interleaved, nonsequential continuous,
and nonsequential interleaved disk windings, as well as helical
and layer windings, have been used in transformers. The IW was
introduced three decades ago, and has been used over the last
20 years to increase winding series capacitance and decrease
the initial impulse voltage distribution coefficient, α. It has been
proven theoretically that IW exhibits lower series capacitance
than interleaved winding, and larger series capacitance than con-
tinuous disk winding. The larger series capacitance leads to a
more linear impulse voltage distribution along the IW, compared
with interleaved winding.
Series Capacitance of Various Transformer
Windings
To demonstrate the increased series capacitance and the im-
proved initial voltage distribution associated with IW, it is essen-

September/October — Vol. 32, No. 5 33

Figure 2. Cross-section schematic view of continuous disk wind-
ing and its equivalent capacitance network.

tial to analyze its behavior theoretically. This section presents

the calculation of series capacitance for continuous disk wind-
ing, interleaved disk winding, and IW.
Figure 3. Cross-section view of voltage distribution along a pair
Series Capacitance of Continuous Disk Winding of disks.
Figure 2 shows a continuous disk winding containing two
disks, and its equivalent capacitive network. The series capaci- where Cdd is the disk-to-disk capacitance. Ctt and Cdd may be
tance has two components, namely, the total series capacitance calculated as shown in (7) to (12). Figure 4(a) shows modified
Ct between the turns, and the total series capacitance Cd between insulation parameters based on [12], and Figure 4(b) shows the
the disks. To calculate the equivalent series capacitance in a con- structure of the disk-to-disk spacer.
tinuous disk winding, we apply the energy summation method.
According to this method, the sum of the energies stored in the Ctt = εrεoDπ[(h + 2δt)/(2δt)], (7)
capacitances associated with the two disks is the total energy
stored within those disks. It is assumed that the number of con- where h is the height of the conductor used for calculating turn-
ductor turns in each disk is N. The number of series capacitors to-turn capacitance, δt is the thickness of the inter-turn insula-
between turns, as shown in Figure 2, will be 2N − 2 for one pair tion, and εt is the permittivity of the paper insulation.
of disks.
It can be shown [12] that Ct is given by
(1/Cdd) = (2/Cdd–paper) + [1/(Cspacer‖Cdd–oil)], (8)
Ct = (1/2)Ctt[(N − 1)/N 2], (3) Dπε d + aw s (ε s − ε d )
C dd = Dπ w ε oε t , (9)
Dπε tδ d + 2δ t Dπε d + aw s (ε s − ε d )
where Ctt is the turn-to-turn capacitance and N is the number
of turns in each disk. Calculation of Cd is based on the voltage
distribution shown in Figure 3.
According to Figure 3, starting from conductor number 1 or
number 20, and moving toward the middle of the winding (con-
ductor number 10 or 11), the voltage will change linearly and
monotonically. Hence, the steady state voltage distribution for
conductors in the upper and lower disks is given by

Uupper(n) = U[(2l − x)/(2l)]

Ulower(n) = [x/(2l)] (4)

where n is the turn number and l is the total length of the con-
ductor in one disk. The equivalent energy Ed and the equivalent
capacitance Cd between two disks are given by

1 1 2
E d = C dU 2 = C dd U upper (n ) −U lower (n )
2 2
2 (5)
1 l x
= C ddU 2 ∫ 1 −  dx ,
2 0 l

Cd = (Cdd/3)l, (6) Figure 4. (a) Modified insulation parameters [12]. (b) Disk-to-
disk spacer.

34 IEEE Electrical Insulation Magazine

where

Cdd–paper = εtεo(Dπw/δt), (10)

Cspacer = εsεo(awsw/δd), (11)
Cdd–oil = εdεo[(Dπ − aws)w/δd], (12)

where D is the mean winding diameter, εd is the permittivity of

the oil, εs is the permittivity of spacer, a is the number of hori-
zontal spacers in a disk, ws is the width of the spacer, hs is the
thickness of single spacer, and w is the width of conductor in-
cluding its paper insulation. δd = ηhs is the disk-to-disk distance
after the winding has been pressed, where η is the number of
vertical spacers between a pair of disks. Finally the equivalent
series capacitance Cs for a pair of disks in continuous disk wind-
ing is given by

Cs = Cd + Ct, (13)

where Cd relates to the capacitances between the disks, and Ct

relates to the series capacitances between the conductors.

Series Capacitance of Interleaved Disk Winding

In an interleaved winding Cs is considerably greater than in
continuous disk winding, and therefore the voltage stress on the
windings is reduced. Various construction methods for interleav-
ing the disk windings are available. The series capacitance of the
simplest interleaved winding is given by
Cs = NidCtt[(N − 1)/4], (14)
where Nid is the number of interleaved disks, usually 2, and N is
the number of turns in each disk. The required manual welding
work increases rapidly with increasing Nid. The increased series
capacitance in interleaved windings results in a more linear volt-
age distribution across the winding structure. However, the turn-
to-turn potential difference increases during steady state opera-
tion, and the insulation must be able to withstand this potential
difference.
Series Capacitance of (IW)
Figure 5 shows the configuration of a disk winding incor-
porating electrostatic shields in each disk, i.e., in an IW. These
shields can be made from copper or aluminum. The shield turns
are inserted between the main conductors of the winding at pre-
determined places, and the shield or shield turns of each disk are
insulated from the conductors. In a pair of disks, the electrostatic
shield conductors of the upper disk are joined to the electrostatic
shield conductors of the lower disk at the outermost shield turn,
as shown in Figure 5(a). As an example, the shield between up-
per disk conductors 1 and 2 is connected to the shield between
lower disk conductors 19 and 20; both shields are insulated from
the main conductors. The equivalent RLC network, shown in
Figure 5(b), is a ladder network, which is commonly used in
transient studies.
The total energy of a pair of disks is given by
Figure 5. Intershield disk winding. (a) Cross-section schematic
view of intershield winding. (b) Equivalent RLC network for in-
tershield winding.
Etotal = Ed + Et, (15)
where Ed is the total disk-to-disk energy similar to (5), and Et is
the total turn-to-turn energy. Calculation of Ed using total disk-
to-disk energy has been presented in [8]. Et is given by
Et = (1/2)CsU 2
= SEtt–shield–upper + SEtt–shield–lower + 2(N − S − 1)Ett, (16)
where U denotes the voltage across a pair of disks, S is the num-
ber of shield turns in each disk, and Ett-shield-upper and Ett-shield-lower
are the energies stored in the upper and lower shields, respec-
tively. These energies are given by
Ett–shield–upper = (1/2)Ctt([U − Vx]2 + {U − Vx − [U/(2N)]}2),
(17)
Ett–shield–lower = (1/2)Ctt({U − Vx − 2N[U/(2N)]}2
+ {U − Vx − (2N − 1)[U/(2N)]}2), (18)
Ett = (1/2)Ctt[U/(2N)] , 2
(19)
where Vx is the potential of the connected shields between pairs
of disks.

September/October — Vol. 32, No. 5 35

 Table 1. Manufactured winding parameters (dimensions in mm)
Number of disks 10
Number of conductor in each disk
Height of conductor without insulation
Width of conductor without insulation
Thickness of conductor insulation at one side
Height of shield conductor without insulation
Width of shield conductor without insulation
Thickness of shield conductor insulation at one side
Average diameter 610
Substituting (17), (18), and (19) into (16), we obtain
Et = (S/2)Ctt([U − Vx]2 + {U − Vx − [U/(2N)]}2) side-down view of IW with six shield turns in each disk.
+ (S/2)Ctt({U − Vx − 2N[U/(2N)]}2
+ {U − Vx − (2N − 1)[U/(2N)]}2) (20) was then used to calculate the total series capacitance. The
+ (N − S − 1)Ctt[U/(2N)]2. (20)
The relation between stored energy and number of shields is
extracted from (20). Finding Vx and calculating the total series
capacitance of the winding is described below.
Test Objects and Measurements
Test Objects
A continuous disk winding, an interleaved disk winding, an
intershield disk winding with one shield turn in each disk, and
an intershield disk winding with six shield turns in each disk
were manufactured according to the specifications given in Ta-
ble 1. Figure 6 shows the manufactured windings. In IW, the
shield turns were made of copper and placed between the wind-
ing main conductors at predetermined locations, while the shield
or shield turns of each disk were insulated from the main con-
ductors of the winding. The interleaved disk winding produc-
tion time was four times that of the continuous winding with the
same technical specifications, while the production time for the
IW with six shield turns in each disk was approximately twice
that of the continuous winding.
Measurement of Vx and Calculation of IW Total
Series Capacitance
Vx [in (19) and (20)] is the voltage between any two “neigh-
boring shields.” In this context, the term “neighboring shields”
indicates that one shield is attached to one disk and the other
shield is attached to the disk underneath [Figure 5(a)].
To simulate the condition of the transformer winding under
lighting strike, and to find Vx, it is necessary to perform the stan-
dard reduced impulse voltage test [9] for manufactured wind-
ings. Consequently, the procedure described in the Initial Im-
pulse Voltage Distribution subsection below was carried out, and
36 IEEE Electrical Insulation Magazine
8
8.5
3.15
0.5
8.5
1.4
0.5
Figure 6. Manufactured windings. (a) From top to bottom: in-
terleaved, continuous, intershield winding (IW) with one shield
turn in each disk, IW with six shield turns in each disk. (b) Up-
measured Vx was substituted into (20) to determine the total se-
ries capacitance, and this capacitance was then compared with
the corresponding capacitance for windings without shields.
Performing the standard impulse voltage on manufactured wind-
ings, albeit at reduced magnitude, helped in examining the influ-
ence of shield turns on the initial impulse voltage distribution
along the winding.
Initial Impulse Voltage Distribution
A test setup was developed to carry out the necessary tests
on each winding. Standard reduced impulse voltages (1.2/50 μs)
were applied to the manufactured windings, and the initial im-
pulse voltage distributions were measured. In addition, Vx was
measured and then (20) was employed to determine total turn-
to-turn energy. Table 2 gives the initial impulse voltage distri-
bution percentages at the line end of the windings, and on the
even-numbered disks. Figure 7 also shows the voltage distribu-
tions along the winding body as functions of coil height fraction.
Figure 7(a) shows that the IW with six shield turns in each
disk yields the best approximation to a linear initial voltage
distribution. Measurements indicated that Vx would be approxi-
mately U/2, where U was defined in (16) as the voltage across a
pair of disks. U/2 was substituted in (20), and the total series ca-
pacitance for the IW was obtained using (21) and (22). It follows
from (22) that the total series capacitance of the IW depends on
the number of shield turns. The (N − 1)/N 2 terms in (22) are the
same as in (3), and the 2S(N − 1)/N terms relate to the effect of
the shields.
Et = (1/2)CtU 2 = (1/2)CttU 2([(N − 1)/N 2] + {[2S(N − 1)]/N})
(21)
Ct = (1/2)Ctt[(N − 1)/N 2] + (1/2)Ctt{[2S(N − 1)]/N}
= (1/2)Ctt([(N − 1)/N 2] + {[2S(N − 1)]/N}) (22)
 Table 2. Initial voltage distribution (%)1

IW with one IW with six

Measure Continuous shield turn Interleaved shield turns
point winding in each disk winding in each disk

Line end 100 100 100 100

Disk 2 43.6 58.2 74.2 76.2

Disk 4 21.2 34.4 52.7 55.1

Disk 6 10.1 19.3 34 35.8

Disk 8 4.4 9 17 17.7

Disk 10 0 0 0 0

Vx NA 77.1 NA 87.4
1
Test voltage (U) = 1 kV. IW = intershield winding. NA = not applicable.

When the number of shield turns is increased, the total series

capacitance increases, and consequently α decreases. The stress
on the upper disks of the IW with six shield turns in each disk is
less than the corresponding stress in the interleaved winding. Ta-
ble 1 shows that the initial impulse voltage distribution along the
upper disks of the IW with six shield turns in each disk is closer
to the linear distribution than the corresponding distributions in
the other windings, and so its α is smaller. It should be noted that
our previous simulation study on IW [13] also found a similar
improvement in transformer initial impulse voltage distribution
when shield turns were inserted in the winding.

FRA of the Windings

Test Setup
FRA is increasingly being used in diagnostics of power equip-
ment, especially for the identification of winding faults in trans-
formers [14]–[16]. FRA is considered as a high accuracy, fast,
economical, and nondestructive method of evaluating winding
characteristics [17]. A swept sinusoidal waveform within a pre-
determined frequency band, typically 20 Hz to 2 MHz [18], is
injected between the ends of the winding being tested, as shown
in Figure 8. The 50 Ω resistors represent the internal impedance
measuring cable. The frequency response magnitude k is given
by

k = 20 log10 (Vout/Vin) = 20 log10 [50/(Zwinding + 50)], (23)

where Zwinding is the winding impedance.

FRA measurements on each of the four winding types were
performed to compare the frequency response characteristics of
different winding designs.

Frequency Response Measurement Results

Figure 7. (a) Initial impulse voltage distribution along wind-
ing body. (b) and (c) Distributed voltage measurement points
(dashed lines). IW = intershield winding.
The rolled magnetic core of a transformer exhibits less mag-
netic reluctance than the air–metal core of individual windings,
and consequently, its self-inductance is higher, leading to higher
inductive reactance at low frequencies. Therefore, the first mini-
mum in an FRA trace, due to the main self-inductive reactance,

September/October — Vol. 32, No. 5 37

Figure 8. Frequency response analysis (FRA) test setup for end-
to-end winding measurement.

occurs at lower frequencies, mostly around 1 kHz. Hence FRA

measurement over the range of 20 Hz to 2 MHz would be tech-
nically sufficient to reveal the characteristics of a winding with
a magnetic core [19], [20]. However, in our study the windings
were located in air, with a metal cylinder placed inside each
winding to simulate the role of the transformer core [Figure
6(a)]. The air–metal core shifts the resonant points in the FRA
spectrum toward higher frequencies, because of its small self-
inductive reactance. For this reason the input sinusoidal signals
in our FRA measurements were swept over the frequency range
20 Hz to 12 MHz.
The results are shown in Figure 9. Obviously, different fre-
quency responses are to be expected since the windings are of
different designs. The FRA response of the IW with one shield
turn in each disk [Figure 9(b)] is similar to that of the continuous
disk winding response [Figure 9(a)], while the response of the
IW with six shield turns in each disk [Figure 9(d)] approaches
that of the interleaved winding [Figure 9(c)].
In relation to the influence of the initial impulse voltage dis-
tribution coefficient (α) on the frequency response spectrum, it
has been suggested [21] that the FRA spectra of windings with
lower series capacitance should exhibit more oscillations at mid-
frequencies (800 kHz–10 MHz in our study) than windings with Figure 9. Frequency response analysis spectra. (a) Continuous
high series capacitance. disk winding, (b) intershield winding with one shield turn in
Consider Figure 9(a). When the frequency increases from each disk, (c) interleaved disk winding, (d) intershield winding
20 Hz to 800 kHz, the inductive reactance of the transformer with six shield turns in each disk.
winding increases. Therefore, according to (23), the frequency
response magnitude (k) will fall smoothly. Oscillations appear in
the mid-frequency range 800 kHz to 10 MHz. The fluctuations, Practical Aspects of IW Fabrication
i.e., the number of maxima and minima, in the mid-frequency An IW is essentially a continuous disk winding with added
range in Figure 9(b) is less than that in Figure 9(a). The higher shield conductors between the main conductors of each disk.
series capacitance value for the interleaved winding [Figure When an IW is incorporated in a power transformer, applied im-
9(c)] results in an almost oscillation-free spectrum, specifically pulse voltages will “see” a uniform impedance, and thus the im-
over the frequency range of 800 kHz to 8 MHz. The same feature pedance mismatch problems associated with multiple-α wind-
appears in Figure 9(d), as would be expected since the IW with ings will not arise.
six shield turns in each disk also has a high series capacitance. One of the disadvantages of IW is that the shield conductors
We conclude that increasing the series capacitance in IW can lead to a heavier winding with larger outer diameter. This disad-
lead to less oscillation in its frequency response, i.e., a better vantage can be overcome, at least partially, by using thin shield
FRA spectrum. It follows that we can vary the number of maxi- conductors. In addition, manufacturers usually vary the number
ma and minima in the FRA response of an IW winding by vary- of shield turns along the winding, starting with a high number
ing the number of shield turns inserted between main conduc- at the critical line-end and reducing it over several disk pairs, to
tors. However, the same is not true of continuous or interleaved avoid reflections due to an abrupt change in the capacitance dis-
windings.

38 IEEE Electrical Insulation Magazine

tribution. Using aluminum shield conductors will yield lighter
windings, but may give rise to other technical problems.
Conclusion
To achieve a more uniform initial voltage distribution along a
transformer winding during fast impulsive surges such as light-
ning, one approach is to use the intershield winding arrange-
ment. We have shown the following:
• Inserting shield turns between transformer winding main
conductors increases the total series capacitance of wind-
ings, resulting in a lower α value and thus a more linear
initial impulse voltage distribution.
• The initial impulse voltage distribution along an intershield
disk winding is more uniform than that along a continuous
disk winding. It is also more uniform than the distribution
along an interleaved disk winding, provided the intershield
disk winding contains a sufficient number of shield turns.
• Different winding designs show different FRA responses.
It was shown that increasing the number of shield turns in
an IW disk results in fewer oscillations in the frequency re-
sponse.
Acknowledgments
We thank the reviewers for their valuable comments and
guidance.
References
[1] M. S. Naderi, M. Vakilian, T. Blackburn, B. T. Phung, H. Zhang, and
O. H. Nam, “A comparison between partial discharge propagation in
multiple-α and single-α transformer winding,” in Electr. Insul. Conf.
Electr. Manuf. Expo, Oct. 2005, pp. 108–111.
[2] O. Moreau, R. Dorr, and Q. Su, “Comparison between the high frequency
characteristics of transformer interleaved and ordinary disk windings,”
IEEE Conf. High Volt. Equip., Jan. 2000, pp. 2187–2192.
[3] R. C. Degeneff and J. C. Crouse, “Electrostatic shielding of nonsequential
disc windings in transformers,” US Patent 4243966, Jan. 6, 1981.
[4] R. A. Hinton, K. W. Doughty, and W. N. Kennedy, “Electrostatic shield-
ing of disc windings,” US Patent 4042900, Aug. 16, 1977.
[5] G. E. Sauer, “Impulse voltage distribution improving partial-turn electro-
static shields for disc windings,” US Patent 4017815, Apr. 12, 1977.
[6] R. E. Gearhart, “Tapering electrostatic shields for disc windings,” US Pat-
ent 3983522, Sep. 28, 1976.
[7] K. Ohta, K. Kurita, and S. Morita, “Shielded conductor for disk windings
for inductive devices,” US Patent 3820048, Jun. 25, 1974.
[8] R. M. Del Vecchio, B. Poulin, and R. Ahuja, “Calculation and measure-
ment of winding disk capacitances with wound-in-shields,” IEEE Trans.
Power Del., vol. 13, no. 2, pp. 503–509, Apr. 1998.
[9] A. Greenwood, Electrical Transients in Power System. New York, NY:
Wiley Interscience, 1991.
[10] A. De and N. Chatterjee, “Part winding resonance: Demerit of interleaved
high voltage transformer winding,” in IEE Proc.-Electro. Power Appl.,
vol. 147, no. 3, May 2000, pp. 167–174.
[11] S. Pramanik and L. Satish, “Localisation of discrete change in a trans-
former winding: A network-function-loci approach,” IET Electr. Power
Appl., vol. 5, no. 6, pp. 540–548, 2011.
[12] K. Karsai and D. Kerenyi, Large Power Transformer. New York, NY:
Elsevier, 1987, pp. 187–214.
[13] M. Bagheri, M. Vakilian, A. Hekmati, and R. Heidarzadeh, “Influence of
electrostatic shielding of disc winding on increasing the series capaci-
tance in transformer,” in IEEE PowerTech ’07, 2007, pp.1780–1784.
[14] N. Hasheminia, A. Abu-Siada, and S. Islam, “Improved power trans-
former winding fault detection using FRA diagnostics—Part 2: Radial
September/October — Vol. 32, No. 5 39
deformation simulation,” IEEE Trans. Dielectr. Electr. Insul., vol. 22, no.
1, pp. 564–570, 2015.
[15] K. P. Badgujar, M. Maoyafikuddin, and S. V. Kulkarni, “Alternative
statistical techniques for aiding SFRA diagnostics in transformers,” IET
Gener., Transmis. Distrib., vol. 6, no. 3, 2012, pp. 189–198.
[16] A. Reykherdt and V. G. Davydov, “Effects of test cable ground extensions
on repeatability of frequency response analysis measurement on power
transformer,” IEEE Electr. Insul. Mag., vol. 28, no. 3, pp. 26–31, 2012.
[17] M. Bagheri, M. S. Naderi, and T. Blackburn, “Frequency response analy-
sis and short circuit impedance measurement in detection of mechanical
defects within large power transformer,” IEEE Electr. Insul. Mag., vol.
29, no. 3, pp. 33–40, 2013.
[18] D. A. K. Pham, T. M. T. Pham, H. Borsi, and E. Gockenbach, “A new
diagnostic method to support standard frequency response analysis
assessment for diagnostics of transformer winding mechanical failure,”
IEEE Electr. Insul. Mag., vol. 30, no. 2, pp. 34–41, 2014.
[19] IEEE Guide for the Application and Interpretation of Frequency Re-
sponse Analysis for Oil-Immersed Transformers, IEEE Std. C57.149,
2012.
[20] Measurement of Frequency Response, IEC Standard 60076-18, Edition
1.0, 2012-07.
[21] Z. Wang, J. Li, and D. M. Sofian, “Interpretation of transformer FRA
responses—Part I: Influence of winding structure,” IEEE Trans. Power
Del., vol. 24, no. 2, pp. 703–710, 2009.
Mehdi Bagheri (S ’07, M ’12) received
his MSc degree in power engineering from
Sharif University of Technology, Tehran,
Iran, in 2007. He joined the Iran Trans-
former Research Institute (ITRI) in Tehran
as a research engineer, and was head of the
Test and Diagnostic Department between
2008 and 2010. In 2014 he gained his PhD
from the University of New South Wales
(UNSW), Sydney, Australia. From 2015 to 2016 he served as a
postdoctoral research fellow in the Electrical Engineering De-
partment, National University of Singapore, working closely
with Rolls-Royce Pte. Ltd. on condition monitoring and predic-
tive maintenance of marine transformers and filters. He is cur-
rently an assistant professor in the School of Electrical and Elec-
tronic Engineering, University of Nazarbayev, Astana,
Kazakhstan.  He is a member of the IEEE Dielectrics and Elec-
trical Insulation Society (DEIS). His research interests include
field and marine applications of high-voltage engineering, con-
dition monitoring and diagnosis of power transformers and elec-
trical rotating machines, transients in power systems, and power
quality.
B. T. Phung (M ’87, SM ’12) gained a PhD
in electrical engineering from the Univer-
sity of New South Wales (UNSW), Sydney,
Australia, in 1998. He is currently an as-
sociate professor in the School of Electri-
cal Engineering at UNSW. He has over 30
years of practical research/development
experience in partial discharge measure-
ment and analysis, and in on-line condition
monitoring of high-voltage equipment. Much of his work has
involved collaborative projects between UNSW and Australian
power utilities. His research interests include electrical insula-
tion (dielectric materials and diagnostic methods), high-voltage
engineering (generation, testing, and measurement techniques),
electromagnetic transients in power systems, and power system
equipment (design and condition monitoring methods).
Mohammad Salay Naderi (M ’10, SM
’11) is currently the managing director of
Smart Energy Solutions Group Pty. Ltd. in
New South Wales, Australia. He received
his BSc in electric power engineering
(1996) and MSc (1998) in high voltage en-
gineering from Sharif University of Tech-
nology in Tehran, Iran. In 2006 he com-
pleted his PhD at the same university. He
worked with Iran Power Generation, Transmission and Distribu-
tion Management Company (Tavanir Holding Co.) as deputy
director general of the Technical Supervision Bureau of Trans-
mission System from 2005 to 2010. In 2010 he joined the School
40 IEEE Electrical Insulation Magazine
of Electrical Engineering and Telecommunications at UNSW as
lecturer in the Energy Systems Group. In 2013 he joined ABB,
and then High Voltage Services, as senior HV testing and com-
missioning engineer. He is involved in advance testing and com-
missioning of power transformers and switchgear, and provides
a consultancy service to multi-national companies in areas in-
cluding power products manufacturing, strategic relationship
management between industry and academia to foster new busi-
ness, and science and commercial opportunities. His technical
interests include high voltage engineering, condition monitoring
of power equipment, and sustainable/smart energy systems anal-
ysis and development.