Electric Power Systems Research 145 (2017) 55–62

Contents lists available at ScienceDirect

Electric Power Systems Research

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

Study on locating transformer internal faults using sweep frequency

response analysis
Mahdi Khanali a,∗ , Amir Hayati-Soloot b , Hans Kristian Høidalen b , Shesha Jayaram a
a
High Voltage Engineering Lab (HVEL) Department of Electrical and Computer Engineering, University of Waterloo, Waterloo, ON N2L 3G1, Canada
b
Electrical Power Engineering Department, Norwegian University of Science and Technology, Trondheim N-7491, Norway

a r t i c l e i n f o a b s t r a c t

Article history: As an extensive network of resistances, capacitances, and inductances, a transformer has inherent char-
Received 22 February 2016 acteristic parameters that are functions of frequency. The form of these functions is predicated on the
Received in revised form geometric design of the transformer and materials that comprise it. Any change in the structure of a
17 September 2016
transformer will be reflected in its frequency response characteristics. Of the possible changes that can
Accepted 21 November 2016
take place in a transformer structure, an internal short circuit is one of the far-reaching incidents that
has been recently reported for many wind-farm transformers. Detecting the location of an internal short
Keywords:
circuit that has occurred in a transformer winding is therefore beneficial in the repair process and also
Sweep frequency response analysis
Internal short circuit
in improving future designs. In an effort to identify trends with inter-turn fault locations and frequency
Inter-turn fault location responses, this research investigates the effect of the location of deliberately initiated internal faults on
Transfer function parameters such as transfer voltages and input impedances by means of sweep frequency response anal-
Transformer fault diagnosis ysis (SFRA). The analysis of three different model transformers with different core and winding designs
shows several trends in frequency response patterns, depending on the location of the internal short
circuits. The paper discusses such trends as a potential use of SFRA in locating inter-turn winding failures
that may result in noticeable short circuits.
© 2016 Published by Elsevier B.V.

1. Introduction using frequency response analysis (FRA) as the most prominent

In wind-farms, connecting and disconnecting a transformer
from the grid using fast switching breakers is essential due to vari-
ations in wind speed. Transformers operating in wind-farms are
exposed to fast transients at a multitude of times due to frequent
switching operations, thereby stressing the winding insulation to
high frequency stresses. The severity of such stresses are signif-
icantly high in wind-farm transformers due to the interaction
between capacitance and inductance of both connecting cables and
transformers involving fast operating vacuum circuit breakers and
power electronic converters [1–3]. On the other hand, detection of
transformer insulation faults by using frequency response analysis
(FRA) has been introduced and developed by previous researchers
as a method that adopts changes in frequency responses to diagnose
physical and/or chemical changes in the transformer. For this pur-
pose, both online and offline procedures [4] have been developed
∗ Corresponding author.
E-mail addresses: mhd.khanali@gmail.com (M. Khanali), amir.hayati@ieee.org
(A. Hayati-Soloot).
measurement methods.
Secue and Mombello [5] reviewed literature related to the
application of sweeping frequency response analysis (SFRA) for
diagnosis of transformer failures. This review explains a number of
aspects that must be considered with respect to implementation of
SFRA, such as, frequency range and connection of non-tested termi-
nals. The authors have also discussed the sensitivity of SFRA as well
as the associated sources of uncertainty and inaccuracies that influ-
ence measured data. It has been concluded that SFRA can be a useful
technique for detecting winding movement and loss of clamping
pressure. In Ref. [6], a combination of SFRA and an impedance mea-
surement method is proposed as a means of enhancing diagnosis
of mechanical failure in windings.
In addition to the work related to mechanical failures mentioned
above, other studies have focused on detection of inter-turn short
circuits in general. In Ref. [7], changes in maximum input admit-
tance values are described with respect to short circuits in electrical
machines. Wilk and Adamczyk [8] found that variations in ampli-
tude at resonance frequencies represent a U-shaped trend as an
inter-turn fault develops; this occurs because the amplitude of the
transfer function is closely related to the amount of power loss due
to short circuit resistance. Since this trend is related to energy, as

http://dx.doi.org/10.1016/j.epsr.2016.11.016
0378-7796/© 2016 Published by Elsevier B.V.
56 M. Khanali et al. / Electric Power Systems Research 145 (2017) 55–62
Table 1
Transformers specification.
Geometrical dimensions Transformer I
Rated output 500 kVA, three phase
Voltage 240 V/11 kV
Frequency 50 Hz
Oil immersed/not immersed Not immersed
Number of layers for HV winding 12
Number of turns for HV winding 1334
the fault develops, consumed energy increases in the early stages,
while the opposite effect occurs during later stages of fault develop-
ment. Resonance occurred at higher frequencies as the inter-turn
fault developed.
Utilising FRA measurements, authors of this paper (Soloot et al.)
have previously investigated the effects of winding design on res-
onance overvoltages in transformer windings [9]. The FRA method
was adopted to measure voltage-drops between adjacent layers
and discs of transformer windings through a wide range of fre-
quency. The results showed that depending on the design of the
windings, transformers show different patterns of internal stresses.
It was concluded that disc winding has a higher internal stress
at resonance frequencies compared to that of pancake and layer
windings; layer structure showed the highest resonance voltage at
winding terminals. In another work by other authors of this paper
(Khanali and Jayaram) [10], a comparison is made between fre-
quency responses of electrostatically shielded and non-shielded
layer type transformers with the focus being on identification
of differences in their resonant overvoltage behaviours. Further,
it has been shown that resonances that occur with many fre-
quency components can lead to high stresses and eventually cause
layer-to-layer short circuits. Such resonance initiated high stresses
particularly in transformers connected to wind-farms are also
reported by Banda and Van Coller [11].
Through the literature, it is evident there is less focus on detec-
tion of the position of internal short circuits using FRA method.
Knowing the location of an internal fault can benefit operators and
manufacturers by: reducing cost and time required for repair, and
also providing statistical information about the areas of the wind-
ings in which internal faults occur more often. The latter would be
useful for more effective insulation reinforcement in future designs.
Greater reliability and efficiently reinforced insulation is especially
valuable for transformers located in offshore wind-farms because
of higher maintenance and repair costs. In addition, due to the
exposure of wind turbine transformers to high dV/dt voltages from
frequent switching, inter-turn and inter-layer insulation of such
transformers is under higher stress compared to their power-grid
counterparts, as an internal short circuit is more likely to happen.
With an experimental approach on model transformers, the aim
of the research presented is to identify the relationship between
locations of an internal fault and patterns of transformer fre-
quency responses. To estimate the location of short circuits, the
FRAs of faulty windings, with short circuits positioned at different
locations, are compared with one another and also with healthy
winding as a reference. Comparison is based on use of statistical
parameters along with visual observations of trends in frequency
spectra. This research also compares sensitivities of different trans-
fer functions with respect to location of the fault.
2. Experimental
Three custom designed model transformers (Table 1) have been
used in this study with special features to facilitate the creation of
artificial short circuits at different locations of the windings. This
feature is made possible by inserting several connection leads along
the HV winding at different layers. Detailed designs of all three
Transformer II Transformer III
9 kVA, single phase 1 kVA, single phase
345 V/19.8 kV 115 V/6.6 kV
60 Hz 60 Hz
Oil filled Oil filled
37 41
7480 10,496
Table 2
Fault numbering based on the position of short circuits according to Fig. 1.
Fault number Location of short circuit (between
connection leads)
1 2 and 6
2 4 and 8
3 6 and 10
4 8 and 12
5 10 and 14
6 12 and 16
7 14 and 18
8 16 and 20
9 18 and 23
transformers are described in Refs. [9,10,12]. The proposed scheme
of using SFRA in identifying trends with fault locations in different
layers are carried out using the model transformer I in detail while
the other two model transformers II and III, are used to verify the
applicability of the SFRA approach to different designs.
Unlike conventional FRA tests on transformers wherein the
responses are used to detect faults such as winding deformations
or core dislocations, in this investigation, frequency responses of
different parameters are measured against a known defect. For this
purpose, nine short circuits were created at different locations by
short-circuiting a certain number of HV winding turns. Details of
the windings of model transformer I that is used in the analysis
is given in Fig. 1. It is also presumed that access to the terminals
of each individual winding is available for measurement. It should
be noted here that although model transformer I is a three-phase
design, only one phase is considered in this study, while the unused
phase windings are short circuited. As the effect of the core for
magnetic coupling between windings of different legs (phases) is
negligible for frequencies above 10 kHz (due to the huge drop in
core permeability) [13]. Windings of different legs have insignifi-
cant mutual effects on each other.
Transformer I is designed in a way that the number of turns
between each two consecutive connections is 56 except for the taps
(21–26) which have 32 turns between one another. This means that,
in terms of number of short circuited turns, all implemented faults
have the same number of turns (112) except the fault in the out-
ermost layer, which has eight more turns. Fault numbering along
with short circuit locations are given in Table 2.
Both transfer functions; HV/LV and LV/HV are measured, respec-
tively, from HV and LV terminals while the other side is excited. In
addition, input HV impedances with LV terminals shorted are also
utilised to correlate measured impedances with defect locations.
®
An Agilent network analyser E5061B and high-frequency oscillo-
scope probes P2220 are used for voltage transfer measurements,
®
as well as special Ionphysics current sensors CM-100-6L-IR50 for
wide-band impedance measurements (Fig. 2).
3. Fault analysis
Variations that occur in transformer frequency responses due to
changes in the location of fault are investigated. The network anal-
yser generates sinusoidal waveform with variable frequencies at
 M. Khanali et al. / Electric Power Systems Research 145 (2017) 55–62
Fig. 1. Details for model transformer I. (a) Transformer winding with connection leads from different locations, (b) winding schematic and positions of connecting leads.
Fig. 2. Test setup for FRA measurement. (a) Transfer function, (b) input HV
impedance (1 – transformer with grounded tank and core, 2 – network analyser,
3 – high-frequency voltage probe, 4 – high-frequency current sensor).
terminal S (Fig. 2). Connections of R and T channels are determined
based on the parameter for which frequency response is mea-
sured. For example, to measure LV/HV voltage transfer function,
R and T channels are connected to HV and LV terminals respec-
tively (Fig. 2a). On the other hand, as it is demonstrated in Fig. 2b,
R and T channels are respectively connected to the current sensor
and HV terminal to measure input impedance.
To analyse the FRA measurement results, both qualitative obser-
vations on trends, and quantitative analysis are used. Trends in
57
frequency responses based on patterns in the shifts of resonance
frequencies and changes in resonance magnitudes have been inves-
tigated in detail. On the other hand, quantitative analysis, which
derives numeric values for comparing FRA measurements, is based
on the “difference factor” which is defined below.
A measure that can be adopted to compare faulty windings with
the reference healthy winding is a criterion derived from absolute
sum of logarithmic errors; hereafter referred to as the difference
factor. The higher the difference factor is, the larger are the differ-
ences between responses; hence it is easier to differentiate between
graphs.

fmax
|20log10 R0 − 20log10 Rsc k |
ıASLE (Rsc k , R0 ) = (1)
N
f =fmin
where N is the number of readings, ıASLE is a quantitative indi-
cation of the difference between Rsc k , responses of transformer
winding with short circuit at location k, and R0 , the response of a
healthy transformer. This difference factor can also be generalised
to include other types of faults. Compared to other mathematical
criteria, the difference factor calculated using the above formula
was found to be the most sensitive criterion to diagnose even badly
scaled responses irrespective of fault type [14].
4. Results
Both transfer voltage and input impedance were measured
using procedures described in Sections 2 and 3. The results are
presented using 3D and 2D plots. In the 3D plots, X, Y, and Z axes
represent frequency, location (fault number) of the short circuit,
and magnitude of the parameters, respectively. For clarity, with 3D
plots, the magnitude of parameters are indicated by coloured maps
wherever it was helpful. Together, the 2D and 3D plots facilitate to
identify effects of fault location on transformer frequency response
patterns. Sub-sections 4.1, 4.2 and 4.3 represent the results for
model transformer I while 4.4 shows complementary results for
model transformers II and III.
58 M. Khanali et al. / Electric Power Systems Research 145 (2017) 55–62
Fig. 3. HV over LV transfer function. (a) 3D plot, (b) 2D plot (the long arrow indicates
the trend of the curves for short circuits #1–#9. The solid curve corresponds to
healthy winding).
4.1. HV/LV transfer voltage
Fig. 3 shows HV voltage over LV voltage transfer function (volt-
age transfer from LV side to HV side) for faults at different locations
in the winding. The 2D graphs in Figs. 3 and 5 present normalised
values of the frequency response in a logarithmic scale. Values are
normalised with respect to the transformer’s original turns ratio
(HV/LV = 83.375 @50 Hz) to yield a comparison between response
values and transformer rated voltage ratio. It can be seen that, com-
pared to a healthy transformer response; represented using a solid
curve, when the location of the fault is in the inner layers, first
resonance occurred at lower frequencies. The most evident charac-
teristic of the graph is that for fault #1, transfer voltage tends to be
enormously high (58 dB more than its normal value) at 11.8 kHz. For
faults in subsequent layers (#2, #3, and #4), magnitudes of transfer
Fig. 4. Difference factor of HV/LV transfer voltage.
voltage at their first resonance frequencies are respectively 29 dB,
15 dB, and 6 dB more than their normal value, revealing a decreas-
ing trend. Although HV/LV transfer voltages for the outer layers do
not indicate a specific trend, some standalone critical frequencies,
such as 180 kHz with a resonance magnitude of 51 p.u. for fault #7,
was noted. Such standalone peaks can be further employed as a
means of matching fault more precisely.
Fig. 4 shows that the difference factor, with respect to transfer
voltage, decreases for faults in the outer layers. This means that
frequency responses of transformers whose faults are located in
the outer layers are closer to the frequency response of a healthy
transformer. Observing the trends shown in Fig. 3 and the difference
factor in Fig. 4, it can be determined that HV/LV magnitude is quite
sensitive to faults that occur in the inner layers, although, it is less
sensitive to those faults that occur in the outer layers.
4.2. LV/HV transfer voltage
Fig. 5 shows the LV voltage over HV voltage transfer function
(voltage transfer from HV side to LV side). It can be observed that at
10 kHz, transfer voltage magnitude drastically decreases by moving
the fault from inner layers to middle ones, this trend continues by
moving the fault to outer layers but with a lower slope. The range of
these changes are from 70.8 dB for the innermost fault to 9.5 dB for
the outermost fault. On the other hand, minimum values of trans-
fer voltages (the first anti-resonance values) and their frequencies
increase by moving the short circuit location from outer layers to
inner ones. Finally, in both Figs. 3 and 5, frequency response curves
show interpretable differences up to 1 MHz while beyond that fre-
quency; responses are almost the same for all fault locations.
Similar to what is shown in Fig. 4, Fig. 6 shows that the difference
factor decreases by moving the short circuit location to outer layers.
This implies that detecting inter-turn short circuits in inner layers
can easily be identified by comparing FRA responses. Furthermore,
the overall higher value of difference factor in Fig. 6 compared to
that in Fig. 4 reveals that inter-turn short circuits can be detected
more promisingly using LV/HV transfer function compared to a
HV/LV one.
4.3. Input impedance from HV side
Fig. 7 shows the frequency spectra of input impedance from HV
side for internal faults at different locations in the winding. Unlike
what was observed in the transfer voltage graphs (Figs. 3–6), a
symmetrical pattern exists between input impedance responses
 M. Khanali et al. / Electric Power Systems Research 145 (2017) 55–62
Fig. 5. LV over HV transfer function. (a) 3D plot, (b) 2D plot (the long arrow indicates
the trend of the curves for short circuits #1–#9. The solid curve corresponds to
healthy winding).
of short circuits associated with faults in inner and outer layers.
The trends observed with input impedance also match with the
variation in the difference factor calculated (Fig. 8). As a result, a
quantitative index can be generated for quick evaluation on fault
location evaluation from SFRA data.
Fig. 6. Difference factor of LV/HV transfer voltage.
59
4.4. Evaluation of SFRA approach
To evaluate the applicability of the proposed method on
different transformers, the SFRA analyses conducted on model
transformer I are performed on two other oil-filled 1 kVA and 9 kVA,
model transformers II and III respectively (described in Table 1).
Results obtained from transfer function and impedance measure-
ments show that similar trends exist in the FRA of transformers II
and III when the fault location moves from inner layers to the outer
ones. Input impedance from the HV side of transformers II and III
are presented here as an example. Fig. 9 shows similar symmetric
patterns in Fig. 7 for transformer I, can also be seen for transformers
II and III. Existence of systematic trends in frequency response of
three model transformers despite huge differences in their designs
(core designs, insulation systems, power ratings and voltage levels),
verifies the applicability of the proposed method as a diagnostic
approach for estimating internal fault locations.
5. Discussion
Today, FRA method is widely used to detect a variety of faults in
transformers using commercially available instruments that have
been developed for field tests. The high sensitivity of this method,
allows one to detect even gentle displacements of windings on
transformer core [15]. Compared to defects such as winding dis-
placement and deformation, internal short circuits cause significant
changes in the frequency response of power transformers. For
instance, the difference between frequency responses of a healthy
winding with 20% deformed winding is limited to only 5% frequency
difference in resonance frequency [16]. In contrast, the difference
between the HV impedance frequency response of the winding
with a fault in the first layer and the winding with a fault in the
second layer is near 90% in magnitude and 10% difference in res-
onance frequency. Changes in frequency response that occur due
to internal short circuits dominate changes that occur due to other
defects such as displacement and deformation.
The FRA test results presented above reveal that when a short
circuit is located in the innermost layer of the winding, very high
voltage amplifications can be expected for certain frequencies. For
the case of model transformer I, the highest voltage amplification
happened with a fault between layers #1 and #2, where the HV/LV
transfer voltage tends to be enormously high at 11.8 kHz. Such high
magnitudes with resonant cases have also been observed in Ref.
[17]; where high magnitudes were recorded for transfer functions
60 M. Khanali et al. / Electric Power Systems Research 145 (2017) 55–62

Fig. 7. Magnitude of input impedance from HV side (LV terminal short circuit). (a) 3D view, (b) 2D plot, (c) coloured map representation.

at certain frequencies. This high amplitude can be related to the
formation of an equipotential cylinder composed of shorted turns
of HV winding close to the LV winding. High capacitance between
this virtual cylinder and the LV winding results in high resonance
for transferred voltage. From the insulation point of view, such phe-
nomena can result in very high stresses inside the transformer even
when a very low content of the associated frequency exists in the
feeding voltage. With respect to practicality, when such high gains
in transfer voltages are observed (Figs. 3 and 5), there may be a
concern that saturation of the core may affect measurements. That
said, since frequencies are much higher than power frequency and
excitation voltage is too low, flux density is very low and is not
significant to cause any core saturation.
Further investigation of Fig. 7 (input impedance from HV side,
when LV terminal is shorted) reveals that connections of non-tested
windings are influential on the trend of frequency response changes
as a function of fault location. When impedance measurements are
performed with LV side short-circuited, the effect of magnetisa-
 M. Khanali et al. / Electric Power Systems Research 145 (2017) 55–62
Fig. 8. Difference factor of input impedance from HV side (LV terminal open circuit).
Fig. 9. Magnitude of input impedance from HV side (LV terminal short circuit). (a)
Coloured map representation for 1 kVA transformer, (b) coloured map representa-
tion for 9 kVA transformer.
tion inductance through the core and LV winding space can be
neglected. This means the flux path generated by HV winding is
mainly restricted to the space out of LV winding and within the HV
winding itself. As a result, and due to the symmetry of the HV wind-
ing, the winding with a fault just next to one end of the winding
(i.e. outermost layer fault) shows the same impedance frequency
response as the winding with a fault just next to the other end of
the winding (i.e. innermost layer fault). The same trend exists for
other faults equally distant from the terminals.
Comparing trends in transfer ratios and input impedances, it
was observed that two parameters do not follow the same pattern
with respect to fault location. This observation is in compliance
with what can be concluded from the literature stating that transfer
voltages have no direct relationship with input impedance mea-
61
surement [18]. Therefore, it would be beneficial to perform both
tests for transformer diagnosis as the observations complement
each other.
6. Conclusion
As an approach to indication of the position of short circuit
in transformer winding, frequency responses of a transformer for
faults at different locations were investigated on a model trans-
former with layered windings. The repeatability of the results was
verified by conducting similar tests on two other oil-filled trans-
formers with same winding configuration (layer winding) but with
different number of layers and turns, and with different voltage and
power ratings.
Comparisons between frequency responses revealed that
changes in the location of internal faults result in identifiable trends
as functions of the distance of the fault from the core. For exam-
ple, amplitude of both HV/LV and LV/HV transfer voltages showed
a decreasing trend at ∼10 kHz when locations are changed from
close-to-core to close-to-HV terminal positions. On the other hand,
amplitude of the HV impedance exhibited a symmetrical pattern
with respect to the location of short circuits with similar patterns
for close-to-core and close-to-HV terminal positions. Using FRA
method, short circuits were classified into two categories: close-to-
core faults and close-to-terminal faults. Although trends observed
in the responses of input impedance and transfer voltages are inde-
pendent, the results can be used complementarily in the detection
of fault location. Further, it was demonstrated that quantitative
analysis of statistical parameters, such as the difference factor,
could also be used in addition to trends observed in FRA patterns.
Acknowledgments
The authors greatly appreciate the source of funding: the
Norwegian Research Center for Offshore Wind Technology
(NOWITECH) and the Natural Sciences and Engineering Research
Council of Canada (NSERC). They are also thankful of Møre Trafo
Co. and Moloney Electric Inc. for their assistance in manufacturing
the model transformers for the study.
References
[1] B. Gustavsen, Study of transformer resonant over-voltages caused by
cable-transformer high-frequency interaction, IEEE Trans. Power Deliv. 25
(April) (2010) 770–779.
62 M. Khanali et al. / Electric Power Systems Research 145 (2017) 55–62
[2] J. Aubin, S.A.C. Hall, Colloquium on transformer reliability and electrical
transients, Report on SC A2 Colloquium – Moscow 2005, April 2006, 18–24.
[3] M. Nagel, T. Leibfried, Investigation on the high frequency, high voltage
insulation properties of mineral transformer-oil, in: 2006 IEEE Conference on
Electrical Insulation and Dielectric Phenomena, 15–18 October, 2006, pp.
226–228.
[4] V. Behjat, A. Vahedi, A. Setayeshmehr, H. Borsi, E. Gockenbach, Diagnosing
shorted turns on the windings of power transformers based upon online FRA
using capacitive and inductive couplings, IEEE Trans. Power Deliv. 26
(October) (2011) 2123–2133.
[5] J.R. Secue, E. Mombello, Sweep frequency response analysis (SFRA) for the
assessment of winding displacements and deformation in power
transformers, Electr. Power Syst. Res. 78 (2008) 1119–1128.
[6] D.A.K. Pham, T.M.T. Pham, H. Borsi, E. Gockenbach, A new method for
purposes of failure diagnostics and FRA interpretation applicable to power
transformers, IEEE Trans. Dielectr. Electr. Insul. 20 (2013) 2026–2034.
[7] M. Florkowski, J. Furgał, Detection of winding faults in electrical machines
using the frequency response analysis method, Meas. Sci. Technol. 15 (2004)
2067–2074.
[8] A. Wilk, D. Adamczyk, Investigations on sensitivity of FRA method in
diagnosis of interturn faults in transformer winding, Proc. of IEEE
International Symposium on Industrial Electronics (ISIE) (2011) 631–636.
[9] A.H. Soloot, H.K. Hoidalen, B. Gustavsen, Influence of the winding design of
wind turbine transformers for resonant overvoltage vulnerability, IEEE Trans.
Dielectr. Electr. Insul. 22 (2015) 1250–1257.
[10] M. Khanali, S.H. Jayaram, Effectiveness of electrostatic shielding in
suppressing the impact of fast transients on transformer insulation, in: 2015
IEEE Conference on Electrical Insulation and Dielectric Phenomena (CEIDP),
18–21 October, 2015, pp. 652–655.
[11] C.A. Banda, J.M. Van Coller, Investigation into Resonant Overvoltages in Wind
Turbine Transformers due to Switching Surges, 2015 International Conference
on Power Systems Transients (IPST), 15–18 June 2015, Paper ID 31.
[12] M. Khanali, S. Jayaram, J. Cheng, Effects of voltages with high-frequency
contents on the transformer insulation properties, in: 2013 IEEE Electrical
Insulation Conference (EIC), 2–5 June, 2013, pp. 235–238.
[13] J. Merrikhi, J.S. Moghani, E. Fallah, Laminated iron core inductor model with
flux skin effect, in: 2006 2nd International Conference on Power Electronics
Systems and Applications, Hong Kong, 2006, pp. 77–78.
[14] J.W. Kim, B.K. Park, S.C. Jeong, S.W. Kim, P.G. Park, Fault diagnosis of a power
transformer using an improved frequency-response analysis, IEEE Trans.
Power Deliv. 20 (2005) 169–178.
[15] M. Bagheri, M.S. Naderi, T. Blackburn, T. Phung, Frequency response analysis
and short-circuit impedance measurement in detection of winding
deformation within power transformers, IEEE Electr. Insul. Mag. 29 (3) (2013)
33–40.
[16] A.P. Purnomoadi, D. Fransisco, Modeling and diagnostic transformer condition
using sweep frequency response analysis, in: ICPADM 2009. IEEE 9th
International Conference on the Properties and Applications of Dielectric
Materials, 19–23 July, 2009, pp. 1059–1063.
[17] S.M. Islam, G. Ledwich, Locating transformer faults through sensitivity
analysis of high frequency modeling using transfer function approach, 16–19
June, in: Conference Record of the 1996 IEEE International Symposium on
Electrical Insulation, vol. 1, 1996, pp. 38–41.
[18] J.R. Secue, E. Mombello, Sweep frequency response analysis (SFRA) for the
assessment of winding displacements and deformation in power
transformers, Electr. Power Syst. Res. 78 (2008) 1119–1128.