0% found this document useful (0 votes)

32 views

Comprehensive Study On Transformer Fault Detection

This document provides an overview of a comprehensive study on transformer fault detection via frequency response analysis (FRA). It discusses how FRA works by comparing the transformer's frequency response in healthy and faulty conditions to identify faults. The document outlines various transformer modeling approaches used in FRA and methods for analyzing FRA results. It also introduces advanced complementary methods to address FRA's limitations and suggests areas for further research to improve transformer fault detection accuracy and efficiency using FRA.

Uploaded by

Available Formats

Download as PDF, TXT or read online on Scribd

0% found this document useful (0 votes)

32 views

Comprehensive Study On Transformer Fault Detection

Uploaded by

Available Formats

Download as PDF, TXT or read online on Scribd

You are on page 1/ 31

This article has been accepted for publication in IEEE Access.

This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2023.3300378

Date of publication xxxx 00, 0000, date of current version xxxx 00, 0000.
Digital Object Identifier 10.1109/ACCESS.2022.Doi Number

Comprehensive Study on Transformer Fault

Detection via Frequency Response Analysis
Seyed Ebrahim Hosseini Kakolaki1,*, Vahid Hakimian2,*, Javad Sadeh1, and
Elyas Rakhshani3, Senior Member, IEEE
1
Faculty of Engineering, Ferdowsi University of Mashhad, Mashhad, Iran
2
Faculty of Electrical and Computer Engineering, Tarbiat Modares University, Tehran, Iran
3
Department of Electrical Sustainable Energy, Delft University of Technology, 2628 CD Delft, The Netherlands

Corresponding authors: Elyas Rakhshani (e-mail: [email protected]) and Javad Sadeh (e-mail: [email protected])
*
The first two authors contributed equally to this paper.

ABSTRACT The sudden outage of a transformer due to a fault can cause irreparable damage to the electricity
industry. Hence, by conducting momentarily inspections of the transformer's condition, faults can be promptly
detected, and the transformer can be disconnected from the power grid to prevent subsequent failures in this
equipment. Detecting faults at an early stage can also result in reduced repair costs. One recent promising
technique for fault detection is Frequency Response Analysis (FRA), which compares the transformer's
response in healthy and faulty conditions for understanding the occurrence of transformer faults. This paper
presents a comprehensive and accurate modeling approach for the behavior of the transformer at different
frequencies, followed by an exposition of the requirements for implementing this method in order to find the
fault type, severity, and location. Additionally, various methods for analyzing the results of frequency
response are introduced and discussed. In this regard, attempts have been made to introduce advanced
complementary methods to address the weaknesses and limitations of the frequency response method. Finally,
the concepts are summarized, and suggestions for further research with applications in this field are presented
and compared.

INDEX TERMS Transformer fault detection, Frequency response analysis, Hybrid model, Ladder model,
Configuration-transfer function pair, Artificial intelligence, Online frequency response analysis.

I. INTRODUCTION vibration of a transformer. However, these are the chemical

A transformer consists of several components, including and electrical methods that provide more useful information
windings and cores (known as electrically and magnetically about the condition of the transformer. Various physical and
active parts, respectively), bushings, solid and liquid chemical methods have been studied [2], including the
insulators (mainly insulator between high voltage windings method of assessing the viscosity of insulation, polarization
and transformer body, insulator between high voltage and indicator of insulation, as well as the dissolved gas analysis
low voltage windings, insulator between low voltage (DGA) method and the optical spectroscopy technique,
windings and transformer core) and possibly tap changer. which are discussed in the references [3]–[5]. For the sake of
The occurrence of untimely and unexpected failure in any of brevity, mentioning other methods has been avoided.
the mentioned components can occur due to various stresses Furthermore, there are various electrical fault detection
such as electromagnetic stresses, dielectric stresses, thermal methods that relay on the signals received from the
stresses and chemical factors [1]. transformer, including methods based on dielectric
In general, fault detection methods in power transformers dissipation factor (DDF), leakage reactance, polarization,
can be divided into three categories: physical, chemical, and and insulation partial discharge (PD) detection. These
electrical methods. Physical methods are based on the methods can be classified into basic and advanced
measurement of quantities such as the temperature and

VOLUME XX, 2017 1

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
This article has been accepted for publication in IEEE Access. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2023.3300378

categories. One of the advanced electrical methods for fault

detection of transformers is the FRA method. This method,
due to its non-destructive nature and high sensitivity in fault
detection, has gained significant popularity among
researchers in this field. The subject of this paper focuses on
the FRA method.
A transformer can be primarily modeled using three
elements: resistor, inductor, and capacitor, each representing
various parts of the transformer. As a result, a unique
frequency response can be obtained for a healthy and non-
defective transformer (which referred to as a “signature” or
“fingerprint”). Consequently, any mechanical or electrical
defects, originating from various sources, can alter the FIGURE 2. Power transformer frequency responses comparison [6].
transformer's structure, thereby changing the values of its
modeling elements and consequently leading to a One of the main advantages of FRA testing is that it enables
corresponding change in the frequency response of the quick and accurate identification of the type and location of
equipment. Mechanical failure usually occurs due to faults. By identifying the faulty component, the duration and
improper transport of this equipment, winding deformation cost of repair can be significantly reduced. Additionally,
due to short circuit in the power grid and failure of the knowledge of the fault severity is also critical in determining
bushing or transformer core. On the other hand, electrical the appropriate course of action. For example, if a short-
faults have a variety of causes, such as loss of insulation circuit fault occurs multiple times between the winding loops
between the windings and inter-turn short-circuit in one part of the transformer, and with considerable
connections. The process of conducting the FRA test, as seen severity, it may be necessary to strengthen the insulation of
in FIGURE 1, involved by considering the two terminal pairs that part to prevent future failures. It is important to note that
of the transformer as input-output. Then a signal is applied the ultimate goal of research in this field is to achieve
to the input terminal and the response signal from the output effective and reliable fault detection. By identifying the type,
of the transformer is measured for different frequencies. severity, and location of faults, it becomes possible to
Then the ratio of the output signal to the input is considered prevent catastrophic transformer failures that can lead to
as the transfer function or the frequency response of the significant economic and environmental damage. Therefore,
transformer. In FIGURE 2, the magnitude and phase of the all efforts are made with the FRA test to achieve these goals,
transformer frequency response are drawn in two healthy and and researchers are continually seeking new and innovative
faulty states. The presence of any abnormality in this methods to improve the accuracy and efficiency of
diagram indicates the occurrence of a fault in this equipment. transformer fault detection which is the topic that is going to
In the following sections, further discussion is provided be addressed in this paper. The rest of the paper is consisting
regarding the characteristics of the applied voltage signal as of the following sections:
input, taking into consideration the configuration of the In section II, various methods of modeling transformers at
transformer for conducting this test. different frequencies are presented to provide a
In the field of transformer fault detection, the FRA test has comprehensive understanding of the behavior of this
emerged as a promising method to identify the type, severity, equipment to the readers.
and location of faults. By comparing the frequency response Regarding FRA, transformer modeling holds importance in
of a transformer in both healthy and faulty conditions, it is two ways:
possible to determine the changes that occur in the 1) As shown in section IV, one of the FRA interpretation
transformer's behavior due to the presence of a fault. methods is based on transformer modeling and
comparison of equivalent circuit parameters between the
healthy and faulty states.
2) Different faults can be simulated by modeling the
transformer and with altering the parameters of the
equivalent circuit. Subsequently, the FRA traces
obtained from the model can be compared with those
recorded from the actual transformer. This enables the
identification of fault type, severity, and location, which
will be discussed in section IV [7], [8].
In section III, the requirements for implementing the FRA
method are discussed. Specifically, the appropriate
configuration and transfer function for this task are
examined, along with the proposition of comprehensive
criteria for their suitability. In the following, the
FIGURE 1. FRA measurement layout [6].
2

characteristics and type of input reference signal to be number and its real and imaginary parts will be frequency
applied to the transformer for FRA will be discussed and the dependent [10]. Reference [10] shows that the resistance,
strengths and weaknesses of each of them will be expressed. capacitance and inductance values of the transformer model
Then, different solutions to achieve the frequency response are dependent on the frequency, and this issue complicates
of the transformer in a healthy state are presented, and it is the modeling of the transformer for the purpose of FRA and
shown that even if the transformer frequency response in the fault detection. Nevertheless, it is possible to consider the
healthy state is not available, it is still possible to detect the values of these parameters constant and independent of
fault severity. In section IV, first, various methods for frequency by accepting a small amount of reduction in
analyzing FRA results are introduced and compared. modeling accuracy.
Subsequently, attempts are made to propose different
methodologies which can cover the weaknesses of the FRA
method. In the following, the effect of winding type, residual
flux and other factors affecting FRA are investigated and
after fully understanding the behavior of these factors, the
problem of determining the appropriate frequency range for
the FRA method is expressed. The study demonstrates that
various types of faults affect specific regions of the
frequency spectrum, and they can also have distinct impacts
on the equivalent circuit parameters.
Finally, the FRA online method will be introduced and (a)
compared with the offline ones, and the challenges of the
online method will be presented. Section V is devoted to
summarizing the key findings and presenting
recommendations for further research in this field.

II. TRANSFORMER FREQUENCY MODEL

In FRA, transformer modeling at different frequencies can
be used as an approach for fault identification. In this
approach, first the transformer is modeled using an
equivalent circuit. After modeling, if there are mechanical or
electrical faults, one or more parameters of this model will
change and thereby altering the frequency response of this (b)
model. Eventually, by comparing and matching the FRA FIGURE 3. (a) Schematic of the FRA experiment for a ferromagnetic core.
trace obtained from the actual transformer with the one (b) Real and imaginary parts of the magnetic permeability coefficient of
generated from the equivalent circuit, the type and severity the core [3].

of any mechanical fault in the transformer can be identified.

In this section, the behavior of dielectrics and ferromagnetic B. Transformer modeling
materials in the presence of time-varying electric and B.1 Hybrid model of transformer
magnetic fields is explained. After describing the behavior
According to [11], based on the theory of the long
of these materials in the presence of time-varying fields,
transmission line (in which the line parameters are
transformer modeling at different frequencies is done.
distributed along the line and at small distances ∆x), a coil
has been modeled which itself consists of several disks and
A. Behavior of ferromagnetic and dielectric materials
in the presence of time-varying fields each disk consists of several loops (shown in FIGURE 4). As
Reference [9] shows that the magnetic permeability each disk in this method is modeled based on the theory of
coefficient of a material is frequency dependent (see traveling waves and the equivalent circuits of the disks are
FIGURE 3). The reason for this phenomenon is that connected to each other, the method is referred to as the
magnetic dipoles always tend to be in the direction of the hybrid method. In fact, in this method, the winding of each
magnetic field. As the frequency increases, since the phase is considered as several transmission lines (each disk,
direction of the field changes rapidly, these dipoles fall one transmission line) and each transmission line is modeled
behind the applied magnetic field and cannot quickly follow as a long transmission line based on the theory of traveling
the direction of its change. This lag of the dipoles leads to a waves. The parameters related to the mentioned model can
delay and manifests itself as a complex number. Therefore, be extracted through analytical relations (formulas) or in
the magnetic permeability coefficient of a material will be a order to achieve higher accuracy through finite element
complex number whose real and imaginary parts depend on method (FEM) [10], [12]–[14]. The mentioned hybrid model
frequency. The same behavior occurs for a dielectric in the is very complex and it is possible to present another model
presence of a time-varying electric field, which means that with a slight reduction in its accuracy, named ladder model
the electrical conductivity of these materials is a complex [15] .On the other hand, due to the simplifying assumptions
3

FIGURE 4. Transformer hybrid model with disk winding [11].

that exist in analytical relationships to determine the the values of the parameters are calculated through analytical
parameters of a hybrid model, the accuracy of modeling is relations [16]. By considering the capacitors’ voltage and
reduced. Also, parameterization can be determined through inductor current as state variables in the ladder model,
analytical relationships, but this requires access to different transfer functions can be defined. Using meta-
information about the transformer's structure and geometry, heuristic optimization algorithms, the parameter values of
which is not usually the case. Because transformer the ladder model are extracted. In this reference, the
manufacturers do not provide this information to users. objective function of the optimization problem is considered
Incorporation of different optimization techniques can be as a weighted sum of the differences between the three
considered as an alternative approach for solving the recent defined transfer functions and their measured values. By
problem. minimizing the mentioned objective function, the equivalent
circuit parameters of the ladder model are derived. As
previously mentioned, the model parameters can be
B.2 Transformer ladder model and methods of determining
extracted through simulation in FEM software. However,
its parameters
this method is only possible if complete information
Reference [15] presents the transformer ladder model. The regarding the transformer's structure is available. In [17],
idea is that in the hybrid method, each disk is considered as after using analytical formulas as the initial conditions for
a long transmission line and normally should be modeled the bacterial swarming optimization algorithm (BSA), it has
using distributed parameter model, but by accepting a slight been concluded that the parameters obtained from the
reduction in modeling accuracy, in this case, each or more optimization with BSA had more accurate performance in
disks are modeled as a lumped model. FIGURE 5 shows the transformer modeling in comparison with the analytical
ladder model of phase X of a three-phase transformer. Again, formulas. Therefore, it is concluded that in the ladder model,

FIGURE 5. The ladder model of phase X of a three-phase transformer [9].

it would be very desirable if the FEM could be used to find algorithm, deals with the problem of constructing a ladder
its parameters. But if only analytical relations are available, model based on driving-point impedance (DPI) as a network
it is better to use them as the initial conditions for the BSA transfer function in a similar way to the reference [17]. The
to achieve the parameters more accurately. In addition, the created model has acceptable accuracy. In reference [20], the
study used a genetic algorithm (GA) along with BSA. The determination of ladder model parameters was compared
results showed that the parameters obtained through BSA using analytical relations, particle swarm optimization (PSO)
were the most successful in accurately modeling the algorithm, and improved particle swarm optimization (IPSO)
transformer, followed by those obtained through GA and algorithm. The results showed that the parameters optimized
analytical relations, in descending order. It is worth noting with the IPSO algorithm were better than those obtained with
that in order to reduce the computational burden in the the PSO algorithm. Moreover, the PSO algorithm was found
optimization problem, the matching between the ladder to contribute to more accurate modeling of transformers
model transfer function and the measured one is not compared to using analytical relations. Reference [21] also
performed for the entire frequency range and only the deals with determining the parameters of the transformer
important points of the frequency response, i.e. the points ladder model using two optimization algorithms, GA and
corresponding to the resonance frequencies (peaks of the PSO, and the results show that this approach can accurately
transfer function) and anti-resonance frequencies (troughs of estimate the parameters of the transformer ladder model in
the transfer function), are considered. both healthy and fault states.
The reference [18] also uses a GA to adapt the frequency So, the frequency response resulting from the simulation of
response of the ladder model with the frequency response the model follows the measured frequency response very
obtained from the measurement. But this adaptation is not accurately.
like the reference [17] that is done only for the resonance and In reference [22], the authors attempt to determine the
anti-resonance frequencies, nor is it like the other references parameters of the ladder equivalent circuit using GA and
related to the transformer ladder modeling that is done for emphasize that in practice, the transformer winding structure
the whole frequency range of the adaptation, but between the is not homogeneous and therefore, the assumption of
two. That is, it is done for a series of frequencies that include equality of the parameters of the ladder model sections,
resonance and anti-resonance frequencies. The reference reduces the modeling accuracy. Therefore, it is better to
[19] using the artificial bee colony (ABC) optimization consider the parameters of each section as a separate variable
5

in the optimization problem. Also, if the optimization It is worth mentioning that the series capacitor value can be
problem is solved unconditionally, it may provide answers calculated through FEM or analytical relations. However,
that are practically impossible. For example, it is clear that this requires information about the structure and geometry of
the mutual inductance between the first and second sections the coil. While using the optimization algorithms, the values
must be greater than the mutual inductance between the first of the parameters of the ladder model, including the series
and third sections of the ladder model. Thus, in this capacitor, can be obtained. Reference [24] presents another
reference, firstly, an optimization problem with inequality method based on the estimation of the transfer function from
constraints is solved. Secondly, two different case studies, the measured frequency response, in order to extract the
namely homogeneous and non-homogeneous models for parameters of the ladder model, including the series
transformer windings, are considered. Finally, it has been capacitor. This reference by measuring the earth capacitance,
concluded that in practice, the parameters of different estimates the series capacitor. The process of determining the
sections in the ladder model should be treated as different value of the series capacitor is shown in FIGURE 6. First,
and distinct variables in order to improve the accuracy of the the impedance of the excitation point of the transformer is
model (non-homogeneous winding). But in [23], transformer measured and its amplitude and phase are obtained. Then,
is modeled by considering both equivalent circuits, i.e. the using the method presented in [25], the best transfer function
hybrid model and the ladder model, and determining their is adapted to the zero-pole-gain format (equation (1)), where
parameters through analytical relations as well as the GA, τ is real zero, 𝑧𝑖 and 𝑝𝑖 are zeroes and poles of transfer
and show that as expected, the results of hybrid model is function, respectively, which appear as complex conjugates.
𝑎
closer to the measured actual value and also the parameters While the value of the gain 0 is equal to lim 𝑠𝑍(𝑠) [24].
𝑏0 𝑠→∞
of both optimization models are closer to the actual measured
values than the parameters obtained through analytical 𝑎0 (𝑠 − 𝜏) ∏𝑛−1 ∗
𝑖=1 (𝑠 − 𝑧𝑖 )(𝑠 − 𝑧𝑖 )
relations. 𝑍(𝑠) = ( ) (1)
𝑏0 ∏𝑛𝑖=1(𝑠 − 𝑝𝑖 )(𝑠 − 𝑝𝑖∗ )

FIGURE 6. The process of estimating the value of the series capacitor of the transformer winding ladder model based on the impedance of the
excitation point.

On the other hand, the ladder model with n sections, where n calculated by an iterative method. By the way, this algorithm
is equal to the number of natural frequencies of the excitation is sensitive to its initial conditions, i.e., the initial values of
point impedance, is considered for the winding. For high the poles and also, the number of iterations and, more
frequencies (𝑠 → ∞), only the capacitors of the equivalent importantly, the degree of the estimated transfer function.
circuit are effective, and therefore, the impedance 𝑍(𝑠) The reference [28] provides a solution to improve this
(including 𝐶𝑔 and 𝐶𝑠 ) is extracted from the ladder model. By algorithm.
𝑎
multiplying 𝑍(𝑠) by s and setting it equal to 0 , an equation As explained before, the hybrid and ladder models both
𝑏0 describe the transformer's behavior over the entire frequency
is derived in terms of 𝐶𝑔 and 𝐶𝑠 . range. However, as shown in the next sections, to detect
By measuring the value of 𝐶𝑔 , the series capacitor 𝐶𝑠 is faults and their locations, modeling the transformer at low
calculated. Finally, the values of inductance and resistance and medium frequencies is necessary.
of the ladder model can also be measured from equation (2)
and for low frequencies (𝑠 → 0) [25]. B.3 Low and medium frequency model of transformer
In this section, two different transformer models with two
lim 𝑍(𝑠) = 𝑅𝑑𝑐 + 𝑠𝐿𝑒𝑞 (2)
𝑠→0 approaches are presented in low (lower than 2 kHz) and
medium (between 2 kHz and 20 kHz) frequencies.
By measuring the equivalent inductance 𝐿𝑒𝑞 and through a
repetition-based algorithm, the self and mutual inductance
B.3.1 Duality-based lumped equivalent circuit of the
values of different parts of the ladder model are extracted.
transformer
The resistance of each section of the ladder model is also
In [29]–[36], the modeling of the transformer in low and
obtained by dividing 𝑅𝑑𝑐 by the total number of sections.
medium frequencies is discussed, which is based on the
Another point that should be mentioned is that the number of development of the transformer reluctance model. The
parts of the ladder model is equal to the number of natural modeling process is presented as follows.
frequencies seen in the impedance of the excitation point.
Considering the paths of flux in the transformer, at first, its
However, not all natural frequencies are visible in the
reluctance model is extracted (FIGURE 7-a and FIGURE 7-
frequency response diagram due to factors such as high
b). Then, based on the duality principle, each reluctance is
damping at high frequencies resulting from the skin effect. replaced by an inductance, voltage source by current source
To address this issue, a repetition-based algorithm is used to and node by mesh (FIGURE 7-c and FIGURE 7-d). On the
calculate the number of natural frequencies in the excitation
other hand, due to the ratio of turns of transformer coils, any
point impedance, as well as the appropriate number of
current source becomes an ideal transformer. Finally, by
sections for the ladder model [26].
adding the resistors that model transformer losses and the
It is worth noting that the unknown values of transfer
capacitors between each of the high voltage and low voltage
function (gain, zeroes and poles) can be obtained in various coils as a lumped element along with ground capacitors, the
ways, such as the vector fitting (VF) method, which operates final model for a three-phase transformer is derived
on the basis of the least squares estimation (LSE) method
(FIGURE 8). The values of inductances and the resistances
[27]. The basis of the VF method is that it adapts the best
of equivalent circuit can be extracted through three tests
transfer function of degree n which is equal to the number of
including short-circuit, open circuit and zero sequence tests.
measured impedance poles, to the transfer function

(a) (b)

FIGURE 7. Model extraction procedure of duality based lumped equivalent circuit of transformer [33].

FIGURE 8. Lumped duality based equivalent of an Yy6 transformer [33].

In terms of measurement, the capacitors of this model are duality based lumped model does not have enough accuracy,
divided into two categories: measurable and non- which is also confirmed by reference [30].
measurable. Measurable capacitors include ground Based on [32], it is suggested to use the duality based lumped
capacitors and capacitors between high and low voltage model for the analysis in low and medium frequencies,
windings, and non-measurable ones include series capacitors because its parameters are easily determined through
and capacitors between high voltage windings as well as measurement. But for the analysis in high frequencies, the
capacitors between low voltage windings. hybrid and ladder models are used for their high accuracy.
The values of measurable capacitors are extracted through Also, the values of the capacitors of different segments of the
capacitive impedance test. The value of the series capacitor hybrid model can be easily found from the values of their
cannot be measured directly and, therefore, an estimated corresponding capacitors in the duality based lumped model,
value is used. It is possible to estimate the capacity of the taking into account the number of segments and without the
series capacitor with high accuracy with the approach need to use analytical relations. This can be seen as an
mentioned in the previous section [24]. Therefore, the values advantage, because analytical relationships require complete
of the parameters of the duality based lumped model are information about the structure of the transformer [36].
calculated by measuring the impedance of the excitation
point [29]–[33]. Since the parameters of duality based B.3.2 Multi-cell equivalent circuit of the transformer
lumped model are calculated based on different input In [37], taking into account the simplifying assumptions in
impedances, this approach is known as impedance method the transformer reluctance model described in previous sub-
[32]. Also, the values of the parameters of the duality based section, using the impedance measured in each phase of the
lumped model can be calculated through the analytical transformer and with an optimization approach, the
relationships and the information provided by the parameters of the model at low and medium frequencies are
transformer manufacturers (manufacturer data sheet) [33], determined. The process of modeling and determining its
[34], [36]. parameters is presented as follows.
Reference [35] uses the zero-crossing method based on First, the input impedance of the different phases of the
waveform data to determine the parameters of the duality transformer is measured (FIGURE 9).
based lumped model. In [33], a comparison is made between As it can be seen in FIGURE 9, two common features in all
the low frequency model of the transformer and the hybrid three-limb core-type transformers can be pointed:
model and the results show that both duality based lumped 1) The length of the magnetic path of the middle phase is
model and hybrid models can be used for FRA analysis in less than the length of the magnetic path of the lateral
low and medium frequencies with very good accuracy [31]. phases.
But in high frequencies, the hybrid model is more suitable 2) The magnetic path of the lateral phases is equal to each
for analysis. In some medium and high frequencies, the other due to the existing symmetry, and for this reason,

FIGURE 9. (a) Flux flow paths in three-phase transformer. (b) Amplitude and phase of input impedance measured in different phases [37].

the reluctance of the lateral phases are equal to each presented model are obtained by minimizing the following
other. objective function for each phase (here phase U) and for m
samples of the frequency response (here 𝑍𝑈 ) [37].
This factor causes the measured frequency response curve of
2
lateral and middle phases to be different. Therefore, two 𝑚 1 1
|𝑌𝑈 (𝜔𝑖 ) − − 𝑗. (𝐶𝑈 𝜔𝑖 − )|
resonance peaks are created in the frequency response of 𝑅𝑈 𝐿𝑈 𝜔𝑖 (4)
𝐸=∑
lateral phases due to asymmetric magnetic paths C1 and C2 |𝑌𝑈 (𝜔𝑖 )|2
𝑖=1
in FIGURE 9. Meanwhile, a resonance peak is created in the
frequency response of the central phase due to the Comparing the results of the model simulation with the
symmetrical magnetic paths C3 and C4. measured values indicates that the proposed model can
3) After measuring the input impedance of different accurately capture the transformer behavior in low and
phases, considering the flux paths, the reluctance model medium frequencies.
of the transformer is extracted (FIGURE 10-a). Based References [38] and [39] also use the impedance measured
on the duality principle and by adding capacitor and in each phase of the transformer. Also, their method is based
resistor, which respectively model the capacitive effects on the multi-cell equivalent circuit, which each cell is
of windings and transformer losses, the equivalent equivalent to a parallel R-L-C circuit. As a result, for each
circuit of FIGURE 10-b is obtained. The elements of phase of the transformer shown in FIGURE 11, the values of
each phase, in general, can be shown with an impedance the model elements in FIGURE 10-b are extracted.
(FIGURE 10-c). By comparing the reluctance model of
FIGURE 10-a with the reluctance model of section
II.B.3.1, it can be seen that the complexity of the model
presented in the previous section has been reduced. Now
the problem is how by measuring the observed
impedances of each phase of the transformer which are
denoted by 𝑍𝑚𝑈 , 𝑍𝑚𝑉 and 𝑍𝑚𝑊 respectively, the (a)

impedances of each phase of the model, i.e. 𝑍𝑈

(including 𝑅𝑈 , 𝐿𝑈 and 𝐶𝑈 ), 𝑍𝑉 (including 𝑅𝑉 , 𝐿𝑉 and 𝐶𝑉 )
and 𝑍𝑊 (including 𝑅𝑊 , 𝐿𝑊 and 𝐶𝑊 ) are extracted.
According to FIGURE 10-c, each of the measured
impedances 𝑍𝑚𝑈 , 𝑍𝑚𝑉 and 𝑍𝑚𝑊 can be expressed through
(b)
three equations, in terms of the impedances of each phase of
the model, i.e. 𝑍𝑈 , 𝑍𝑉 and 𝑍𝑊 (equation (3)) [37].
𝑍𝑚𝑈 = 𝑍𝑈 ∥ (𝑍𝑉 + 𝑍𝑊 )
{ 𝑍𝑚𝑉 = 𝑍𝑉 ∥ (𝑍𝑊 + 𝑍𝑈 ) (3)
𝑍𝑚𝑊 = 𝑍𝑊 ∥ (𝑍𝑈 + 𝑍𝑉 )
(c)
In this case, by ordering these three equations, it is possible
FIGURE 10. (a) Transformer magnetic circuit. (b) Transformer core
to obtain the impedances 𝑍𝑈 , 𝑍𝑉 and 𝑍𝑊 and of course, their model. (c) Transformer core model represented by three impedances ZU,
corresponding admittances in terms of 𝑍𝑚𝑈 , 𝑍𝑚𝑉 and 𝑍𝑚𝑊 . ZV and ZW [37].
Finally, the values of the elements of each phase of the
9

appropriate configuration and transfer function pair to

analyze the frequency response, selecting the appropriate
voltage range for the FRA method, types of reference signals
that can be used for the FRA method is discussed and
compared. Furthermore, different methodologies to achieve
the transformer frequency response in a healthy state with a
review over the possibility of detecting the fault severity
FIGURE 11. Four-cell equivalent model [38].
especially when the transformer frequency response is not
available in the healthy state, are presented and discussed.
A parallel connected R-L-C circuit has one resonance
frequency. Therefore, a two-cell equivalent circuit is A. Selection of the appropriate configuration and
considered for the lateral phases and a single-cell equivalent transfer function pair
circuit is considered for the middle phase. Regarding FRA, it is important to note that a transformer has
Having the input impedance of each phase obtained through multiple terminals, and the conditions of these terminals can
measurement, the values of the elements of each cell are vary depending on whether they are grounded or open. The
calculated through the minimization of the objective main problem is to find the best configuration and,
function, similar to equation (4). Finally, the values of the consequently, the appropriate transfer function to perform
elements of the model in FIGURE 10-b can be easily the FRA test. Each configuration is equivalent to a new
extracted from the values of the multi-cell equivalent circuit network with a number of natural frequencies. Due to the
elements through mathematical relationships. possible cancelation of zeroes and poles, all or part of the
natural frequencies of such a network appear in the different
transfer functions. In order to increase the sensitivity and
III. FRA REQUIREMENTS
fault detection capability of the FRA test, it is necessary that
In this section, a comprehensive review on the requirement
all or most of the natural frequencies of the network are
of implementation of the FRA method is presented and
available. Therefore, as shown in FIGURE 12, it is essential
analyzed. In this war, various topics such as selecting the
to determine the best configuration and transfer function pair

FIGURE 12. Various configurations along with measurable quantities [40].

TABLE 1.
Classification of different network function-configuration pairs based on the highest number of natural frequencies in order to determine the
best network function-configuration to perform FRA test [40].

to achieve this goal. For this purpose, reference [40] visual comparison, because the interpretation of the results
investigates different types of configuration and transfer by numerical indices is more accurate and logical than the
function pairs (FIGURE 12) according to the number of interpretation based on observation. The results indicate that
natural frequencies that appear in them and divides them into bfloat-oc and pfsg-oc configurations, which are placed in the
four categories: good, reasonable, medium and bad (TABLE good and reasonable category in [40], have the best
1). Another important point is that, among those performance for identifying mechanical faults.
configuration and transfer function pairs suitable for FRA According to the standards available to researchers and
test, some of them have more sensitivity for a specific fault craftsmen so far, four configurations are usually
and are more suitable for detecting that type of fault. recommended as shown in FIGURE 13 by selecting the
Therefore, it is possible to choose the appropriate transfer voltage as the transfer function [43], [44]: end-to-
configuration and transfer function pair for a specific fault to end open circuit (EEOC), end-to-end short circuit (EESC),
perform the FRA test [41]. Reference [42] deals with which capacitive inter-winding (CIW), and inductive inter-winding
configuration with the specified transfer function is more (IIW) configuration.
sensitive to the occurrence of three types of AD, RD and
axial bending (AB). To achieve this goal, five different
configurations with three transfer functions are considered.
The procedure used in this reference to determine the best
configuration and transfer function pair is such that various
faults are applied and by observing the frequency response
changes, it is investigated which configuration and transfer FIGURE 13. Different FRA configurations: a) EEOC b) EESC c) CIW d) IIW
function pair has the highest sensitivity for the specified [43].
fault. The results show that for AD and RD faults, transfer
voltage ratio configuration and for AB fault, end-to-end It is worth noting that although there are not many standards
voltage ratio configuration have more sensitivity and are and technical reports available on the topic of FRA, some
more suitable. It is worth mentioning that the two references are particularly valuable. For example, a technical
configurations of transfer voltage ratio and end-to-end report published by CIGRE focuses on the interpretation and
voltage, being in the good category, have also been approved procedure of performing FRA tests [45]. A similar guideline
by reference [40]. Reference [7] follows the same procedure has been presented by IEEE with a focus on oil-immersed
for the AD, RD and disc space variation (DSV) faults, but to transformers [46]. Chinese National Development and
select the best one, it resorts to numerical indices rather than Reform Commission (NDRC) has issued a standard about

the primitive requirements in FRA on transformer winding in the time domain and therefore, there is a need to convert
deformation [47]. IEC has also published a standard in the short time Fourier transform in order to transfer it to the
association with the measurement techniques and equipment frequency domain. Besides, the resolution is adjustable in the
in FRA [44]. Finally, one of the most recent practical SFRA method, but in the IFRA method, the resolution is
guidelines about FRA is presented in reference [48] which fixed. In relation to the subject of adjustable resolution in the
discusses the measurement fundamentals and methods of SFRA method, it's important to note that various standards
result interpretation in FRA. require a minimum resolution of at least 200 points per
decade or higher in order to accurately detect resonance
B. Different methods for performing FRA test frequencies. [51]. In the IFRA method, it is possible to
Another issue is the determination of the type of reference measure several outputs, and therefore, several transfer
signal for performing the FRA test, which consists of two functions can be defined. However, in the SFRA method,
types [49], [50]: only one output can be measured at any time. Testing by
1) Impulse frequency response analysis (IFRA) method: In IFRA takes less time compared to SFRA which may take up
this case, a wide range of frequencies is injected into the to 10 minutes.
transformer by a low voltage impulse signal, in order to Considering the advantages and disadvantages of the
perform FRA. Then, the output signal is measured and mentioned methods, SFRA method is usually used to analyze
after passing through the filter to eliminate noise, using the transformer frequency response.
the short time Fourier transform, it is transferred to the
frequency domain and thus, the transformer frequency C. Selection of the appropriate voltage amplitude for
response is extracted. the SFRA test
2) Sweep frequency response analysis (SFRA) method: In As mentioned earlier, in the SFRA analysis, a sinusoidal
this method, a sinusoidal signal with a low amplitude signal with a low amplitude, from low frequencies to high
from low frequencies to high frequencies is swept and frequencies, is applied as a reference signal to the
its output is measured. In this case, the output-to-input transformer. This section discusses why a sinusoidal signal
ratio is considered as a transfer function. with a low amplitude and close to zero is applied to the
transformer. FIGURE 14 shows the frequency response
The advantages and disadvantages of the two mentioned curve of the transformer for the EEOC configuration and two
methods can be examined from different aspects. different input amplitudes (5 and 10 volts). As can be seen,
The SFRA method requires less measurement equipment the frequency responses differ only until the first anti-
than the IFRA method (oscilloscope, Rogowski coil, and resonance frequency (less than 10 kHz) is reached. After that
other equipment). SFRA method has a higher signal-to-noise point, the responses match each other. Before the first anti-
ratio compared to the IFRA method. Also, in the SFRA resonance frequency, there is only an inductive effect with
method, the response exists in the frequency domain in the resistance (no capacitive effect is seen). Since the applied
form of amplitude and phase. But in the IFRA method, it is voltages are much lower than the nominal voltage of the

FIGURE 14. Comparison of FRA curves for two different input voltage values [52].

transformer, the core losses can be ignored, and the of any defects and faults (e.g., right after the installation
equivalent circuit considered as inductance. According to the of the transformer). Since this information is usually
magnetization curve of the core and the magnetic rarely available, two other methods are used.
permeability coefficient shown in FIGURE 15, the operating 2) Construction-based approach: Due to the symmetrical
point G is for applying 5 volts and operating point J is for property that exists in the construction of some three-
applying 10 volts. There is a significant difference in phase transformers, the result of performing the FRA
magnetic permeability coefficient values between these two test on other phases of the transformer can be used as a
points. In fact, there is a direct correlation between the input fingerprint. It should be noted that this similarity is not
voltage amplitude and the permeability of the transformer enough for evaluation in transformers with zigzag
core. Therefore, with a slight change in the input voltage coupled windings.
amplitude from 5 to 10 volts and while the transformer is 3) Type-based approach: If the fingerprint of the
healthy, the inductance experiences significant changes. transformer is not available through the above two
This change in inductance leads to a change in the frequency methods, it is possible to use the FRA test results of
response of a healthy transformer, and it is wrongly assumed similar transformers in terms of design and construction
that a fault has occurred inside the transformer, while the that were produced in the same factory. Due to the
transformer is healthy. However, if the input voltage existence of tolerance in manufacturing, various
amplitude is close to zero, inductance changes can be ignored statistical evaluation methods are provided, with the
for small changes in the input amplitude. It is worth noting help of which, the effect of the design and construction
that from the first anti-resonance frequency onwards, this process is separated from the effect of defects and
problem will not exist for a change in the input voltage failures.
amplitude [52].
These three methods are summarized in the FIGURE 16.

FIGURE 15. Iron core's magnetization curve with its magnetic

permeability coefficient [52].

In relation to the low voltage amplitude applied for the SFRA

test, it should also be noted that if the voltage amplitude is
too low, the response may be affected by ambient noise and
FIGURE 16. Different methods for measuring the frequency response of
impair performance [51]. Fortunately, the analyzer can solve the healthy transformer [53].
this problem by increasing the number of measurements
(repetitions) and eliminating the noise.
E. Fault severity detection in case of lack of access to
D. Transformer FRA in the healthy state the fingerprint
The FRA method relies on comparing the frequency The FRA test involves comparing the frequency response of
response of a transformer in both healthy and faulty states. a transformer in two states: healthy and faulty, for a specific
Therefore, it's crucial to have the precise frequency response configuration. In some cases, the frequency response of a
of the healthy transformer, which is called fingerprint of the healthy transformer is not available. In this situation, some
available transformer. Because without access to the methods have been presented in the references to identify the
fingerprint of the transformer, it is almost impossible to fault without using the frequency response of the healthy
detect the fault type, severity and the location. There are transformer and comparing it with the frequency response of
three different approaches to achieve the frequency response the defective transformer. For instance, reference [54]
of a healthy transformer, which are discussed as follows [53]. proposes a method is proposed in which the frequency
1) Time-based approach: This approach is the most response of the transformer is compared in a faulty state for
accurate and reliable method to obtain the frequency different configurations (standard configurations, i.e. EE,
response of a healthy transformer (fingerprint). In this EESC , CIW and IIW) with each other, the severity of the
case, the fingerprint is obtained from the previous FRA mechanical fault can be detected. This reference uses two
test data, when it is ensured that the transformer is free
13

indicators, including CC and Euclidean distance (ED) to this case is called a “reference circuit”. After a fault occurs,
evaluate and compare the frequency response of the a new circuit is synthesized again, and by comparing its
defective transformer for two configurations EEOC and elements with the elements of the reference circuit, the fault
EESC and for axial displacements (AD) and radial location and severity is determined. The results indicate the
deformations (RD) faults. In this regard, this result has been high accuracy of this method in fault detection. In [41], by
obtained that with the increase in the fault severity, the extracting the transformer model in both healthy and fault
values of the mentioned indicators change monotonically. states, the location of various faults such as disc deformation,
Hence, the fault severity can be identified with this method. winding short circuit, aging of insulation, etc. has been well
Also, by comparing the frequency response of the identified. In [26], based on an iterative algorithm, it deals
transformer for CIW and IIW configurations, a similar result with the synthesis of the transformer ladder model based on
has been obtained. the measured excitation point impedance, before and after
the fault occurs. Then, based on changing the parameters of
IV. INVESTIGATION AND INTERPRETENTION OF the ladder model, the fault type, severity, and location have
FREQUENCY RESPONSE been identified. The good performance of the mentioned
Various factors can lead to faults and changes in the method in fault detection has also been confirmed in [57]. In
transformer frequency response, including: inter-turn short- [39], using the low and medium frequency models of the
circuit fault, DSV fault, AD fault, RD fault, bushing fault, transformer, the fault involving the magnetic core and its
etc.. This section will first discuss various methods of severity have been correctly identified. In [58], based on the
analyzing FRA results, followed by an introduction on the improved lumped circuit (ILC) model, the changes in energy
methods that come along with FRA method and cover its transfer and distribution in the elements of the mentioned
weaknesses. Subsequently, it will examine the effect of model for inter-turn fault and its effect on the frequency
several important factors on the transformer frequency response of the transformer have been investigated. Short
response, including the type of winding, the important issue circuit resistance plays a key role in changing the frequency
of selecting the frequency range for FRA and its sub-ranges response when a short circuit occurs between the turns of the
as well as the effect of different faults on equivalent circuit coil. Because it causes a change in the path of energy transfer
parameters. It is also shown that different faults will mainly between adjacent turns, which increases the resonance
affect which of the identified sub-ranges of frequency. frequencies. On the other hand, due to the increase in losses
Finally, the online FRA method is discussed and a in the transformer winding, it leads to a change in the
comparison with to the offline method is presented as well. frequency response amplitude at resonance frequencies. The
results show that by reducing the short circuit resistance
A. Different methods of interpreting FRA results between the winding turns, the frequency response
The comparison of the frequency response of the transformer amplitude at the resonance frequencies experiences a U-
in both healthy and defective states and its interpretation can shaped trend and also, the resonance frequencies increase.
be done by an expert through visual inspection. Reference Reference [59] examines the frequency responses of the
[55] the frequency response of a transformer for open circuit windings of a single phase transformer with a large voltage
and short circuit faults is examined to develop a guideline ratio. It also approximates the first anti-resonance frequency
approach. The reference discusses the interpretation and of the FRA diagrams of the low voltage (LV) and high
analysis of frequency response based on human experience, voltage (HV) windings and highlights the relationships
which may not always be accurate, as operators may not have between FRA properties and the electrical components in the
the necessary skills to evaluate the results in some cases. To equivalent circuit network. The derived results present an
address this, researchers have developed methods to remove intuitive quantitative relationship between the values of the
the human interface, which is going to be discussed in the electrical winding components and the corresponding FRA
following subsections. characteristics.
Each of the following FRA interpretation methods has TABLE 2 illustrates the relationship between the parameters
advantages and disadvantages. One of the drawbacks of FRA of the electrical model of the transformer and the mechanical
analysis as a fault detection tool is that there is no specific faults [21]. In other words, based on this table, it is shown
procedure and standard for analyzing and interpreting the that each of the mechanical faults of the transformer leads to
measurements. Reference [56] compares and states the the change of which of the parameters of the model.
advantages and disadvantages of some of the methods
mentioned below. TABLE 2.
Transformer model parameters and its associated mechanical faults [21].
A.1 Model-based FRA Parameter TYPE OF FAULT
Reference [57] deals with identifying the fault location based Self / Mutual Winding / Core deformation, shorted turns
on the synthesis of a lumped-parameter ladder network. For Inductance and disk space variation
Disk space variation, winding buckling and
this purpose, the impedance of the excitation point is Shunt Capacitance
loss of clamping pressure
measured by performing the FRA test for the healthy Series Capacitance Insulation degradation and disk movement
transformer and based on it, a lumped-parameter ladder Resistance Shorted turns
network is synthesized using an iterative algorithm, which in
14

TABLE 3.
Different statistical indicators to analyze FRA results [60].

of course, improving their performance in fault detection.

Because different faults do not necessarily affect the whole
A.2 FRA based on statistical indicators frequency range and only affect certain intervals of it. The
In order to remove the operator from the process of analyzing results indicate that DABS, which is introduced in TABLE
the results of FRA, it is possible to compare the frequency 3, F-test and T-test indicators, which are detailed introduced
responses in two healthy and defective states using statistical in [60], are more sensitive in detecting transformer winding
indicators. In this section, by introducing different statistical faults, among which F-test, due to its greater sensitivity, can
indicators, it is tried to interpret the results of FRA and make well reflect the difference between two healthy and faulty
a comparison between these indicators. Various indicators states. In [61], the inter-turn fault and the interpretation of
can be used for this purpose, which are presented in TABLE FRA have been investigated based on two indicators, CC and
3 [60]. spectrum deviation (SD).
In connection with the discussion of indicators, their In this case, similar to reference [60], the entire frequency
sensitivity, monotonicity and linearity are crucial for the range in the FRA test is divided into four decade frequency
changes in fault severity. Because it is only in this way that bands and the mentioned indicators are calculated in each of
it is possible to identify the fault type, severity and location these four bands. The following outcomes have been
based on the indicator value. obtained:
Reference [60] introduces and examines statistical indicators o The lowest amount of correlation corresponds to the
and, statistical tests are applied to FRA records in normal and lowest frequency band. Because for the inter-turn fault,
faulty conditions in order to interpret the results of applying the frequency response changes in this band are more.
SC fault for different severities. Examination and o Reliability and sensitivity of SD indicator in detecting
comparison of statistical indicators and statistical tests are inter-turn fault is more than CC indicator. Because with
divided into two subsections. Once, the statistical indicators the increase in the severity of the mentioned fault, the
and statistical tests are applied for the whole frequency range shape of the frequency response is almost constant, but
and in the second case, the whole frequency range is divided the resonance frequencies are moved to higher
into four decade frequency bands and statistical indicators frequencies, which is mentioned in the reference [60]
and statistical tests are calculated in these four bands. This about the weakness of the CC indicator in such a case.
action leads to increasing the sensitivity of the indicators and

Reference [62] tries to find the best indicator in terms of o For the radial displacement fault, all indicators except
sensitivity, capability and reliability by examining different MM have a positive performance to identify the fault in
indicators. For this purpose, CC, ASLE, DABS and MM all frequency bands. Also, the values of the indicators in
indicators are first introduced to interpret FRA. In this case, the second band have more deviations compared to the
the entire frequency range of the FRA test is divided into first and third bands.
three-decade bands and the values of these indicators are
In reference [68], the objective is to identify the location and
calculated in each of these frequency bands. The results show
extent of the RD fault. The study reveals that when a fault
that ASLE and MM indicators are more confident and
occurs, the direction of the shift of the frequency response
reliable in order to identify faults and make decisions,
amplitude at resonance frequencies can be used as an
compared to CC and DABS. The weakness of the CC
indicator to identify the fault location. Consequently, the RD
indicator has also been confirmed in [60] and [61]. Reference
fault location can be determined using this method. The
[13] also compares the three indicators, namely CC, SD and
authors of the reference propose the use of the index of
ASLE, for detecting minor AD and RD faults. The study
frequency deviation (IFD) and the index of magnitude
demonstrates that the SD and ASLE indicators outperform
deviation (IMD), to assess and identify the fault severity. But
the other in identifying these faults. Moreover, the
since these indices are not always proportional to the degree
effectiveness of the SD indicator is further confirmed by the
of deformation, SD index, as its proper performance has been
findings of [60]. Reference [63] also introduces weight
confirmed in [13], [61], [67], has been used to determine the
functions Fa and Ff as follows:
𝑎2𝑗 fault severity.
𝐹𝑎 = ∑ × 𝑤𝑎𝑗 (5) Reference [69] compares the reliability and sensitivity of
𝑎1𝑗
𝑗 several indicators, including standardized difference area
𝑓2𝑗 (SDA), correlation factor (ρ), indices of frequency and
𝐹𝑓 = ∑ × 𝑤𝑓𝑗 (6) amplitude deviation (IFD and IAD), weight functions (Wa,
𝑓1𝑗
𝑗 Wf), SD, weighted normalized difference (WND), and
In the above equations, a1j and a2j, are the transformer stochastic spectrum deviation (σ), in identifying the fault
frequency response amplitude, before and after the type, severity and location for disc-space variation (DSV),
displacement of the jth resonance frequency, respectively. AD and RD mechanical faults. In order to increase the
The values of f1j and f2j are the jth resonance frequency before sensitivity of the mentioned indicators, the entire frequency
and after displacement due to AD fault in the transformer, range has been divided into three bands and the following
respectively. Furthermore, waj and wfj are the weighted results have been obtained:
coefficients of the jth resonance frequency for amplitude and o The SDA indicator shows an irregular behavior in
frequency, respectively. These two indicators are used to relation to the fault type, severity and location, and
interpret the FRA test and the results show that this approach, therefore, it is not suitable for the interpretation of FRA.
independent of the defined transfer function, is a reliable o The correlation factor indicator also has less reliability
method to evaluate the FRA test. In [64], different indicators than other indicators to identify the fault level, which
of sum squared error (SSE), sum squared ratio error (SSRE), has been confirmed in [60]–[62].
sum squared max-min ratio error (SSMMRE), CC and ASLE o Wa and Wf indicators are able to identify the fault
are compared for different faults that the results indicate high location and severity. The high capability of these two
accuracy of the ASLE indicator. The high capability of the indicators is also confirmed by [63].
ASLE indicator in fault detection is also confirmed in [13] Reference [70] deals with determining the location of the
and [62]. Similarly, in [65], it is concluded that CC and MM inter-turn fault by using the absolute sum of logarithmic error
indicators are not sensitive to the variation between normal (ASLE) indicator. The study considers two transfer
and short circuit conditions, and conversely, ASLE, MSE functions, namely transfer voltage and input impedance, to
and SD indicators are sensitive and show a trend due to fault. perform the FRA test. The results show that when the fault
The reference [66] by introducing the R-square technique, location changes from the inner layers to the outer layers of
which is based on two indicators, sum of squares of the the coil, the values of the ASLE indicator changes in the form
regression (SSR) and total sum of squares (SST), shows that of a symmetrical pattern (for the input impedance transfer
this methodology performs better in comparison with cross function) or in the form of a decreasing trend (for the voltage
correlation coefficient (CCF) for vertical displacement transfer function). Therefore, the mentioned indicator can
problems of deviations. identify the fault location very well, as mentioned in the
Reference [67] examines and compares CC, MSE, ASLE, references [13] and [62] about the reliability and appropriate
MABS, MM and SD indicators for inter-turn fault and radial sensitivity of this indicator to identify the fault.
displacement. Again, in order to increase the sensitivity, the In previous references, in order to increase the sensitivity of
entire frequency range is divided into three-decade bands. statistical indicators, the entire frequency range of the FRA
The results show that: test is divided into smaller bands and the presented indicators
o For the inter-turn fault, all indicators in the first band are calculated in each of these bands. Reference [71] seeks
and all indicators except MM in the second band had an to identify specific frequency bands in which the frequency
acceptable performance.
16

response changes are proportional to the fault severity. In that the use of frequency response phase data in the
other words, suppose TFbase is the transformer frequency calculation of indices improves their performance in fault
response in a healthy state and TFi is the frequency response detection. For this purpose, the CC and ED indicators were
of the faulty transformer. In this case, the difference between applied to the real and imaginary parts of the frequency
these two frequency responses is equal to [71]: response and the results show that this action has led to an
∆𝑇𝐹𝑖 = 𝑇𝐹𝑏𝑎𝑠𝑒 − 𝑇𝐹𝑖 (7) increase in the sensitivity of these indicators. Also, by
introducing the complex distance (CD) indicator, these
This reference intends to find bands of the frequency references show that this indicator has more sensitivity than
spectrum of the FRA test in which ∆𝑇𝐹𝑖 can be written in the other indicators. In fact, the CD indicator calculates the
form of a monotonic function relative to the fault severity. complex distance between samples of two frequency
Such frequency ranges are called diagnostic bands. responses by simultaneously using data related to the
Therefore, it is recommended to interpret and analyze the amplitude and phase of the frequency response, while the ED
frequency response in these bands. indicator only calculates the Euclidean distance between the
Until now, various numerical indices have been employed amplitude samples of two frequency responses. It is worth
for interpreting FRA. Although these indicators are mentioning that the sensitivity of the CD indicator is even
straightforward and their implementation is easily possible, higher than the sensitivity of the ED-real part and ED-
they have two drawbacks. First, researchers have not imaginary part indicators, as shown in FIGURE 17. Finally,
provided explanations about the origin of these indices. reference [8] shows that the CD indicator performs better
Secondly, most of these indicators do not show regular and than the ED indicator in terms of uncertainty.
linear behaviors [72]. Reference [72] uses a probabilistic Reference [74] further explores and compares different
reasoning to extract the ED indicator, which can identify the indicators and configurations of the FRA test, especially
fault type and severity. First, a comparison is made between from their influence of uncertainties point of view. In this
the ED indicator and other indicators such as correlation reference, the indicators of standard deviation (SD), integral
coefficient, weight functions, etc., and the results show that of absolute difference (IA), absolute sum of logarithmic error
the ED indicator is more linear than other indices in detecting (ASLE), root mean square error (RMSE), standard deviation
the severity of AD and RD faults. (σe), stochastic spectrum deviation (σ), SSE, SSRE, and
Also, in order to detect the fault type, first the entire SSMMRE are taken into account.
frequency range is divided into several frequency bands and The results show that the end-to-end configuration with SSE
the ED indicator is calculated in each of these frequency and σe indicators are less subject to uncertainties.
bands, which is called regional Euclidean distance (RED). In [75] and [76], statistical indicator-based boundary
The ratio of maximum ED to minimum ED, which is called conditions for various operational states of transformers, i.e.
MMR, can act as a discriminator between AD and RD faults. normal, suspicious, and critical, are explored using a pattern
The aforementioned indicators were only applied to the recognition-based method, named bolstered error estimation,
frequency response amplitude. References [8] and [73] show and the level of confidence in decisions made are also

FIGURE 17. Comparison of different numerical indicators versus axial displacements for: (a, c) the inductive inter-winding and (b, d) the capacitor
inter-winding connections. (a, b) are experimental results whereas (c, d) are simulation results [8].

estimated. The aim of [77] is to try to collect all the indicators fault detection process is shown in FIGURE 18. First, using
in a single context, compare and classify them based on their the transformer model, the RD fault is simulated in various
characteristics and introduce more appropriate indicators as locations and for different severities. Then, the transfer
standard indicators. This reference also defines some criteria function of the transformer is estimated in the form of a
for a suitable index such as monotony, linearity, sensitivity, rational function in two states, healthy and fault, using the
dependence on data points and index ratio. In [78], VF method. In this case, the Nyquist diagram can be drawn
experimental FRA measurements are performed under for the occurrence of faults in various places. Then, based on
various faults. It is concluded that some faults, e.g., inter- the shift direction of the Nyquist diagram, a feature is
disk and short circuit, affects the FRA response at medium extracted for the occurrence of faults in different places.
and high frequency ranges, while some other faults, e.g. axial Based on this feature, the fault location can be identified.
winding displacement, radial deformation, and winding After determining the fault location, the fault severity is also
bushing fault, affects the FRA response at high frequency calculated based on the sum of absolute displacement of
range. Besides that, the loss of clamping pressure displays an poles (SDP) indicator. This reference states that the SDP
effect across the entire frequency range. The results also indicator has the necessary sensitivity and linearity to
indicates that the ASLE, along with other investigated identify the fault severity.
indicators, performs well in detecting small variations.
Reference [79] utilizes bootstrap sampling technique to
simulate the effect of many experiments and overcome the
difficulty of limited available measured data in FRA
interpretation and classification, through establishing a large
set of statistical indicators-specific decision boundaries.
Reference [80] proposes an innovative fault segmentation
and localization technique based on FRA data. This
technique is based on regression analysis. The proposed
scheme specifies different transformer conditions, such as
healthy windings, axial and radial winding deformations,
core deformation and electrical faults.
Reference [81] emphasizes the importance of clustering and
optimization methods to improve the performance of the
current FRA interpretation. In this regard, this reference
extracts the necessary features from the measured transfer
functions of various healthy and faulty transformers. Then,
using the interval maximum to global maximum (IMGM)
indicator and employing the k-means method, the collected
data are sorted into different clusters and the centers of the
clusters are determined using the grasshopper optimization
algorithm (GOA). At the end, the distance of the new data
from the centers of the sorted clusters is determined and the
closest cluster to the new data determines the condition of
the transformer. Reference [82] deals with sensitivity
analysis of amplitude, phase, real and imaginary components
of transfer function for different faults. Moreover, the FIGURE 18. Algorithm of fault detection mentioned in [83].
accuracy of different indicators and four different
components of transfer function is investigated. The In [84], first, the transfer function is estimated in the form of
conclusion drawn from the analysis is that in comparison equation (8) for the transformer in two healthy and fault
with the amplitude, the phase, imaginary, and real states [84].
components of transfer function have more sensitivity to the 𝑎𝑚 𝑠 𝑚 + ⋯ + 𝑎1 𝑠 + 𝑎0
faults, respectively, and are more accurate for classification 𝑓(𝑠) = (8)
which is confirmed in references [8] and [73]. 𝑏𝑛 𝑠 𝑛 + ⋯ + 𝑏1 𝑠 + 𝑏0
In this case, by extracting the polynomial coefficients of the
A.3 FRA based on transfer function estimation numerator and denominator of equation (8), it is possible to
The method of determining the fault type, severity and calculate the indicator of faulted-intact relation which is
location by estimating the transfer function is also possible. shown as equation (9) [84].
The estimation-based approach is presented and discussed in
∑𝑚
𝑖=1|𝑎𝑓𝑖 | ∑𝑛𝑖=1|𝑏𝑓𝑖 |
this section. In [83], an attempt has been made to determine 𝐹𝐼 = [ ] × [ ] (9)
the location and extent of the RD fault by using the ∑𝑚
𝑖=1|𝑎𝑛𝑖 | ∑𝑛𝑖=1|𝑏𝑛𝑖 |
estimation of the transfer function of the transformer. The

In equation (9), m and n are the polynomial degrees of the

numerator and denominator of the fraction of the
transformation function, respectively. Also, (afi and bfi) and
(ani and bni) are, respectively, the ith polynomial coefficient
of the numerator and denominator, mentioned in equation
(8), in the faulted condition and the normal condition. In this
reference, it has been shown that by using this indicator, the FIGURE 19. Block diagram of proposed method in [88].
type, severity and location of AD, RD, DSV and SC faults
can be detected. On the other hand, by comparing this Reference [89] uses a support vector machine (SVM) in
indicator with the SDP indicator, which is used in [83] to order to distinguish AD, RD, DSV and SC faults. In order to
determine the fault severity, it has been concluded that the improve the classification performance, two different sets of
SDP indicator can hardly identify the fault severity, while the indicators are considered as input to the SVM classifier. In
FI index, in addition to estimating the fault severity, has the the first set, there are index of frequency ratio (IFR) and
ability to detect the fault type and location. index of amplitude ratio (IAR), which are extracted from the
Reference [85] also calculated the severity of the winding frequency response of the transformer in healthy and faulty
deformation fault by using the estimation of the transfer states. Also, in the second set, there are VF-based indicators
function and drawing the Nyquist diagram. In this reference, which are extracted from the coefficients of the estimated
after drawing the Nyquist diagram of the transformer transformation function of the transformer in healthy and
frequency response, in two healthy and defective states, a faulty states. The results show that the accuracy in the
special feature is extracted from it, based on which the fault classification of the mentioned faults is higher when the
severity can be identified, and this approach is also features of the first set are applied as input to the SVM. The
confirmed in [84]. The evaluation of this method has also study also included a comparison between the SVM
been done on two real transformers and the results show the classifier and ANN, and the findings demonstrate that SVM
applicability of this approach for practical use. In [86], with provides more precise fault classification for the mentioned
the same approach as those of presented in [85], the location faults.
of the inter-disk fault has been detected. The results show Reference [90] deals with identifying AD, RD, DSV and SC
that this method has succeeded in identifying the fault faults. In order to compare the transfer function of the
location. Reference [27] also deals with the detection of transformer in healthy and faulty conditions, a new method
short-circuit fault severity for the transformer inter-turn fault for calculating indicators is proposed, which is called
using transfer function estimation. In this reference, the input windowed calculated-index (W-index). The basis of this
impedance is well estimated using the VF method. Then, the method is that a moving window with a certain length scans
location of the main pole of the estimated transfer function, the entire frequency range of the FRA test, and the index
which corresponds to the maximum value of the frequency value (here CC) is calculated in each of these windows. The
response, has been used as an indicator to identify the fault value of W-index (here W-CC) in each of these windows is
severity. Reference [87] also uses VF and based on shift of used as a feature for fault classification. Therefore, the W-
the poles and the zeroes of the frequency response, detects index creates a large number of features per observation. Due
the fault severity in which the principle component analysis to the considerable number of observations and features
(PCA) has been used for this purpose. corresponding to them, the dimensions of the classification
problem increase. Using the Linear/Fisher discriminant
A.4 FRA based on artificial intelligence analysis (LDA/FDA) technique, the dimensions of the
Artificial intelligence (AI)-based diagnostic methods offer problem are reduced and applied to an SVM for
researchers another way to achieve the main purpose of classification. The results show that the identification of the
performing FRA. Reference [88] deals with the fault type, severity and location of occurrence is done with
identification of AD, RD and disc-to-disk short circuit faults. 100% accuracy. Reference [91] also uses fuzzy logic
The implementation process is explained as follows. First, algorithm (FLA) to detect the type of fault and its severity.
the transformer model is extracted considering its The results show that this algorithm is able to distinguish
specifications and geometric dimensions. Then, the different types of faults and their severity from the normal
transformer frequency response is obtained in healthy and state of the transformer, regardless of the transformer
defected cases. In the next step, based on the frequency configuration. On the other hand, the low sensitivity of this
response obtained from the previous step, cross-correlation method to noise makes it a reliable method for this purpose.
(CC), ED, SSE, and SSRE features are calculated. These Reference [92], considering RD, DSV and SC faults for the
features are used for artificial neural network (ANN) transformer winding, uses SVM to identify the fault type and
training. The results show that this approach has the ability severity. Correlation coefficient, ED, maximum of
to distinguish the mentioned faults from each other with high difference (MAX), integral of absolute difference (IA), SSE,
accuracy, and also, the fault location and extent are well SSRE, SSMMRE and RMSE are applied as input for SVM
identified. Block diagram of the mentioned approach is training, and its parameters are adjusted by PSO algorithm.
shown in FIGURE 19. The obtained results are as follows:

o when the classifier is used as a separator of the geometric dimensions features and eleven combined features
mentioned three types of faults, the fault identification of invariant moment and texture analysis are extracted from
accuracy is 96.3%. the polar plot image. Then, based on the features extracted
o when SVM is used to classify different degrees of DSV from the polar plot image, three indicators city-block
fault, its accuracy is 95.24%. distance (CBD), root mean square (RMS) and image
o when SVM is used to discriminate different degrees of Euclidean distance (IED) are calculated. Finally, based on
RD fault, its accuracy is 70%. these three indicators, the fault type, severity, and location
are identified. The implementation steps of the presented
Reference [93] uses machine learning and numerical analysis
method are shown in FIGURE 20. It is worth noting that the
techniques to train a predictive engine for smart
mentioned approach has the ability to identify minor faults,
interpretation of FRA data in order to analyze fault severity.
while other common methods have a great weakness in
The proposed technique, which can be utilized for offline or
identifying such faults. Reference [102] also used the same
online application of FRA, is effective for SC inter-turn or
approach presented in [101] to detect the severity and
inter-disk fault detection and fault severity classification in
location of short circuit faults and the results show that this
transformer winding. Reference [94] proposes a two-
approach is capable of identifying minor short circuit fault
dimensional Hilbert ID that considers the extraction of a
levels. The difference is that instead of AD and DSV faults,
multi-window feature for deep vision fault positioning of the
reference [102] deals with inter-turns fault.
transformer winding. In this regard, sweep frequency
response data containing complex fault characteristics are
obtained. Also, different intelligent locating methods are
compared and the proposed deep vision fault location
method is on average 6.51% higher than other methods. In
[95], recognized intelligent classifiers, namely probabilistic
neural network (PNN), decision tree, SVM, and k-nearest
neighbors, are utilized to classify transformer faults.
Similarly, [96] has utilized SVM, radial basis function (RBF)
neural network and statistical k-nearest neighbor method for
fault classification with different strategies and
configurations. Reference [97] divides frequency responses
into four frequency ranges based on frequency resonances
and anti-resonances, extracts the features of the four
frequency ranges by Lin’s Concordance Coefficient (LCC)
indicator, trains the Multilayer Perceptron (MLP) neural
network by the extracted features, and finally, identify and
differentiate the types of winding faults.
In reference [98], the machine learning application called
generalized regression neural network (GRNN) was used to FIGURE 20. Flow chart of the presented DIP technique in [101].
identify the fault and the results show that the introduced
novel index which is called fitting percentage (FP) leads to By introducing the principles of optical FRA and noise
100% accuracy of GRNN in RD fault detection. In reference removal and implementing this method, reference [103]
[99], a new methodology based on the combination of SVM takes advantage of its several unique benefits. These benefits
and synthetic minority oversampling technique (SMOTE) include high immunity against electromagnetic interference,
data preprocessing algorithm is used for fault detection, and direct display of winding deformation, no effect of connected
the results indicate the high accuracy of this method. electrical equipment, and high reliability in harsh
Reference [100] also identified the location and severity of environments. So, the winding deformation in a power
the disk space variation (DSV) fault using the self- transformer is detected.
organizing map (SOM) neural network and shows that this
method, like the reference [98], was able to achieve the A.6 FRA based on discrete wavelet transform
mentioned goal with a 100% accuracy. In order to reduce the effect of noise and measurement
disturbances, references [6] and [104] present a new
A.5 FRA based on digital image processing approach for fault detection, based on discrete wavelet
Reference [101] uses the digital image processing (DIP) transform (DWT). The process of this advanced
technique to identify AD and DSV faults using the image of methodology is presented as follows:
the polar plot of the frequency response of the transformer in o In the first stage, by applying the DWT, the frequency
two healthy and faulty states. First, pre-processing steps are response of the transformer in two healthy and faulty
performed with the aim of adjusting the size and color of the states is decomposed up to seven levels. In this case,
image, as well as image segmentation with the aim of seven smoothed versions are extracted.
removing unwanted effects including noise. Then, four o Then, starting from the seventh level and reaching the
20

first level, a comparison is made between the smoothed for this purpose. It seems that using the digital image
version of present and reference responses. If the processing (DIP) method to evaluate and interpret the FRA
difference between the two is greater than a threshold at standard can be a suitable alternative to the two evaluation
each level, an abnormal difference is detected, and the methods adopted in [32] and [33] for the interpretation of the
corresponding frequency band is stored as an abnormal FRA standard. Because the DIP method is able to identify
band. It is worth noting that at lower resolution levels, minor mechanical faults as well [101], [102].
this frequency band is not scanned again.
o In the next step, the abnormal bands that have been B.2 Characteristic impedance
detected at each level are also examined using indicators The FRA method is very sensitive to the measuring setup at
such as CC and min-max index (MM) to determine high frequencies. Also, due to the effectiveness of this
whether this abnormality is related to failure or not. This method from external factors, reference [105] provides
work is done with the aim of increasing the robustness another method to evaluate the condition of the transformer.
of the presented approach. In this method, considering the transformer winding as a long
transmission line, its characteristic impedance (Zc) is
o Finally, based on the results of the previous step, the obtained as the signature of the transformer. The
fault type is identified. characteristic impedance of the transformer winding is
The results indicate that this approach is highly reliable in unique and independent of external factors. Any change in
detecting faults, as it effectively removes the effects of noise the geometry of the transformer winding due to mechanical
and disturbances. faults leads to a change in the Zc curve at different
frequencies. By comparing the Zc curve of the transformer
B. FRA supplementary methods winding in two healthy and faulty states, it is possible to
Undoubtedly, no fault diagnosis method for transformers can determine the fault type, severity and location. For this
claim to be perfect and flawless. Every method has its own purpose, a sinusoidal signal from low frequencies to high
limitations and weaknesses that can affect its effectiveness frequencies is applied to the input terminal of the transformer
in detecting faults accurately. To address these and by measuring the voltage and current in its other
shortcomings, supplementary methods have been proposed terminal, the characteristic impedance is obtained as follows
and developed over time to improve the accuracy and [105]:
reliability of transformer fault diagnosis. Some of the main 2 (𝜔)
𝑉𝑖𝑛 2 (𝜔)
− 𝑉𝑜𝑢𝑡
supplementary methods are explained below. 𝑍𝑐 = √| 2 2 (𝜔) |
(10)
𝐼𝑖𝑛 (𝜔) − 𝐼𝑜𝑢𝑡
B.1 Impedance method The wavelet coherence indicator is also used to compare two
Research shows that the standard FRA assessment cannot Zc curves (before and after the occurrence of the fault). In
always be sufficient to identify faults. Therefore, the this reference, two types of faults, AD and RD, are
reference [32] states that the impedance method (IM), which investigated and the results show that the characteristic
is based on the low and medium frequency model of the impedance has a higher sensitivity to the occurrence of small
transformer, supports the standard FRA assessment. This mechanical faults compared to FRA.
reference has investigated the sensitivity of standard FRA
and IM for the occurrence of AD and RD mechanical faults. B.3 Vibroacoustic method
For AD fault, the condition of transformer is reported as In this method, the accelerometer sensor is installed on
normal by EEOC and EESC configurations assessment different parts of the transformer tank and its vibrations are
based on correlation coefficients, while in IM, significant recorded. Reports indicate that, if the shape of the coils does
change of leakage inductance and slight change of capacitor not change while only loosely closed, the FRA method
between high and low voltages coil indicate the occurrence cannot detect the fault; but can be detected by the vibration-
of fault in transformer. Also, for RD fault, no fault is reported based method (VM) [106], [107].
by the assessment of EEOC, EESC, CIW and IIW
configurations, while the decrease in the value of earth C. Effect of several important factors on the
capacitor in IM indicates the occurrence of RD fault. Finally, transformer frequency response
it has been concluded that the combination of standard FRA In the following, the effect of some factors such as winding
and IM is the best solution for detecting mechanical faults of type, transformer structure and some other factors affecting
the transformer winding. the frequency response are explained.
Also in [33], by comparing the changes of the parameters of
the transformer ladder model extracted from the FRA C.1 Effect of winding type, structure and residual flux
standard test and the parameters of the IM model in low and of transformer
medium frequencies, it has been concluded that, for the AD In [108], the impact of different winding structures on end-
and RD minor faults, the parameters of the ladder model do to-end FRA responses has been investigated. In general, the
not have enough sensitivity to identify mechanical faults. ratio of series capacitance (Cs) and shunt capacitance (Cg) is
Meanwhile, IM model parameters can be sensitive indicators very important in determining the FRA response for a
21

specific winding structure because this ratio will determine transformer winding. Therefore, the position of the tap
the shape and position of resonance and anti-resonance changer should remain constant during the FRA test.
frequencies. The results of this research are presented as
follows:
o In the frequency response of windings with low Cs
value, the stable magnitude trend with a series of well-
defined resonances and quasi-resonances is observed.
Continuous disc winding is in this category of windings
(FIGURE 21-a).
o In the frequency response of windings with high C s
value, the increasing trend of magnitude with little
resonances and anti-resonances is seen. Inter-leaved
disc-type winding is an example of a winding with high
(a)
Cs (FIGURE 21-b).
o The frequency response of the inter-shielded disc
winding starts with a series of "U"-shaped resonances
and quasi-resonances. Then the response becomes
"camel hump"-shaped with damped resonances and
anti-resonances (FIGURE 21-c).
Reference [109] with a more accurate modeling of the
transformer winding compared to reference [108] and with a
mathematical approach based on the theory of traveling
waves, examines the frequency response of two different
types of windings, i.e. continuous disc winding and (b)
interleaved disc winding, in intermediate frequencies. The
results show that at intermediate frequencies, the frequency
response of continuous disc winding experiences more
oscillations than those of interleaved disc winding. This
result is also confirmed in [108]. In addition to the type of
transformer phase winding, the interactions between the
winding under test and other windings (other windings of the
same phase and other phases) in the D-connection state due
to magnetic coupling and capacitive effect between them
affects the frequency response (resonance and anti-
resonance frequencies) [110].
Reference [111] investigates the effect of magnetization (c)

condition of the transformer core on FRA measurement. In FIGURE 21. Frequency response of different windings:
(a) Continuous disc winding. (b) Interleaved disc winding. (c) Inter-
order to avoid the effects of core magnetization, the FRA test shielded disc winding [108].
should not be performed immediately after disconnecting the
transformer from the power grid. Therefore, the FRA test is
recommended to be performed at least 24 hours after the C.3 Effect of transformer bushing modeling
transformer is de-energized. In order to simultaneously In [114], the effect of transformer bushing modeling on its
remove and measure the residual flux of a three-phase FRA signature has been investigated. In this reference, the
transformer core, reference [112] suggests an improved flux- transformer bushing is added as a T-model to the transformer
controlled variable frequency-constant voltage (VFCV) ladder model. Considering the bushing model, the resonance
strategy along with the core flux equalization using a low frequency of the transformer transfer function at low and
power electronic device. medium frequencies do not change. But due to the fact that
bushing increases the capacitive components of the
C.2 Effect of transformer tap changer transformer, the bushing modeling at high frequencies leads
Reference [113] investigates the effect of transformer tap to the change of resonance frequencies and their
changer setting on FRA. The FRA results for open circuit, corresponding amplitudes. Therefore, when modeling the
short circuit and inductive inter-winding configurations transformer for FRA, the bushing model should also be
clearly show that the tap changer setting influences the considered in order to increase the simulation accuracy.
frequency response of the transformer in low and medium Reference [115] also investigates the effect of bushing fault
frequencies. These changes in the frequency response at low on FRA signature and in this way determines the minimum
frequencies occur due to changes in the inductance of the failure level that can be revealed visually from FRA
response. It also shows that the impact of the bushing fault
22

appears at high frequencies, which is also confirmed by D. Selection of the appropriate frequency range for the
[114]. In [23], transformer bushing is added as T-model to FRA and automatically determining its sub-ranges
the transformer hybrid model and by using CC and spectrum Although it is generally possible to divide the frequency
deviation (σ) indices to interpret the results, it is shown that range into three sub-ranges of low frequency (LF), medium
considering the bushing model causes the frequency frequency (MF), and high frequency (HF), it is not clear what
response to change only at high frequencies. The results of range of frequencies each of these sub-ranges encompasses.
references [114] and [115] also confirm this. This depends on the structure of the transformer and its
configuration, the type of winding and its voltage level, and
C.4 Effect of temperature and moisture therefore, a general rule cannot be considered for it [42]. On
References [116] and [117] investigate the effect of the other hand, a proper definition of these sub-ranges can
temperature and moisture content changes on FRA. Changes improve the capability of the indicators presented earlier. In
in temperature and moisture lead to changes in the this section, the following topics are discussed:
transformer frequency response in medium and high o Selection of the appropriate frequency range for FRA
frequencies. So that with increasing temperature, resonance testing and the factors affecting the limitation of this
frequencies in FRA spectra shift towards lower frequencies range.
and some resonant magnitudes are damped, just like the o The mentioned frequency spectrum is divided into
moisture migration from the paper into the oil insulation several different sub-ranges and what frequency range
when the temperature remained unchanged. Also, with each of these sub-ranges includes.
increasing temperature, the increasing trend is evident in the o Which of these sub-ranges are affected by each of the
total capacitance variation of transformer windings. In these electrical and mechanical faults and the consequent
references, the CC and standard deviation (SD) indicators are change in the parameters of the equivalent circuit?
used to interpret FRA for changes in temperature and In general, the frequency range can be considered as
moisture content, and the results show that these indicators FIGURE 22 [123]. Based on this figure, the following can be
do not have sufficient accuracy in interpreting FRA, and deduced:
therefore they should be modified in some way. Also, in 1) The LF range itself is divided into two smaller ranges
[118] and [119], changes in FRA signature due to LF1 and LF2. In the LF1 range, only the transformer
temperature changes of transformer conductors and moisture core magnetic inductance (Lm) will have a dominant
content of its insulation have been investigated. The results effect on the frequency response as can be observed
show that, regardless of the type of transformer winding, from its nearly linear behavior. Since no capacitor is
with changes in temperature and moisture, the FRA signature effective in this range, the oscillations in the frequency
changes have the same pattern. So, with the increase in response occur solely due to interactions mutually with
temperature and moisture, the transformer frequency Lm inductance. However, in the LF2 range, due to the
response shifts in the medium and high frequencies towards interactions between the Lm inductance and the Cg,w
the lower frequencies. These results are also confirmed by parallel capacitors, the combination of these two
[116] and [117]. Reference [120] also compares the effect of elements will have a dominant effect in this range.
moisture changes in transformer paper insulation and also 2) In the MF range, the penetration depth of the magnetic
winding deformation on the transformer frequency response. field inside the transformer core is reduced. So, the core
Finally, it has been concluded that the effect of moisture behaves like an aerial core with a specific inductance
migration can be seen in the medium and high frequency (L). Due to its interactions with series capacitors (Cs)
bands, but in connection with mechanical defects, this effect and parallel capacitors (Cg,w) as well as the effect of
can be seen in each or every part of the entire frequency mutual coupling between the windings (LM) and the
range. A new moisture sensor and measurement method is capacitor between the high-voltage and low-voltage
presented in [121], which is based on the introduction of windings, large fluctuations occur in this range.
considering capacitive effects inside the transformer tank 3) The HF range is also divided into two smaller ranges,
and the analysis of its frequency response with the FRA test. HF1 and HF2. In the HF1 range, series capacitors (C s)
The sensor should be immersed in transformer oil, away will have the predominant effect. In the HF2 range,
from high voltage areas. Since the moisture of the oil which is not shown in FIGURE 22 and includes
depends on the temperature, the temperature of the oil must frequencies above 1 MHz, the measurement system also
also be measured. Also, reference [122] evaluates the affects the frequency response.
capability and accuracy of its presented method in
Reference [123] presents an automatic method for
determining the moisture content of the transformer by
determining the mentioned frequency ranges. The principle
obtaining the results of FRA, extracting the required
of this method is based on the resonance frequencies
features, using the k-means method for placing these features
corresponding to passing through the zero phase of the
in three clusters (dry, wet, and excessively wet), and
frequency response. In other words, based on the critical
optimizing the cost function of the k-means method using the
points of passing through the zero phase of the frequency
PSO algorithm.
response, the proposed method determines the frequency
ranges of each sub-range.
23

ranges represented by F1, F2, F3, and F4. The third

classification level covers general electrical and mechanical
faults, denoted by EM and MM, respectively. However,
there are some faults affecting all frequency ranges and their
interpretation can be based on the behavior of the resonance
peak.
References [124] and [125] examine the effect of different
faults on frequency ranges and present that RD fault affects
all frequency ranges, while AD fault affects intermediate and
higher frequencies. The effect of SC fault and fault between
disks is also evident at low and medium frequencies,
respectively. Reference [126] examines the RD fault and
models this fault by changing the capacitor between the high
voltage and low voltage windings. The results indicate that
this type of fault affects the entire frequency range. Actually,
the results of [124] and [125] are approved by another
method. Reference [127] also obtains a similar result for the
RD fault.
Some researchers believe that only the mutual inductance
FIGURE 22. Frequency range defined by the expert [123]. between the high-voltage and low-voltage windings changes
when an AD fault occurs. Therefore, they change this
As previously mentioned, the structure of the transformer, parameter in the modeling, so that they can model the AD
the type of winding or its configuration for the FRA test may fault and then, compare the output with the frequency
change. Because the number of resonance frequencies and response of the transformer in the healthy state and evaluate
their location change. In this regard, this method can adapt different methods to determine its severity.
itself to the new conditions and specify frequency ranges. The authors indicate in [128] that when an AD fault occurs
This means that the method is adaptive. FIGURE 23 shows on a high voltage coil, not only the mutual inductance but
the effect of various electrical and mechanical faults on each also the capacitor between the high voltage coil and the
of the frequency ranges. The first classification level transformer body, as well as the capacitor between the high
considers some faults that affect only one frequency range voltage coil and the low voltage coil changes and these
and change the frequency response in that range, denoted by changes are evident. Therefore, in AD fault modeling, this
R1, R2, R3, R4, and R5, and corresponding to the LF1, LF2, point should be considered so that different methods of fault
MF, HF1, and HF2, respectively. The second classification severity detection are evaluated and compared correctly. The
level includes some other faults that affect two adjacent authors of references [129]–[131] also believe that the AD

FIGURE 23. Classification of fault types inside the transformer [123].

fault affects the intermediate frequencies. This type of fault, detection of incipient faults are possible. Reference [139]
in addition to changing the inductance matrix of the explains several online transformer FRA methods, for
equivalent circuit, changes the capacitive matrix that detecting several faults, specially the inter-turn fault. These
changing of these two parameters should be considered to methods include (DGA) [140], current harmonic analysis
model this type of fault. In [132], it is shown with the same [141], vibration analysis [142], the differential current based
process, for modeling the RD fault correctly, two changes on negative sequence and Extended Park Vector [143],
must be considered; changing the values of the equivalent [144], ΔV-I locus diagram [145], and flux leakage method
circuit capacitances, and changing the circuit inductance [146]. These methods have been usually evaluated
matrix. The effect of this fault is evident in the whole qualitatively with a focus on feasibility. In this regard,
frequency range. In [23], using the hybrid transformer model reference [147] by leveraging both experiment and
and optimizing its parameters, the effect of different theoretical analysis, tries to completely evaluate several
transformer faults on the frequency response is evaluated and fundamental aspects of online FRA methods, including
the obtained results indicate that AD and RD faults have their sensitivity, repeatability, and anti-interference.
main effect on the medium frequencies (approximately 1 The IFRA method itself can be done in two ways: the
MHz). controlled and the uncontrolled signal application methods,
while the SFRA method is performed only in a controlled
E. Online FRA method manner. In the controlled signal application method, the
The FRA test requires the transformer to be de-energized and pulse signal is generated by a pulse wave generator. The
disconnected, and laboratory instrumentation must be used advantage of the controlled method is that firstly, the signal
to determine the transfer function. As a result, this may incur application time and secondly, the signal shape and its
high costs and reduce network reliability. This means that presence time (which will determine its frequency content)
this test cannot be carried out repeatedly to implement will be controlled and the aforementioned difficulties will be
precautionary actions. It is only performed when the slightly reduced. In the uncontrolled signal application
transformer is suspected of being faulty, and corrective method, lightning or switching signals that occur in the
actions are taken if necessary. In order to solve this problem, substation are used. The advantage of this method is that
it prompted the researchers to provide methods that detect there is no need for a pulse signal generator. But the
the internal defects of the transformer without disconnecting drawback is that firstly these signals are not always available,
it from the power grid and determine the severity of the fault and one has to wait for a certain moment for one of the two
and the location inside the transformer online. voltage waves to occur and secondly, the shape and duration
As mentioned earlier, FRA testing (including its online of presence of such uncontrolled signals are not adjustable.
method) can be performed in two ways, SFRA and IFRA. Reference [148] first deals with the necessary tools to
But according to the reports, the online testing has some implement the online FRA test and then compares it with the
problems [133], [134]. Some of these difficulties are as offline one, and shows that in the online FRA due to the
follow: presence of noise, the low frequency information of the
o The moment of applying the FRA signal is very frequency response cannot be trusted. This makes it
important due to its effect on the obtained frequency challenging to detect the occurrence of faults that affect this
response, frequency range, such as a short-circuit fault.
o The possibility of interference on network controller
signals by reference signal (with frequencies from V. CONCLUSION AND RECOMMENDATIONS FOR
several Hz to several MHz), FUTURE IMPROVEMENTS
o The possibility of damage to the analyzer device in case In this paper, while expressing the need for the importance
of directly connecting it to the transformer, of transformers in power grids and detecting the occurrence
o The effect of transformer load on online FRA fault of various faults in it, the frequency behavior of the
detection capability, etc. transformer is explained, and various models are presented
to simulate its actual behavior against different frequencies.
Also, several advantages for an online FRA test are worth- Moreover, the requirements of FRA for implementation are
mentioning, such as [135]: explained and it is shown that the four configurations named
o Lower cost of testing, EE, EESC, CIW and IIW together with the transfer voltage,
o No requirement for disconnecting the transformer from form the best and most sensitive configuration-transfer
the grid. function pair for FRA.
Some researchers have focused on online FRA. For example, In addition to that, various methods for FRA are examined.
references [136] and [137] deal with the effect of transformer After analyzing the FRA results using different methods,
load on online FRA fault detection capability. In [138], by complementary FRA methods are introduced and it is shown
utilizing and comparing two methods, IFRA and lightning that three different methods, namely the impedance, the
impulse analysis (LIA), detection of winding deformation characteristic impedance, and the vibroacoustic methods,
and short circuit faults on live transformers is studied. can eliminate the weaknesses of the FRA test in order to
Results show that the diagnosis of transformer windings and identify the faults.

The effect of several factors such as winding type, structure, increasing the ability to detect faults in this method.
residual flux, temperature and moisture on the frequency Therefore, for a two-winding transformer, many
response of the transformer is also investigated and it is terminal connections were considered and for each
shown that some of these factors can significantly change the terminal connection, different transfer functions were
frequency response. After examining the effect of these examined. Then, based on the number of natural
factors on the frequency response, it is pointed out that due frequencies appearing in the transfer function, different
to the effect of these factors, there is a need for an adaptive configuration-transfer function pairs were categorized
and flexible frequency range for FRA. Therefore, using an (TABLE 1). Finally, the configuration-transfer function
adaptive method, the frequency range is divided into pair that has the highest natural frequency and, of
different sub-ranges in order to increase the sensitivity of course, the highest sensitivity for fault detection, was
statistical indicators to analyze the FRA results. selected for FRA test. Since a large 3-phase transformer
Furthermore, different indicators have been examined from as well as multi-winding transformers have more
the perspective of sensitivity, capability and reliability. Also, terminals than two-winding transformers, it is possible
it is shown that each of the electrical and mechanical faults to define far more terminal connections than a two-
firstly, on which of the specified sub-ranges, have a greater winding transformer to perform the FRA test. Therefore,
effect, and secondly, each of these faults leads to a change in the process mentioned in section III can be implemented
which of the equivalent circuit elements. Finally, the online in order to find the suitable configuration-transfer
FRA method is discussed, and its challenges are mentioned. function pair to perform the FRA test on this class of
The following topics are suggested in order to continue transformers.
further research in FRA:
REFERENCES
o In section IV.A.1, it was stated that some indicators do [1] E. Gockenbach and H. Borsi, “Condition monitoring and
not show a regular and linear behavior with respect to diagnosis of power transformers,” in 2008 International
the changes in fault severity, and therefore cannot be Conference on Condition Monitoring and Diagnosis, 2008, pp.
useful in determining the fault severity value. Since each 894–897.
[2] J. S. N’cho, I. Fofana, Y. Hadjadj, and A. Beroual, “Review of
fault usually does not affect the entire frequency physicochemical-based diagnostic techniques for assessing
response range of the transformer and only affects a insulation condition in aged transformers,” Energies, vol. 9, no. 5,
specific interval of the entire frequency range (which p. 367, 2016.
was stated in section IV.D). Therefore, some papers in [3] M. Badawi et al., “Reliable estimation for health index of
transformer oil based on novel combined predictive maintenance
order to solve this problem recommend the approach of techniques,” IEEE Access, vol. 10, pp. 25954–25972, 2022.
dividing the entire frequency range into several smaller [4] M. M. F. Darwish, M. H. A. Hassan, N. M. K. Abdel-Gawad, and
intervals and calculating each of the indicators presented D.-E. A. Mansour, “A new method for estimating transformer
in section IV.A.1 in that frequency interval where the health index based on ultraviolet-visible spectroscopy,” in 2022
23rd International Middle East Power Systems Conference
specific fault is effective. Because it has been shown that (MEPCON), 2022, pp. 1–5.
the values of the indicators in the interval of the [5] M. M. F. Darwish, M. H. A. Hassan, N. M. K. Abdel-Gawad, and
frequency range where the specific fault is effective, D.-E. A. Mansour, “Application of infrared spectroscopy for
usually show a linear relationship with respect to the discrimination between electrical and thermal faults in
transformer oil,” in 2022 9th International Conference on
changes in the fault severity. The problem that exists is Condition Monitoring and Diagnosis (CMD), 2022, pp. 255–258.
that even with this technique, some indicators still do not [6] J. C. G. Arispe and E. E. Mombello, “Detection of failures within
experience a linear relationship with the changes in the transformers by FRA using multiresolution decomposition,”
fault severity. Therefore, to solve this problem, it is IEEE Trans. Power Deliv., vol. 29, no. 3, pp. 1127–1137, 2014.
[7] M. H. Samimi, S. Tenbohlen, A. A. S. Akmal, and H. Mohseni,
suggested to draw the difference function (ΔTFi) graph “Effect of different connection schemes, terminating resistors and
presented in equation (7) in that frequency range where measurement impedances on the sensitivity of the FRA method,”
the specific fault is effective. Then, in that frequency IEEE Trans. Power Deliv., vol. 32, no. 4, pp. 1713–1720, 2016.
range, a smaller interval should be found so that the [8] M. H. Samimi, S. Tenbohlen, A. A. S. Akmal, and H. Mohseni,
“Improving the numerical indices proposed for the FRA
values of ΔTFi in that smaller interval have a significant interpretation by including the phase response,” Int. J. Electr.
difference with each other, and also the curve-fitting of Power Energy Syst., vol. 83, pp. 585–593, 2016.
the ΔTF diagram according to the severity of different [9] S. D. Mitchell and J. S. Welsh, “The influence of complex
faults, experiences a linear relationship. Such small permeability on the broadband frequency response of a power
transformer,” IEEE Trans. Power Deliv., vol. 25, no. 2, pp. 803–
bands are called diagnostic bands, which was mentioned 813, 2009.
in section IV.A.1. Therefore, the calculation of the [10] K. G. N. B. Abeywickrama, Y. V Serdyuk, and S. M. Gubanski,
indicators in the diagnostic band related to the frequency “Exploring possibilities for characterization of power transformer
range where the mentioned fault is effective, can have a insulation by frequency response analysis (FRA),” IEEE Trans.
Power Deliv., vol. 21, no. 3, pp. 1375–1382, 2006.
significant effect on the linearity and monotonicity of [11] A. Shintemirov, W. H. Tang, and Q. H. Wu, “A hybrid winding
the mentioned indicator. model of disc-type power transformers for frequency response
o In section III.A, it was shown that choosing the analysis,” IEEE Trans. Power Deliv., vol. 24, no. 2, pp. 730–739,
appropriate configuration and transfer function to 2009.
[12] S. Almas, T. Wenhu, and Q. H. Wu, “Modeling of a power
perform the FRA test has a significant effect on transformer winding for deformation detection based on

frequency response analysis,” in 2007 Chinese Control Dielectr. Electr. Insul., vol. 20, no. 6, pp. 2026–2034, 2013.
Conference, 2007, pp. 506–510. [31] D. A. K. Pham, T. M. T. Pham, M. H. Safari, V. N. C. Ho, H.
[13] W. H. Tang, A. Shintemirov, and Q. H. Wu, “Detection of minor Borsi, and E. Gockenbach, “FRA-based transformer parameters
winding deformation fault in high frequency range for power at low frequencies,” in 2012 International Conference on High
transformer,” in IEEE PES General Meeting, 2010, pp. 1–6. Voltage Engineering and Application, 2012, pp. 476–479.
[14] A. Shintemirov and Q. H. Wu, “Transfer function of transformer [32] D. A. K. Pham, T. M. T. Pham, H. Borsi, and E. Gockenbach, “A
winding for frequency response analysis based on traveling wave new diagnostic method to support standard frequency response
theory,” in Proc. Int. Control Conf. (ICC 2006), 2006, p. 6. analysis assessments for diagnostics of transformer winding
[15] S. D. Mitchell and J. S. Welsh, “Modeling power transformers to mechanical failures,” IEEE Electr. Insul. Mag., vol. 30, no. 2, pp.
support the interpretation of frequency-response analysis,” IEEE 34–41, 2014.
Trans. Power Deliv., vol. 26, no. 4, pp. 2705–2717, 2011. [33] D. A. K. Pham and E. Gockenbach, “Analysis of physical
[16] E. Rahimpour, J. Christian, K. Feser, and H. Mohseni, “Transfer transformer circuits for frequency response interpretation and
function method to diagnose axial displacement and radial mechanical failure diagnosis,” IEEE Trans. Dielectr. Electr.
deformation of transformer windings,” IEEE Trans. Power Deliv., Insul., vol. 23, no. 3, pp. 1491–1499, 2016.
vol. 18, no. 2, pp. 493–505, 2003. [34] B. A. Mork, F. Gonzalez, D. Ishchenko, D. L. Stuehm, and J.
[17] A. Shintemirov, W. J. Tang, W. H. Tang, and Q. H. Wu, Mitra, “Hybrid transformer model for transient simulation—Part
“Improved modelling of power transformer winding using I: Development and parameters,” IEEE Trans. Power Deliv., vol.
bacterial swarming algorithm and frequency response analysis,” 22, no. 1, pp. 248–255, 2006.
Electr. Power Syst. Res., vol. 80, no. 9, pp. 1111–1120, 2010. [35] B. A. Mork, F. Gonzalez, D. Ishchenko, D. L. Stuehm, and J.
[18] V. Rashtchi, E. Rahimpour, and E. M. Rezapour, “Using a genetic Mitra, “Hybrid transformer model for transient simulation—Part
algorithm for parameter identification of transformer RLCM II: Laboratory measurements and benchmarking,” IEEE Trans.
model,” Electr. Eng., vol. 88, no. 5, pp. 417–422, 2006. Power Deliv., vol. 22, no. 1, pp. 256–262, 2006.
[19] P. Mukherjee and L. Satish, “Construction of equivalent circuit of [36] S. P. Ang, J. Li, Z. Wang, and P. Jarman, “FRA low frequency
a single and isolated transformer winding from FRA data using characteristic study using duality transformer core modeling,” in
the ABC algorithm,” IEEE Trans. Power Deliv., vol. 27, no. 2, 2008 International Conference on Condition Monitoring and
pp. 963–970, 2012. Diagnosis, 2008, pp. 889–893.
[20] V. Rashtchi, H. Shayeghi, M. Mahdavi, A. Kimiyaghalam, and E. [37] W. Herrera, G. Aponte, J. Pleite, and C. Gonzalez-Garcia, “A
Rahimpour, “Using an improved PSO algorithm for parameter novel methodology for transformer low-frequency model
identification of transformer detailed model,” Int. J. Electr. Power parameters identification,” Int. J. Electr. Power Energy Syst., vol.
Energy Syst., vol. 1, no. 3, pp. 138–144, 2008. 53, pp. 643–648, 2013.
[21] A. Abu-Siada, M. I. Mosaad, D. Kim, and M. F. El-Naggar, [38] W. H. Portilla, G. A. Mayor, J. P. Guerra, and C. Gonzalez-
“Estimating power transformer high frequency model parameters Garcia, “Detection of transformer faults using frequency-response
using frequency response analysis,” IEEE Trans. Power Deliv., traces in the low-frequency bandwidth,” IEEE Trans. Ind.
vol. 35, no. 3, pp. 1267–1277, 2020, doi: Electron., vol. 61, no. 9, pp. 4971–4978, 2013.
10.1109/TPWRD.2019.2938020. [39] W. Herrera, G. Aponte, C. Gonzalez-Garcia, and J. Pleite, “Novel
[22] M. M. Shabestary, A. J. Ghanizadeh, G. B. Gharehpetian, and M. procedure for low frequency FRA traces interpretation,” in 2012
Agha-Mirsalim, “Ladder network parameters determination XXth International Conference on Electrical Machines, 2012, pp.
considering nondominant resonances of the transformer 2344–2350.
winding,” IEEE Trans. Power Deliv., vol. 29, no. 1, pp. 108–117, [40] L. Satish and A. Saravanakumar, “Identification of terminal
2013. connection and system function for sensitive frequency response
[23] T. Y. Ji, W. H. Tang, and Q. H. Wu, “Detection of power measurement on transformers,” IEEE Trans. Power Deliv., vol.
transformer winding deformation and variation of measurement 23, no. 2, pp. 742–750, 2008.
connections using a hybrid winding model,” Electr. Power Syst. [41] J. Zhijian, Z. Minglin, and Z. Zishu, “Fault location of transformer
Res., vol. 87, pp. 39–46, 2012. winding deformation using frequency response analysis,” in
[24] S. Pramanik and L. Satish, “Estimation of series capacitance of a Proceedings of 2001 International Symposium on Electrical
transformer winding based on frequency-response data: An Insulating Materials (ISEIM 2001). 2001 Asian Conference on
indirect measurement approach,” IEEE Trans. Power Deliv., vol. Electrical Insulating Diagnosis (ACEID 2001). 33rd Symposium
26, no. 4, pp. 2870–2878, 2011. on Electrical and Ele, 2001, pp. 841–844.
[25] K. Ragavan and L. Satish, “Construction of physically realizable [42] J. Jayasinghe, Z. D. Wang, P. N. Jarman, and A. W. Darwin,
driving-point function from measured frequency response data on “Winding movement in power transformers: a comparison of
a model winding,” IEEE Trans. Power Deliv., vol. 23, no. 2, pp. FRA measurement connection methods,” IEEE Trans. Dielectr.
760–767, 2008. Electr. Insul., vol. 13, no. 6, pp. 1342–1349, 2006.
[26] K. Ragavan and L. Satish, “Localization of changes in a model [43] R. K. Senobari, J. Sadeh, and H. Borsi, “Frequency response
winding based on terminal measurements: Experimental study,” analysis (FRA) of transformers as a tool for fault detection and
IEEE Trans. Power Deliv., vol. 22, no. 3, pp. 1557–1565, 2007. location: A review,” Electr. Power Syst. Res., vol. 155, pp. 172–
[27] R. M. Youssouf, F. Meghnefi, and I. Fofana, “Frequency response 183, 2018.
analyses via rational function fitting,” in 2012 Annual Report [44] IEC 60076-18, “Measurement of frequency response,” 61010-1 ©
Conference on Electrical Insulation and Dielectric Phenomena, Iec2001, vol. Ed.1, no. Part 18, 2012.
2012, pp. 230–233. [45] “Mechanical condition assessment of transformer windings using
[28] M. Heindl, S. Tenbohlen, A. Kraetge, M. Krüger, and J. frequency response analysis (FRA),” CIGRE Tech. Broch. no.
Velásquez, “Algorithmic determination of pole-zero 342, 2008.
representations of power transformers’ transfer functions for [46] IEEE, “IEEE guide for the application and interpretation of
interpretation of FRA data,” in Paper D-26, 16th Int. Symp. on frequency response analysis for oil-immersed transformers,”
High Voltage Engineering, Johannesburg, 2009. IEEE Std C57. 149, 2012.
[29] D. A. K. Pham, T. M. T. Pham, V. N. C. Ho, H. Borsi, and E. [47] “Frequency response analysis on winding deformation of power
Gockenbach, “Duality-based lumped transformer equivalent transformers,” Chinese Stand. DL/T911, 2004.
circuit at low frequencies under single-phase excitation,” in 2012 [48] F. (OMICRON) Predl, “Interpretation of sweep frequency
International Conference on High Voltage Engineering and response analysis (SFRA) measurement results,” Omi. Sydney,
Application, 2012, pp. 600–603. Aust., 2016.
[30] D. A. K. Pham, T. M. T. Pham, H. Borsi, and E. Gockenbach, “A [49] S. Tenbohlen and S. A. Ryder, “Making frequency response
new method for purposes of failure diagnostics and FRA analysis measurements: a comparison of the swept frequency and
interpretation applicable to power transformers,” IEEE Trans. low voltage impulse methods,” in 18th Int. Symp. on High Voltage

Engineering, Netherlands, 2003. techniques to FRA data for diagnosing integrity of transformer
[50] A. Kraetge, M. Kruger, and P. Fong, “Frequency response windings,” in 2015 International Conference on Condition
analysis—Status of the worldwide standardization activities,” in Assessment Techniques in Electrical Systems (CATCON), 2015,
2008 International Conference on Condition Monitoring and pp. 87–92.
Diagnosis, 2008, pp. 651–654. [68] P. Karimifard and G. B. Gharehpetian, “A new algorithm for
[51] T. Sano and K. Miyagi, “Influence factors on FRA waveforms,” localization of radial deformation and determination of
in 2012 IEEE International Conference on Condition Monitoring deformation extent in transformer windings,” Electr. Power Syst.
and Diagnosis, 2012, pp. 601–604. Res., vol. 78, no. 10, pp. 1701–1711, 2008.
[52] Y. Liu et al., “A study of the sweep frequency impedance method [69] E. Rahimpour, M. Jabbari, and S. Tenbohlen, “Mathematical
and its application in the detection of internal winding short circuit comparison methods to assess transfer functions of transformers
faults in power transformers,” IEEE Trans. Dielectr. Electr. to detect different types of mechanical faults,” IEEE Trans. Power
Insul., vol. 22, no. 4, pp. 2046–2056, 2015. Deliv., vol. 25, no. 4, pp. 2544–2555, 2010.
[53] J. Christian and K. Feser, “Procedures for detecting winding [70] M. Khanali, A. Hayati-Soloot, H. K. Høidalen, and S. Jayaram,
displacements in power transformers by the transfer function “Study on locating transformer internal faults using sweep
method,” IEEE Trans. Power Deliv., vol. 19, no. 1, pp. 214–220, frequency response analysis,” Electr. Power Syst. Res., vol. 145,
2004. pp. 55–62, 2017.
[54] M. H. Samimi, A. A. S. Akmal, H. Mohseni, and S. Tenbohlen, [71] K. Ludwikowski, K. Siodla, and W. Ziomek, “Investigation of
“Detection of transformer mechanical deformations by comparing transformer model winding deformation using sweep frequency
different FRA connections,” Int. J. Electr. Power Energy Syst., response analysis,” IEEE Trans. Dielectr. Electr. Insul., vol. 19,
vol. 86, pp. 53–60, 2017. no. 6, pp. 1957–1961, 2012.
[55] A. A. Pandya and B. R. Parekh, “Interpretation of Sweep [72] K. Pourhossein, G. B. Gharehpetian, E. Rahimpour, and B. N.
Frequency Response Analysis (SFRA) traces for the open circuit Araabi, “A probabilistic feature to determine type and extent of
and short circuit winding fault damages of the power winding mechanical defects in power transformers,” Electr.
transformer,” Int. J. Electr. Power Energy Syst., vol. 62, pp. 890– Power Syst. Res., vol. 82, no. 1, pp. 1–10, 2012.
896, 2014. [73] M. H. Samimi, S. Tenbohlen, A. A. S. Akmal, and H. Mohseni,
[56] J. R. Secue and E. Mombello, “Sweep frequency response “Using the complex values of the frequency response to improve
analysis (SFRA) for the assessment of winding displacements and power transformer diagnostics,” in 2016 24th Iranian conference
deformation in power transformers,” Electr. Power Syst. Res., vol. on electrical engineering (ICEE), 2016, pp. 1689–1693.
78, no. 6, pp. 1119–1128, 2008, doi: 10.1016/j.epsr.2007.08.005. [74] M. H. Samimi, S. Tenbohlen, A. A. S. Akmal, and H. Mohseni,
[57] L. Satish and S. K. Sahoo, “Locating faults in a transformer “Dismissing uncertainties in the FRA interpretation,” IEEE
winding: An experimental study,” Electr. Power Syst. Res., vol. Trans. Power Deliv., vol. 33, no. 4, pp. 2041–2043, 2016.
79, no. 1, pp. 89–97, 2009. [75] Y. Akhmetov, V. Nurmanova, M. Bagheri, A. Zollanvari, and G.
[58] X. Lei, J. Li, Y. Wang, S. Mi, and C. Xiang, “Simulative and B. Gharehpetian, “A new diagnostic technique for reliable
experimental investigation of transfer function of inter-turn faults decision-making on transformer FRA data in interturn short-
in transformer windings,” Electr. Power Syst. Res., vol. 107, pp. circuit condition,” IEEE Trans. Ind. Informatics, vol. 17, no. 5,
1–8, 2014. pp. 3020–3031, 2021, doi: 10.1109/TII.2020.3007607.
[59] B. Cheng, Z. Wang, and P. Crossley, “Using lumped element [76] V. Nurmanova, Y. Akhmetov, M. Bagheri, A. Zollanvari, B. T.
equivalent network model to derive analytical equations for Phung, and G. B. Gharehpetian, “Confidence level estimation for
interpretation of transformer frequency responses,” IEEE Access, advanced decision-making in transformer short-circuit fault
vol. 8, pp. 179486–179496, 2020, doi: diagnosis,” IEEE Trans. Ind. Appl., vol. 58, no. 1, pp. 233–241,
10.1109/ACCESS.2020.3027798. 2021.
[60] V. Behjat and M. Mahvi, “Statistical approach for interpretation [77] M. Tahir, S. Tenbholen, and S. Miyazaki, “Analysis of statistical
of power transformers frequency response analysis results,” IET methods for assessment of power transformer frequency response
Sci. Meas. Technol., vol. 9, no. 3, pp. 367–375, 2015. measurements,” IEEE Trans. Power Deliv., vol. 36, no. 2, pp.
[61] V. Behjat, A. Vahedi, A. Setayeshmehr, H. Borsi, and E. 618–626, 2021, doi: 10.1109/TPWRD.2020.2987205.
Gockenbach, “Sweep frequency response analysis for diagnosis [78] S. M. A. N. Al-Ameri et al., “Understanding the influence of
of low level short circuit faults on the windings of power power transformer faults on the frequency response signature
transformers: An experimental study,” Int. J. Electr. Power using simulation analysis and statistical indicators,” IEEE Access,
Energy Syst., vol. 42, no. 1, pp. 78–90, 2012. vol. 9, pp. 70935–70947, 2021, doi:
[62] H. Firoozi and M. Shishehchian, “Frequency response analysis- 10.1109/ACCESS.2021.3076984.
condition assessment of power transformers using mathematical [79] Y. Akhmetov, V. Nurmanova, M. Bagheri, A. Zollanvari, and G.
and statistical criteria,” in 2009 IEEE 9th International B. Gharehpetian, “A bootstrapping solution for effective
Conference on the Properties and Applications of Dielectric interpretation of transformer winding frequency response,” IEEE
Materials, 2009, pp. 253–256. Trans. Instrum. Meas., vol. 71, no. 1, pp. 1–11, 2022.
[63] E. Rahimpour and D. Gorzin, “A new method for comparing the [80] B. A. Thango, A. F. Nnachi, G. A. Dlamini, and P. N. Bokoro, “A
transfer function of transformers in order to detect the location and Novel Approach to Assess Power Transformer Winding
amount of winding faults,” Electr. Eng., vol. 88, no. 5, pp. 411– Conditions Using Regression Analysis and Frequency Response
416, 2006. Measurements,” Energies, vol. 15, no. 7, p. 2335, 2022.
[64] J.-W. Kim, B. Park, S. C. Jeong, S. W. Kim, and P. Park, “Fault [81] M. Bigdeli and A. Abu-Siada, “Clustering of transformer
diagnosis of a power transformer using an improved frequency- condition using frequency response analysis based on k-means
response analysis,” IEEE Trans. Power Deliv., vol. 20, no. 1, pp. and GOA,” Electr. Power Syst. Res., vol. 202, p. 107619, 2022.
169–178, 2005. [82] H. Tarimoradi, H. Karami, G. B. Gharehpetian, and S. Tenbohlen,
[65] A. S. A. Razzaq, M. F. M. Yousof, S. M. Al-Ameri, M. A. Talib, “Sensitivity analysis of different components of transfer function
and A. J. Mshkil, “Interpretation of inter-turn fault in transformer for detection and classification of type, location and extent of
winding using turns ratio test and frequency response analysis,” transformer faults,” Measurement, vol. 187, p. 110292, 2022.
Int. J. Electr. Electron. Eng. Telecommun, vol. 11, pp. 218–225, [83] P. Karimifard, G. B. Gharehpetian, and S. Tenbohlen,
2022. “Localization of winding radial deformation and determination of
[66] O. E. Gouda and S. H. El-Hoshy, “Diagnostic technique for deformation extent using vector fitting-based estimated transfer
analysing the internal faults within power transformers based on function,” Eur. Trans. Electr. Power, vol. 19, no. 5, pp. 749–762,
sweep frequency response using adjusted r-square methodology,” 2009.
IET Sci. Meas. Technol., vol. 14, no. 10, pp. 1057–4068, 2020. [84] M. Bigdeli, M. Vakilian, and E. Rahimpour, “A new method for
[67] M. Prameela and P. M. Nirgude, “Application of numerical detection and evaluation of winding mechanical faults in

transformer through transfer function measurements,” Adv. [101] O. Aljohani and A. Abu-Siada, “Application of DIP to detect
Electr. Comput. Eng., vol. 11, no. 2, pp. 23–30, 2011. power transformers axial displacement and disk space variation
[85] M. F. M. Yousof, C. Ekanayake, and T. K. Saha, “Frequency using FRA polar plot signature,” IEEE Trans. Ind. Informatics,
response analysis to investigate deformation of transformer vol. 13, no. 4, pp. 1794–1805, 2016.
winding,” IEEE Trans. Dielectr. Electr. Insul., vol. 22, no. 4, pp. [102] O. Aljohani and A. Abu-Siada, “Application of digital image
2359–2367, 2015. processing to detect short-circuit turns in power transformers
[86] M. F. M. Yousof, C. Ekanayake, and T. K. Saha, “Locating inter- using frequency response analysis,” IEEE Trans. Ind. Informatics,
disc faults in transformer winding using frequency response vol. 12, no. 6, pp. 2062–2073, 2016.
analysis,” in 2013 Australasian Universities Power Engineering [103] G. Ma, Y. Liu, Y. Li, X. Fan, C. Xu, and W. Qin, “Optical
Conference (AUPEC), 2013, pp. 1–6. frequency-response analysis for power transformer,” IEEE Trans.
[87] J. C. Gonzales and E. E. Mombello, “Diagnosis of power Power Deliv., vol. 36, no. 3, pp. 1562–1570, 2021, doi:
transformers through frequency response analysis by poles and 10.1109/TPWRD.2020.3011422.
zeros shifts identification,” in 2012 Sixth IEEE/PES Transmission [104] J. C. Gonzales and E. E. Mombello, “Fault interpretation
and Distribution: Latin America Conference and Exposition algorithm using frequency-response analysis of power
(T&D-LA), 2012, pp. 1–8. transformers,” IEEE Trans. Power Deliv., vol. 31, no. 3, pp.
[88] A. J. Ghanizadeh and G. B. Gharehpetian, “ANN and cross- 1034–1042, 2015.
correlation based features for discrimination between electrical [105] A. J. Ghanizadeh and G. B. Gharehpetian, “Application of
and mechanical defects and their localization in transformer characteristic impedance and wavelet coherence technique to
winding,” IEEE Trans. Dielectr. Electr. Insul., vol. 21, no. 5, pp. discriminate mechanical defects of transformer winding,” Electr.
2374–2382, 2014. Power Components Syst., vol. 41, no. 9, pp. 868–878, 2013.
[89] M. Bigdeli, M. Vakilian, and E. Rahimpour, “Transformer [106] E. Kornatowski and S. Banaszak, “Diagnostics of a transformer’s
winding faults classification based on transfer function analysis active part with complementary FRA and VM measurements,”
by support vector machine,” IET Electr. Power Appl., vol. 6, no. IEEE Trans. Power Deliv., vol. 29, no. 3, pp. 1398–1406, 2013.
5, pp. 268–276, 2012. [107] S. Banaszak and E. Kornatowski, “Evaluation of FRA and VM
[90] H. Tarimoradi and G. B. Gharehpetian, “Novel calculation measurements complementarity in the field conditions,” IEEE
method of indices to improve classification of transformer Trans. Power Deliv., vol. 31, no. 5, pp. 2123–2130, 2015.
winding fault type, location, and extent,” IEEE Trans. Ind. [108] Z. Wang, J. Li, and D. M. Sofian, “Interpretation of transformer
Informatics, vol. 13, no. 4, pp. 1531–1540, 2017. FRA responses—Part I: Influence of winding structure,” IEEE
[91] A. Contin, G. Rabach, J. Borghetto, M. De Nigris, R. Passaglia, Trans. Power Deliv., vol. 24, no. 2, pp. 703–710, 2009.
and G. Rizzi, “Frequency-response analysis of power [109] M. Bagheri, B. T. Phung, and T. Blackburn, “Transformer
transformers by means of fuzzy tools,” IEEE Trans. Dielectr. frequency response analysis: mathematical and practical approach
Electr. Insul., vol. 18, no. 3, pp. 900–909, 2011. to interpret mid-frequency oscillations,” IEEE Trans. Dielectr.
[92] J. Liu, Z. Zhao, C. Tang, C. Yao, C. Li, and S. Islam, “Classifying Electr. Insul., vol. 20, no. 6, pp. 1962–1970, 2013.
transformer winding deformation fault types and degrees using [110] D. M. Sofian, Z. Wang, and J. Li, “Interpretation of transformer
FRA based on support vector machine,” IEEE Access, vol. 7, pp. FRA responses—Part II: Influence of transformer structure,”
112494–112504, 2019, doi: 10.1109/ACCESS.2019.2932497. IEEE Trans. Power Deliv., vol. 25, no. 4, pp. 2582–2589, 2010.
[93] V. Nurmanova, M. Bagheri, A. Zollanvari, K. Aliakhmet, Y. [111] N. Abeywickrama, Y. V Serdyuk, and S. M. Gubanski, “Effect of
Akhmetov, and G. B. Gharehpetian, “A new transformer FRA core magnetization on frequency response analysis (FRA) of
measurement technique to reach smart interpretation for inter-disk power transformers,” IEEE Trans. Power Deliv., vol. 23, no. 3,
faults,” IEEE Trans. Power Deliv., vol. 34, no. 4, pp. 1508–1519, pp. 1432–1438, 2008.
2019, doi: 10.1109/TPWRD.2019.2909144. [112] S. Zhang et al., “Improved flux-controlled VFCV strategy for
[94] X. Wu, Y. He, C. Wang, W. Wu, C. Wang, and J. Duan, “Hilbert eliminating and measuring the residual flux of three-phase
ID considering multi-window feature extraction for transformer transformers,” IEEE Trans. Power Deliv., vol. 35, no. 3, pp.
deep vision fault positioning,” IEEE Access, vol. 8, pp. 91276– 1237–1248, 2019.
91286, 2020, doi: 10.1109/ACCESS.2020.2991844. [113] M. F. M. Yousof, C. Ekanayake, and T. K. Saha, “An
[95] M. Bigdeli, P. Siano, and H. H. Alhelou, “Intelligent classifiers in investigation on the influence of tap changer on Frequency
distinguishing transformer faults using frequency response Response Analysis,” in 2015 IEEE 11th International Conference
analysis,” IEEE Access, vol. 9, pp. 13981–13991, 2021, doi: on the Properties and Applications of Dielectric Materials
10.1109/ACCESS.2021.3052144. (ICPADM), 2015, pp. 963–966.
[96] R. S. De Andrade Ferreira, P. Picher, H. Ezzaidi, and I. Fofana, [114] N. Hashemnia, A. Abu-Siada, and S. Islam, “Detection of power
“Frequency response analysis interpretation using numerical transformer bushing faults and oil degradation using frequency
indices and machine learning: a case study based on a laboratory response analysis,” IEEE Trans. Dielectr. Electr. Insul., vol. 23,
model,” IEEE Access, vol. 9, pp. 67051–67063, 2021, doi: no. 1, pp. 222–229, 2016.
10.1109/ACCESS.2021.3076154. [115] O. Aljohani and A. Abu-Siada, “Identification of the minimum
[97] R. Behkam et al., “Mechanical fault types detection in transformer detection of transformer bushing failure based on Frequency
windings using interpretation of frequency responses via Response Analysis (FRA),” in 2016 IEEE 2nd Annual Southern
multilayer perceptron,” J. Oper. Autom. Power Eng., vol. 11, no. Power Electronics Conference (SPEC), 2016, pp. 1–5.
1, pp. 11–21, 2023. [116] M. Bagheri, B. T. Phung, and T. Blackburn, “Influence of
[98] R. Behkam, H. Karami, M. Salay Naderi, and G. B. Gharehpetian, temperature and moisture content on frequency response analysis
“Generalized regression neural network application for fault type of transformer winding,” IEEE Trans. Dielectr. Electr. Insul., vol.
detection in distribution transformer windings considering 21, no. 3, pp. 1393–1404, 2014.
statistical indices,” COMPEL-The Int. J. Comput. Math. Electr. [117] M. Bagheri, B. T. Phung, and T. Blackburn, “Temperature
Electron. Eng., vol. 41, no. 1, pp. 381–409, 2022. influence on FRA spectrum of oil-filled and oil-free single-phase
[99] H. Ezziane, H. Houassine, S. Moulahoum, and M. S. Chaouche, transformer,” in 2015 IEEE 11th International Conference on the
“A novel method to identification type, location, and extent of Properties and Applications of Dielectric Materials (ICPADM),
transformer winding faults based on FRA and SMOTE-SVM,” 2015, pp. 955–958.
Russ. J. Nondestruct. Test., vol. 58, no. 5, pp. 391–404, 2022. [118] A. A. Reykherdt and V. Davydov, “Case studies of factors
[100] O. Elahi, R. Behkam, G. B. Gharehpetian, and M. N. Heris, “A influencing frequency response analysis measurements and power
new data mining application in smart monitoring systems using transformer diagnostics,” IEEE Electr. Insul. Mag., vol. 27, no. 1,
self organizing map neural network to distinguish disk space pp. 22–30, 2011.
variations in distribution transformers,” in 2022 IEEE Electrical [119] M. Bagheri, B. T. Phung, T. Blackburn, and A. Naderian,
Power and Energy Conference (EPEC), 2022, pp. 248–254. “Influence of temperature on frequency response analysis of

transformer winding,” in 2013 IEEE Electrical Insulation [137] N. Shanmugam, S. Gopal, B. Madanmohan, S. P. Balaji, and R.
Conference (EIC), 2013, pp. 40–44. Rajamani, “Diagnosis of inter-turn shorts of loaded transformer
[120] M. Bagheri, B. T. Phung, and T. Blackburn, “Paper moisture under various load currents and power factors; impulse voltage-
variation vs. mechanical deformation impacts on transformer based frequency response approach,” IEEE Access, vol. 9, pp.
frequency response spectrum,” in Proceedings of 2014 40811–40822, 2021, doi: 10.1109/ACCESS.2021.3064347.
International Symposium on Electrical Insulating Materials, [138] R. Kumar, A. Vaijayanthi, R. Deshmukh, B. Vedik, and C. K.
2014, pp. 128–131. Shiva, “A condition monitoring and fault detection in the
[121] J. M. Guerrero, A. E. Castilla, J. A. S. Fernandez, and C. A. windings of power transformer using impulse frequency response
Platero, “Transformer oil diagnosis based on a capacitive sensor analysis,” Int. J. Syst. Assur. Eng. Manag., pp. 1–13, 2022.
frequency response analysis,” IEEE Access, vol. 9, pp. 7576– [139] S. Seifi, P. Werle, A. A. S. Akmal, and H. Mohseni, “Sweep
7585, 2021, doi: 10.1109/ACCESS.2021.3049192. reflection coefficient method for localization of internal faults in
[122] M. Bigdeli, “Hybrid k-means-PSO technique for transformer transformer windings,” Int. J. Electr. Power Energy Syst., vol.
insulation moisture determination in the production stage based 137, p. 107781, 2022.
on frequency response analysis,” Iran. J. Electr. Electron. Eng., [140] A. Abu-Siada and S. Islam, “A new approach to identify power
vol. 18, no. 1, p. 2072, 2022. transformer criticality and asset management decision based on
[123] J. C. Gonzales and E. E. Mombello, “Automatic detection of dissolved gas-in-oil analysis,” IEEE Trans. Dielectr. Electr.
frequency ranges of power transformer transfer functions for Insul., vol. 19, no. 3, pp. 1007–1012, 2012.
evaluation by mathematical indicators,” in 2012 Sixth IEEE/PES [141] P. A. Venikar, M. S. Ballal, B. S. Umre, and H. M. Suryawanshi,
Transmission and Distribution: Latin America Conference and “A novel offline to online approach to detect transformer interturn
Exposition (T&D-LA), 2012, pp. 1–8. fault,” IEEE Trans. Power Deliv., vol. 31, no. 2, pp. 482–492,
[124] A. Abu-Siada, N. Hashemnia, S. Islam, and M. A. S. Masoum, 2015.
“Understanding power transformer frequency response analysis [142] B. García, J. C. Burgos, and Á. M. Alonso, “Transformer tank
signatures,” IEEE Electr. Insul. Mag., vol. 29, no. 3, pp. 48–56, vibration modeling as a method of detecting winding
2013. deformations-part I: theoretical foundation,” IEEE Trans. Power
[125] N. Hashemnia, A. Abu-Siada, M. A. S. Masoum, and S. M. Islam, Deliv., vol. 21, no. 1, pp. 157–163, 2005.
“Characterization of transformer FRA signature under various [143] D. Zacharias and R. Gokaraju, “Prototype of a negative-sequence
winding faults,” in 2012 IEEE International Conference on turn-to-turn fault detection scheme for transformers,” IEEE
Condition Monitoring and Diagnosis, 2012, pp. 446–449. Trans. power Deliv., vol. 31, no. 1, pp. 122–129, 2015.
[126] A. Amini, N. Das, and S. Islam, “Impact of buckling deformation [144] L. M. R. Oliveira and A. J. M. Cardoso, “A permeance-based
on the FRA signature of power transformer,” in 2013 Australasian transformer model and its application to winding interturn arcing
Universities Power Engineering Conference (AUPEC), 2013, pp. fault studies,” IEEE Trans. Power Deliv., vol. 25, no. 3, pp. 1589–
1–4. 1598, 2010.
[127] Y. Li, G. Liu, L. Zhang, L. Zhang, and Z. Lin, “Transformer [145] A. S. Masoum, N. Hashemnia, A. Abu-Siada, M. A. S. Masoum,
winding deformation diagnosis using middle band frequency and S. M. Islam, “Online transformer internal fault detection
response analysis,” in 2007 IEEE International Conference on based on instantaneous voltage and current measurements
Solid Dielectrics, 2007, pp. 677–680. considering impact of harmonics,” IEEE Trans. Power Deliv.,
[128] N. Hashemnia, A. Abu-Siada, and S. Islam, “Improved power vol. 32, no. 2, pp. 587–598, 2014.
transformer winding fault detection using FRA diagnostics–part [146] S. C. Athikessavan, E. Jeyasankar, S. S. Manohar, and S. K.
1: axial displacement simulation,” IEEE Trans. Dielectr. Electr. Panda, “Inter-turn fault detection of dry-type transformers using
Insul., vol. 22, no. 1, pp. 556–563, 2015. core-leakage fluxes,” IEEE Trans. Power Deliv., vol. 34, no. 4,
[129] N. Hashemnia, A. Abu-Siada, and S. Islam, “Impact of axial pp. 1230–1241, 2018.
displacement on power transformer FRA signature,” in 2013 [147] X. Ouyang, Q. Zhou, H. Shang, Y. Zheng, S. Pan, and J. Luo,
IEEE Power & Energy Society General Meeting, 2013, pp. 1–4. “Towards a comprehensive evaluation on the online methods for
[130] J. Jayasinghe, Z. D. Wang, P. N. Jarman, and A. W. Darwin, monitoring transformer turn-to-turn faults,” IEEE Trans. Ind.
“Investigations on sensitivity of FRA technique in diagnosis of Electron., 2022.
transformer winding deformations,” in Conference Record of the [148] V. Behjat, A. Vahedi, A. Setayeshmehr, H. Borsi, and E.
2004 IEEE International Symposium on Electrical Insulation, Gockenbach, “Diagnosing shorted turns on the windings of power
2004, pp. 496–499. transformers based upon online FRA using capacitive and
[131] M. A. Sathya, A. J. Thomas, and S. Usa, “Prediction of inductive couplings,” IEEE Trans. Power Deliv., vol. 26, no. 4,
transformer winding displacement from frequency response pp. 2123–2133, 2011.
characteristics,” in 2013 IEEE 1st International Conference on
Condition Assessment Techniques in Electrical Systems
(CATCON), 2013, pp. 303–307.
[132] N. Hashemnia, A. Abu-Siada, and S. Islam, “Improved power
transformer winding fault detection using FRA diagnostics–part
2: radial deformation simulation,” IEEE Trans. Dielectr. Electr.
Insul., vol. 22, no. 1, pp. 564–570, 2015.
[133] E. Gomez-Luna, G. A. Mayor, C. Gonzalez-Garcia, and J. P.
Guerra, “Current status and future trends in frequency-response
analysis with a transformer in service,” IEEE Trans. Power Deliv.,
vol. 28, no. 2, pp. 1024–1031, 2013.
[134] C. Gonzalez, J. Pleite, V. Valdivia, and J. Sanz, “An overview of
the on line application of frequency response analysis (FRA),” in
2007 IEEE International Symposium on Industrial Electronics,
2007, pp. 1294–1299.
[135] A. Vandermaar, “On-line frequency response analysis system,”
Dev. Specif. EPRI Proj. Manag. Tech. Progress, Dev. Specif., no.
1001942, 2001.
[136] N. Shanmugam, B. Madanmohan, and R. Rajamani, “Influence of
the load on the impulse frequency response approach based
diagnosis of transformer’s inter-turn short-circuit,” IEEE Access,
vol. 8, pp. 39454–39463, 2020.

Seyed Ebrahim Hosseini Kakolaki was born in

Shahrekord, Iran, in 1992. He received the B.Sc.
degree in electrical engineering from Yazd
University, Yazd, Iran, in 2015 and the M.Sc. degree
from Iran University of Science and Technology,
Tehran, Iran, in 2018, respectively. Since 2018 until
2021, he worked as a Ph.D. student and since 2021,
as a research assistant at Ferdowsi University of
Mashhad, Mashhad, Iran. His current research
interests include power system protection, fault
detection, and transient analysis in power systems.

Vahid Hakimian was born in Tabas, Iran, in

1993. He received the B.Sc. degree in electrical
engineering from University of Tehran, Tehran,
Iran, in 2015 and the M.Sc. degree from Tarbiat
Modares University, Tehran, Iran, in 2018,
respectively. He then joined a consulting company
where he works as an engineering consultant in the
electrical power industry of Iran and in the field of
HV substation design. His current research
interests include power system operation and
planning, power system protection, and electrical
power distribution systems.

Javad Sadeh was born in Mashhad, Iran, in

1968. He received the B.Sc. and M.Sc. degrees
in electrical engineering (Hons.) from Ferdowsi
University of Mashhad, Mashhad, Iran, in 1990
and 1994, respectively, and the Ph.D. degree in
electrical engineering from Sharif University of
Technology, Tehran, Iran, with the collaboration
of the electrical engineering laboratory of the
Institut National Polytechnique de Grenoble
(INPG), Grenoble, France, in 2001. Currently, he is Professor in the
Department of Electrical Engineering, Ferdowsi University of Mashhad. His
research interests are power system protection, dynamics, operation as well
as renewable energy technologies.

Elyas Rakhshani is a senior control and power

systems engineer, leading the algorithm and control
group at the HESStec company in Spain. With a
focus on dynamic low-inertia grids, hybrid energy
storage integration, integrating models’ behavior,
and operation, through the development of advanced
control algorithms with their integration into
HESStec's control and power platforms. He obtained
his Ph.D. in Electrical Engineering (cum laude) from
the Universitat Politècnica de Catalunya (UPC),
Barcelona, Spain, in 2016. He also got his master's degree in control systems
in 2008 and his degree in power systems in 2004. In 2018 he received an
extraordinary doctoral award from the permanent committee of the UPC
doctoral school for his contribution on different methodology for emulating
virtual inertia after his Ph.D. He is a Senior member at IEEE and a member
of the editorial board of Power and Energy Society (PES) journals such as
IEEE Transactions on Power Systems, IEEE Power Engineering Letters,
IET Generation Transmission & Distribution, IET Renewable Power
Generation, and IEEE System journal. His research interests are modern
power system control, dynamic stability, hybrid energy storage integration,
converter control applications in power systems, and HVDC control for grid
applications, frequency control.

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/