IEEJ TRANSACTIONS ON ELECTRICAL AND ELECTRONIC ENGINEERING

IEEJ Trans 2014; 9: 302314

Published online in Wiley Online Library (wileyonlinelibrary.com). DOI:10.1002/tee.21971

Paper
A Discrete Wavelet Transform Approach to Discriminating among Inrush
Current, External Fault, and Internal Fault in Power Transformer using
Low-Frequency Components Differential Current Only
Atthapol Ngaopitakkula , Non-member
Chaiyan Jettanasen, Non-member
This paper proposes an algorithm based on discrete wavelet transform (DWT) for discriminating among inrush current,
internal fault, and external fault in power transformers. Fault conditions are simulated using the Alternative Transients
Program/Electromagnetic Transients Program (ATP/EMTP). Daubechies4 (db4) is employed as the mother wavelet to decompose
low-frequency components from fault signals. The ratio between per unit (p.u.) differential current and p.u. time is suggested as
an index. The numerator of the ratio is the difference between the maximum differential current and the minimum differential
current in terms of p.u. with a base value selected at the transformer-rated current. The ratio is calculated for all three phases,
and from a trial and error process the indices for the separation among the internal fault condition, the external fault condition,
and inrush condition are defined. The results obtained from the proposed technique show good accuracy for discriminating faults
in the considered system. In addition, the proposed algorithm uses data of the differential current with a time of quarter cycle
under the analysis. 2014 Institute of Electrical Engineers of Japan. Published by John Wiley & Sons, Inc.
Keywords: interturn fault, power transformer, discrete wavelet transforms, inrush current

Received 15 January 2013 Revised 20 June 2013

1. Introduction
A reliable supply of electric power is essential, and a failure
of any part of the installed equipment is expensive not only
for utilities but also for the manufacturing industry. A power
utility may lose revenues and incur penalties for nondelivery,
while the failure of an industrial transformer, for example, may
lead to lengthy and therefore costly downtime. The differential
relaying principle is used for the protection of medium and
large power transformers. In the literature for fault detection,
several decision algorithms [142] have been developed to be
employed in the protective relay for preventing maloperation
of the protective equipment under different nonfault conditions,
including magnetizing inrush current, ratio mismatch, throughfault current, etc. There are many techniques [142] for detecting
faults, such as artificial neural networks (ANNs) [12,30,39,40],
transient-based protection [13,1820], finite element [14], fuzzy
logic [35], hybrid systems [14,32], and so on. An algorithm for
protecting a transformer with three windings using the increments
of flux linkages (IFLs) has been proposed by Kang et al . [2]. Nine
detectors and a rule are suggested for fault detection, the faulted
phase, and winding identification. Mathematical morphology has
been proposed to identify the inrush current [3]. It is able to
discriminate between inrush and internal fault currents even in
the case of an inrush with a low second-harmonic component and
an internal fault current with a high second-harmonic component.
A novel technique to distinguish the inrush currents from internal
faults in a power transformer is proposed by Ma et al . [4] using
the normalized grille curve (NGC). NGC is an effective tool for
transient signal analysis. The NGC calculation method is first
a

Correspondence to: Atthapol Ngaopitakkul. E-mail:

knatthap@kmitl.ac.th

Faculty of Engineering, King Mongkuts Institute of Technology Ladkrabang, Bangkok 10520, Thailand

derived. Then, the criteria to extract features of the inrush current

and the internal fault in the respective time and frequency domains
are developed in detail. This technique is able to identify inrush
currents rapidly and can be implemented with only a small amount
of computation. A method for distinguishing the fault-originated
transients from the switching transients has been developed [6].
The decision trees (DTs), hidden Markov models (HMM), and
probabilistic neural network (PNN) techniques were compared, and
the PNN classifier gave the best classification results. In Ref. [7],
an index is proposed to discriminate external faults, incipient
faults, inrush currents, and internal faults by using the hyperbolic
S-transform-based method. This method has two steps; in the first
step, external faults are discriminated from other disturbances. If
an internal disturbance is detected, then the absolute deviation
of the S-transform matrix values will be applied. In Ref. [30],
a support vector machine (SVM) classification technique for the
detection of minor internal turn-to-turn faults was presented. The
discrimination between internal turn-to-turn fault, external fault,
and the inrush condition can be made within one-half cycle
from the fault inception time using this technique. According
to this technique, one-half-cycle post-fault current samples are
the inputs for the SVM classifier. In Ref. [37], an online sweep
frequency-response analysis (SFRA) was developed to detect
winding interturn faults of power transformers in service using
the transfer function method.
The idea of application of wavelet transform to fault diagnosis is
not new, and there are a number of research papers related to this
idea [1127]. In Ref. [12], the development and hardware implementation of the wavelet transform and HMM-based classifiers to
distinguish between the transients originating from faults and from
other types of transients was presented. The transient recognition
scheme uses wavelet transforms for the extraction of features, and
the HMM is used for the classification. In Ref. [13], the transient
signal analysis with discrete wavelet transform (DWT) was used

2014 Institute of Electrical Engineers of Japan. Published by John Wiley & Sons, Inc.

DISCRIMINATING AMONG INRUSH CURRENT, EXTERNAL FAULT, AND INTERNAL FAULT IN POWER TRANSFORMERS

to identify and correctly differentiate inrush current from incipient internal faults. The decision algorithm is based on a ratio
index quantified in a certain window of analysis. The ratio index
is defined as the relation between the maximum coefficient from
the first detail of the DWT decomposition and the spectral energy
of the other frequency components present in the same detail.
In addition, different mother wavelets were compared, and the
Daubechies wavelet was found to give excellent performance and
high efficiency in the discrimination of simulated disturbances. An
implementation of d q axis components and wavelet packet transform (WPT) based hybrid technique was introduced by Aktaibi
et al . [2527]. This technique is based on extracting the highfrequency subband contents present in the d q axis components
of the differential currents. It requires only one level of WPT for
the synchronously rotating reference frame (d q axis) components
of the differential current to accurately discriminate inrush currents
from all types of the internal fault currents. In previous research
works [20], the low-frequency component obtained from DWT of
differential current is analyzed. The proposed decision algorithm
gave more satisfactory results for the separation between internal fault and external fault, but the case studies were made with
a power transformer that was connected with Y-Y configuration.
In fact, power transformers connected with -Y are more widely
employed than those connected with Y-Y in power systems, so the
decision algorithm should be proved to discriminate in both -Y
and Y-Y connections.
Therefore, in this paper we develop a decision algorithm for
detecting and discriminating between inrush current, internal fault,
and external fault for a power transformer. The decision algorithm is based on DWT as an alternative to or improvement
upon the existing protective relaying functions. The simulations
and analyses are performed using the Alternative Transients Program/Electromagnetic Transients Program (ATP/EMTP) and MATLAB. ATP/EMTP is a universal program for digital simulation of
transient phenomena of electromagnetic as well as electromechanical nature, thus ATP/EMTP is probably the most widely used
power system transients program in the world today. The scheme
under investigations is a part of Thailands electricity transmission
and distribution system. In addition, the transformer model with
stray capacitances is used.

Power transformer

(a)

Ioperating coil = 0

Power transformer

(b)

Fault

Ioperating coil

Fig. 1. Basic differential scheme for the power transformer.

(a) Normal condition. (b) Internal fault condition
f(n)

50100 kHz

HF

LF

1
2550 kHz

HF

LF

2
12.525 kHz

fs = 200 kHz

HF

LF

3
6.2512.5 kHz

HF

LF

4
3.1256.25 kHz

HF = High pass filter

HF

LF

LF = Low pass filter

03.125 kHz

Fig. 2. Five-level wavelet decomposition tree

can be finely adjusted so that both the high- and low-frequency
components are precisely detected. The results from the wavelet
transform are shown in both the time domain and the frequency
domain. The wavelet transform can expand signals by using either
a shift or a translation time as well as a compression in time or
a dilation of a fixed wavelet function referred to as the mother
wavelet. The wavelet transform that scales the results of the
analysis by a factor of two is called a DWT as expressed in (1).


n k 2m
1 
f (k )
(1)
DWT(m, n) =
2m
2m

2. Theory
2.1. Differential protection The differential principle,
as applied for protecting power transformers, can be described
with the help of Fig. 1. The levels of currents in the primary
and secondary sides of the power transformer are reduced by the
current transformers (CTs). The outputs of these CTs are compared.
The ratios of the primary and secondary CTs are selected such
that the CTs produce the same secondary current for nominal line
current.
The operating coil of the relay is connected to the secondary
windings of the two CTs in such a way that the net current flowing
through it is equal to the difference between the secondary currents
of CTs provided on the two sides of the power transformer. The
net current through the operating coil of differential relay is zero
for normal operation and external faults. An internal fault in the
power transformer breaks this balance and causes a current to flow
in the relays operating coil. This is shown in Fig. 1(b).

nk 2m
2m

= mother wavelet (In this paper, Daubechies 4

where
is selected as the mother wavelet).
Referring to Fig. 2, the original input signal is split up into
two parts in the first stage (scale 1) by passing the signal from
a high-pass and a low-pass filter, which results in two different
versions of the same signal: the portion of the signal corresponding
to low-frequency components of the signal to analyze the low
frequencies is called approximations (low-pass portion), while
the portion of the signal corresponding to the high-frequency
components of the signal to analyze the high frequencies is called
details (high-pass portion). The coefficients of approximations and
details can be calculated by iterating or cascading the single-stage
filter bank to obtain a multiple-stage filter bank. The coefficients
of approximations contain most of the information content of
the original input signal or the general trend. The coefficient of

2.2. Wavelet transforms A wavelet is a small, localized

wave of a particular shape and finite duration that has an average
value of zero. The wavelet transform is a tool that splits up data,
functions, or operators into different frequency components, and
then studies each component with a resolution adjusted to its
scale. The advantage of the transform is that the analysis band
303

IEEJ Trans 9: 302314 (2014)

A. NGAOPITAKKUL AND C. JETTANASEN

a

details contains its local variations or the difference between the

true input and the value of the reconstructed input if it has to
be reconstructed from only the information given in the lowpass output. After passing these data through the filter functions,
the output of the low-pass filter (coefficients of approximations)
consists of the average of every two samples, and the output of
the high-pass filter (coefficient of details) consists of the difference
of every two samples. The high-pass filter obviously contains less
information than the low-pass output. If the signal is reconstructed
by an inverse low-pass filter of the form, then the result is a
duplication of each entry from the low-pass filter output.
For other stages (scale 2, . . . , scale 5), the filters of each
stage have different cut-off frequencies and bandwidths, while the
processed signal is still unchanged. The frequency bandwidth of
the band decreases with growing level scale, which means that
the frequency resolution becomes higher by increasing the level
scale. However, the higher the scale, the longer is the processing
time of the signal. The increase in processing time is a problem
when the scale is high. In general, higher order wavelets tend to
put more information into the low-pass output, and vice versa. If
the average amplitude of the high-pass output is low enough, then
the high-pass half of the signal may be discarded without greatly
affecting the quality of the reconstructed signal.

a
Phase A

Phase A

Phase B
3

Phase B

Phase C
5

Phase C

Primary

Secondary

Primary

Secondary

Fig. 3. The modification on ATP/EMTP model [43] for a threephase transformer with internal faults
Chl

Chg

Clg

3. Power System Simulation using EMTP

3.1. Transformer winding models For a computer
model of a two-winding three-phase transformer having primary
and secondary windings in each phase, BCTRAN is a wellknown subroutine on ATP/EMTP. To study internal faults of the
transformer, Bastard et al . [43] proposed a modification of the
BCTRAN subroutine. Normally, the BCTRAN uses a matrix of
inductances with a size of 6 6 to represent a transformer, but
with the internal fault conditions the matrix is adjusted to be of
size 7 7 for winding-to-ground faults and of 8 8 for interturn
faults [43]. In the research work of Bastard et al . [43], the model
was proved to be validated and accurate as shown by a comparison
with measurement results. However, the effects of high-frequency
components that may occur during the faults are not included in
such a model. Islam and Ledwich [44] described the characteristics
of high-frequency responses of a transformer due to various faults.
It has been shown that the fault types and fault locations have an
influence on the frequency responses of the transformer [44]. As
a result, in this paper the combination of the transformer models
proposed by Bastard et al . [43], as shown in Fig. 3, with the highfrequency model including stray capacitances of the transformer
recommended by IEEE working group [45], as shown in Fig. 4, is
used for simulations of internal faults in the transformer windings.
From Fig. 3, for the phase winding of the transformer with
internal faults, the winding is divided into two parts in case of
winding-to-ground faults and three parts in case of interturn faults.
The process for simulating internal faults based on the BCTRAN
routine of EMTP can be summarized as follows:
First step: Compute matrices [R] and [L] of the power transformer from manufacturers test data [20,31,46] without considering the internal faults [26].

R1 . . . 0

..
..
[R] = ...
.
.
0 R6

L1 L12 L16
L21 L2 L26

[L] = .
..
..
..
..
.
.
.
L61 L62 L6

Primary

115/23 kV Secondary
50 MVA

Fig. 4. A two-winding transformer with the effects of stray

capacitances
Second step: Modify (4) and (5) to obtain the new internal
winding fault matrices [R] and [L] as illustrated in (4) and
(5) [43].

Ra
0
0
0
0
0
0
0
0 Rb 0
0
0
0
0
0

0
0
0
0
0
0
R
c

0
0
0
0
0
0 R2 0

(4)
[R] =
0
0
0
0
0
0 R3 0

0
0
0
0
0 R4 0

0
0
0
0
0
0 R5 0
0
0
0
0
0
0
0 R6

La
Mab Mac Ma2 Ma3 Ma4 Ma5 Ma6
Mba
Lb
Mbc Mb2 Mb3 Mb4 Mb5 Mb6

Mca Mcb
Lc
Mc2 Mc3 Mc4 Mc5 Mc6

M2a M2b M2c

L2
M23 M24 M25 M26
(5)
[L] =
M3a M3b M3c M32
L3
M34 M35 M36

M4a M4b M4c M42 M43

L4
M45 M46

M5a M5b M5c M52 M53 M54

L5
M56
M6a M6b M6c M62 M63 M64 M65
L6
Third step: The inter-winding stray capacitances and earth capacitances of the high-voltage (HV) and low-voltage (LV) windings
can be simulated by adding lumped capacitances connected to the
terminals of the transformer, as shown in Fig. 4.
The capacitances shown in Fig. 4 are as follows:
Chg = stray capacitance between the high-voltage winding and
ground
Clg = stray capacitance between the low-voltage winding and
ground
Chl = stray capacitance between the high-voltage winding and
the low-voltage winding.

(2)

(3)

304

IEEJ Trans 9: 302314 (2014)

DISCRIMINATING AMONG INRUSH CURRENT, EXTERNAL FAULT, AND INTERNAL FAULT IN POWER TRANSFORMERS

3.2. Simulation case studies In this paper, the combination between the transformer models proposed by Bastard
et al [43] and the high-frequency model including capacitances
of the transformer recommended by IEEE working group [45] is
used for simulations of faults in the transformer windings. The
scheme under investigations is a part of Thailand electricity transmission and distribution system as depicted in Fig. 5. A 50-MVA,
115/23-kV three-phase two-winding transformer was employed in
simulations with all parameters and configuration provided by the
manufacturer [20,31,46].

EGAT
RE

PEA

115/23 kV
50 MVA

MK

R+jX
Primary side
current

Secondary side
current

Load

Fig. 5. System used in simulation studies [21,32,47]

(a)

RLC

115/23

Line
RL +/0

RLC

RLC

Line
RL+/0

RLC

(b)

RLC

115/23

RLC

RLC

Line
RL+/ 0

RLC

Line
RL+/0

RLC

(c)
80

20

Line
RL+/0

115/23

80

20

Line
RL+/0

Line
RL+/0

RLC

RLC

Line
RL+/0

RLC

RLC

(d)
I

115/23

RLC

Line
RL+/0

v
Line
RL+/0

RLC

RLC

Fig. 6. Components of a proposed simulation model implemented in ATP/EMTP. (a) The winding-to-ground fault case. (b) The interturn
fault case. (c) The external fault case. (d) The inrush case
305

IEEJ Trans 9: 302314 (2014)

A. NGAOPITAKKUL AND C. JETTANASEN

Table III. Number of case studies and parameters for case of

external fault

Table I. Number of case studies and parameters for case of

winding to ground fault
Parameter of
winding-to-ground
fault
Phase in which
fault occurs
Transformer
winding
Angles of fault
inception
Fault position

Detail

Number
of case
studies

Phases A, B, and C

High voltage and low voltage

(primary and secondary)
0 330 (each step is 30 ) and
phase A voltage is reference
1090% (each step is 10%)
measured from the line end
of the windings as shown in
Fig. 7(a).

Parameter of
external fault
Phase in which
fault occurs
Transformer side
Angles of fault
inception
Type of fault

12
9

Fault location
Table II. Number of case studies and parameters for case of
interturn fault
Parameter of
interturn fault
Phase in which
fault occurs
Transformer
winding
Angles of fault
inception
Fault position of
point ZAF

Fault position of
point ZINF

Detail
Phases A, B, and C
High voltage and low voltage
(primary and secondary)
0 330 (each step is 30 ) and
phase A voltage is reference
1080% (each step is 10%)
measured from the line end
of the windings as shown in
Fig. 7(b).
1080% (each step is 10%)
measured from the line end
of the windings as shown in
Fig. 7(b).

Fault resistance

Number
of case
studies

Detail

Number
of case
studies

Phase A

High voltage and low voltage

(primary and secondary)
0 330 (each step is 30 ) and
phase A voltage is reference
single line-to-ground, double
line-to-ground, line-to-line,
and three-phase faults (AG,
ABG, AB, and ABC)
20, 40, 60, and 80% measured
from the power transformer
5

2
12
4

4
1

Table IV. Number of case studies and parameters for case of

inrush current

3
Parameter of
inrush current

2
12

Angles of fault
inception

36

Magnetizing fluxes

Detail

0 330 (each step is 30 )

and phase A voltage is
reference
80, 90, 100, 110, and 120%

Number
of case
studies
12

current in all three phases, and the zero sequence are calculated,
and the resulting current signals are employed to decompose
the high-frequency (details) and low-frequency (approximations)
components from the simulated current signals using the mother
wavelet daubechies4 (db4) [1820].
From our previous paper, the coefficient details of the resulting
current signals obtained from the DWT are squared. It is clearly
seen that when a fault occurs, the coefficients of high-frequency
components from each scale have a sudden change compared
with those before the occurrence of the faults. This sudden
change is used as an index for the occurrence of faults using
comparison of the coefficients details from each scale. However,
the similarity between the waveforms of the internal faults
and the external fault signals and from coefficients of highfrequency components can be seen obviously. To overcome this
problem, the comparison of the coefficients of the low-frequency
components from each phase is considered. Examples of the
approximated signal of the extracted waveform using DWT for
the differential currents from scale 1 to scale 5 are illustrated in
Figs 1215.
After applying the DWT, Figs 1215 show several examples
of extraction using DWT for the differential currents and zero
sequence current from scale 1 to scale 5. Figs 12 and 13 show a
case of an internal fault at 20% of the winding length and a fault
inception angle of 120 . An example of an external fault case at
20% of the length of the transmission line and fault inception angle
of 120 is illustrated in Fig. 14, while an example of the inrush
current condition is illustrated in Fig. 15.
Generally, during normal condition the amplitude of each phase
must be nearly treated as zero, but during fault condition the
amplitude of phase in which the fault occurs has a sudden change.
By considering Fig. 12(a), when 20% of the length of the highvoltage winding with  connection is considered as an example,

The ATPDraw circuit of the proposed simulation model is shown

in Fig. 6; it can be seen that the transformer, which is a step-down
transformer, is connected between two subtransmission sections.
The BCTRAN model based on test data can be obtained from
the transformer manufacturers. Supporting routine of BCTRANS
can be used to derive a linear representation for three-phase
transformer, using test data of both the excitation test and the
short-circuit test. The magnetizing branch is represented by a
hysteretic nonlinear inductor model generated by the HYSDAT
supporting routine of ATP. To implement or study the transformer
model, simulations were performed with various changes in system
parameters, as shown in Tables I IV.
The primary and secondary current waveforms can then be
simulated using ATP/EMTP, and these waveforms are interfaced
to MATLAB/Simulink for the construction of the fault diagnosis
process. The fault signal in each phase is obtained from the primary
and secondary currents of the transformer, as shown in Figs 811.
The data in these figures correspond to the two protection zones.
The differential currents, which are the difference between the
primary current and the secondary current in all three phases, and
the zero sequence are calculated, and the resulting current signals
are extracted using the DWT.

4. Decision Algorithm and Results

From the simulated signals, the differential currents, which are
the difference between the primary current and the secondary
306

IEEJ Trans 9: 302314 (2014)

DISCRIMINATING AMONG INRUSH CURRENT, EXTERNAL FAULT, AND INTERNAL FAULT IN POWER TRANSFORMERS

(a)

(b)

ZAF
Fault

A
1

ZAF

4
5
1

ZCF

ZINF

ZCF

Fault high voltage coil 1

Fault high voltage coil 1

Fig. 7. Modification on an ATP/EMTP model for a three-phase transformer. (a) The winding-to-ground fault case. (b) The interturn fault
case
(a)

Primary current

x 104

(b)
A
C

A
B

0
500

0.5
1

500
(A)

(A)

0.5

Primary current
1000

0.025

0.05

0.075

1000

0.1

0.025

0.05

Time (s)

0.075

0.1

0.075

0.1

Time (s)

Secondary current

Secondary current
A B

2000

5000

(A)

(A)

1000
0

1000
2000
0

0.025

0.05

0.075

5000

0.1

0.025

0.05
Time (s)

Time (s)

Fig. 8. Primary and secondary currents obtained from simulation for a case of winding-to-ground fault (fault at 20% of winding length
and inception angle of 120 ). (a) The high- and (b) low-voltage winding fault cases
x 104

(a)

Primary current
A

(b)

1500
(A)

(A)

Primary current
3000

0
1500

1
0

0.025

0.05

0.075

3000

0.1

0.025

0.05

Time (s)
Secondary current
A

0.1

0.075

0.1

Secondary current
3000

1500
(A)

(A)

2000

0.075

Time (s)

0
1500

2000
0

0.025

0.05

0.075

0.1

Time (s)

3000

0.025

0.05
Time (s)

Fig. 9. Primary and secondary currents obtained from simulation for a case of interturn fault (fault between 10 and 20% of winding length
and inception angle of 120 ). (a) The high- and (b) low-voltage winding fault cases
the input differential current signal is plotted in the top trace of
the figure. It can be observed that the amplitude of each phase is
different because the phases A and C (coil 1, as shown in Fig. 7(a))
are the ones in which the fault occurs so that the amplitude of
phases A and C increases immediately after the fault occurrence
and has a value more than the other phase (phase B). This indicates
that the decision algorithm can benefit from variations of the
coefficient approximations.
In addition, by considering Fig. 12(a), the input signal implementation is a multisignal trace from each low-pass filter, which
corresponds to a particular scale parameter, as shown in Fig. 12(a).
The traces labeled scale 1, scale 2, . . . , scale 5 in this figure
correspond to the filter output of Fig. 2. It can be seen that the

amplitudes of the coefficients approximations on each scale are

related to the frequency banks according to the scale. As a result,
by observing Figs 1215, it is clear that the coefficients of the
low-frequency components have a sudden change when a fault
occurs compared to those before a fault occurrence at 0.04 s.
However, by performing many case studies, the patterns in the
raw data are hard to discriminate. To overcome this problem,
the ratio between per-unit differential current and per-unit time is
calculated and Phmax , Phsum , and Zchk are performed as comparison
indicators in order to discriminate between the internal fault
condition and the external fault condition. Hence, the complexity
of the patterns is significantly reduced. The ratio is calculated as
follows:
307

IEEJ Trans 9: 302314 (2014)

A. NGAOPITAKKUL AND C. JETTANASEN

(a)

(b)

Primary current
600

Primary current
1000

0
300
600

500
(A)

(A)

300

0
500

0.025

0.05

0.075

1000

0.1

0.025

0.05

Time (s)

0.1

0.075

0.1

Secondary current

Secondary current
2000

0.075

Time (s)

5000

(A)

(A)

1000
0

1000
2000 0

0.025

0.05

0.075

0.1

5000

0.025

0.05
Time (s)

Time (s)

Fig. 10. Primary and secondary currents obtained from simulation for a case of external fault (fault at 20% of transmission line and
inception angle of 120 ). (a) The high- and (b) low-voltage side fault cases
diff
Xmin
= the minimum coefficient from the approximated differential signal of DWT
diff_x
tmax
= the time at which the maximum coefficient of the
approximated differential signal occurs
diff_x
tmin
= the time at which the minimum coefficient of the
approximated differential signal occurs
diff
Xchk
= comparison indicator for separation between internal
fault condition and external fault condition
Irated = rated current of the power transformer
T = the period of the power frequency of the system
Zchk = zero sequence current of comparison indicator for
separation between internal fault condition and external fault
condition
diff
Phmax = maximum value of comparison indicators (Xchk
) used
in classifying the fault condition
diff
Phsum = summation value of comparison indicators (Xchk
) used
in classifying the fault condition
The numerator of the ratio is the difference between the
maximum differential current and the minimum differential current
in terms of per unit with a base value selected at a transformers
rated current. The denominator of the ratio is the difference
diff_x occurs and the time when t diff_x
between the time when tmax
min
occurs, with a quarter cycle of power frequency (in this case,
power frequency = 50 Hz).

Primary current
1000

(A)

500
0
500
1000

0.01

0.02

0.03

0.04

0.05

0.06

0.07

0.08

0.09

0.1

0.07

0.08

0.09

0.1

Time (s)
Secondary current
2000
C

1000
(A)

0
1000
2000
0

0.01

0.02

0.03

0.04

0.05

0.06

Time (s)

Fig. 11. Primary and secondary currents obtained from simulation

for a case of inrush condition (inception angle of 120 )

diff
Xchk
=

diff X diff )
(Xmax
min
Irated
diff_x
diff_x
(tmax tmin )
T

(6)

where
diff
Xmax = the maximum coefficient from the approximated- differential signal of DWT
Phase C

Zero sequence

50
50

100
100

200
200

0.5
0.5

50
50

100
100

200
200

0.5
0.5

50
50

100
100

200
200

0.5
0.2

50
50

100
100

200
200

0.2
0.2

50
50

100
100

0
0

0.05

0.2
0.1
0

0
0.05

50

0.1
0
Time (s)

Scale 1

0
0.5
0.5

Scale 2

0
200
200

200

(b)

100

Scale 3

50

Scale 4

Scale 1
Scale 2
Scale 3
Scale 4
Scale 5

Phase B
0.5

200

0
0.05

100
0.1
0

Phase A

Phase B

0.05

0.1

Phase C

Zero sequence

10
0

0.5
0

10
0

0.1
0

10
10
0

0.5
0.5
0

10
10
0

0.1
0.1
0

10
10
0

0.5
0.5
0

10
10
0

0.1
0.1
0

10
10
0

0.5
0.5
0

10
10
0

0.1
0.1
0

10
10

0.5
0.2
0

10
10
0

0.1
0.1
0

0.2
0.2
0

10
10
0

0.1
0.02
0

10
10
0

Scale 5

Phase A

(a)

10

0.05

0.1

0.2

0.05

10
0.1
0
Time (s)

0.05

0.02
0.1
0

0.05

0.1

Fig. 12. DWT of differential currents for internal fault case (winding phase A-to-ground fault at 20% of winding length and inception
angle of 120 ). Fault occurring at (a) high-voltage winding and (b) low-voltage winding
308

IEEJ Trans 9: 302314 (2014)

DISCRIMINATING AMONG INRUSH CURRENT, EXTERNAL FAULT, AND INTERNAL FAULT IN POWER TRANSFORMERS

0.2
0
0.2

200
0
200

0.02
0
0.02

200
0
200

0.2
0
0.2

200
0
200

0.02
0
0.02

0.2
0
0.2

200
0
200

0.01
0
0.01

0.2
0
0.2

200
0
200

5
0
5

Scale 5

200
0
200
200
0
200

0.05

0.1

0.05

0.1
0
Time (s)

0.05

0.1

Scale 1

200
0
200

Scale 2

0.02
0
0.02

Scale 3

200
0
200

Scale 4

0.2
0
0.2

Scale 4

200
0
200

x 10
0

0.05

Phase A

(b)

Zero sequence
0.02
0
0.02

Scale 1

Phase C
200
0
200

Scale 2

Phase B
0.2
0
0.2

Scale 3

Phase A
200
0
200

Scale 5

(a)

0.1

50
0
50
50
0
50
50
0
50
50
0
50
50
0
50
50
0
50

0.05

Phase B

0.1

0.5
0
0.5
0.5
0
0.5
0.5
0
0.5
0.5
0
0.5
0.2
0
0.2
0.2
0
0.2

0.05

Phase C
50
0
50
50
0
50
50
0
50
50
0
50
50
0
50
50
0
50

0.1
0
Time (s)

0.05

Zero sequence

0.1

0.1
0
0.1
0.1
0
0.1
0.1
0
0.1
0.1
0
0.1
0.1
0
0.1
0.02
0
0.02

0.05

0.1

Fig. 13. DWT of differential currents for internal fault case (interturn fault between 10 and 20% of winding length and inception angle
of 120 ). Fault occurring at (a) high- and (b) low-voltage winding
Phase C

1
1

1
1

5
5

10
5

1
1

1
1

5
5

5
2

1
1

1
1

5
2

2
1

1
0.2

1
0.5

2
1

1
0.5

0.2
0.2

0.5
0.2

1
0.5

0.5

0.2

0.2

0.5

0.05

0.1

0.05

0.1
0
Time (s)

0.05

0.1

Scale 1

10
10

(b)

Scale 2

Scale 3

Zero sequence

Scale 4

Scale 1

Scale 3

Scale 2

Phase B

Scale 4
Scale 5

Phase A

10

Scale 5

(a)

0.05

0.1

Phase A

Phase B

Phase C

Zero sequence

0.5
0

0.2
0

0.5
0

0.1
0

0.5
0.5
0

0.2
0.2
0

0.5
0.5
0

0.1
0.1
0

0.5
0.5
0

0.2
0.5
0

0.5
0.2
0

0.1
0.1
0

0.5
0.5
0

0.5
0.5
0

0.2
0.2
0

0.1
0.1
0

0.5
0.2
0

0.5
0.2
0

0.2
0.2
0

0.1
0.1
0

0.2
0.2
0

0.2
0.2
0

0.2
0.2
0

0.1
0.02
0

0.2

0.05

0.1

0.2

0.05

0.2
0.1
0
Time (s)

0.05

0.1

0.02

0.05

0.1

Fig. 14. DWT of differential currents for external fault case (phase A-to-ground fault at 20% of transmission line and inception angle of
120 ). Fault occurring at (a) high- and (b) low-voltage sides

10
10

20
20

10
10

0.05
0.05

10
10

20
20

10
10

0.05
0.05

10
10

20
20

10
10

0.05
0.05

10
10

20
20

10
10

0.05
0.05

10
10

20
20

10
10

0.05
0.01

10

0.05

0.1

20

0.05
0.1
Time (s)

10

0.05

0.1

0.01

0.05

0.1

Phase B

Phase C

Zero sequence

10
0
10

20
0
20

10
0
10

Scale 1

Phase A

(b)

Zero sequence
0.05

10
0
10

20
0
20

10
0
10

0.05
0.05
0
0.05

Scale 2

Phase C
10

10
0
10

20
0
20

10
0
10

0.05
0
0.05

Scale 3

Phase B
20

10
0
10

20
0
20

10
0
10

0.05
0
0.05

Scale 4

Phase A
10

10
0
10

20
0
20

10
0
10

0.05
0
0.05

Scale 5

Scale 5

Scale 4

Scale 3

Scale 2

Scale 1

(a)

10
0
10

20
0
20

10
0
10

0.02
0
0.02

0.05

0.1

0.05 0.1
Time (s)

0.05
0

0.05

0.1

0.05

0.1

Fig. 15. DWT of differential currents for inrush current case (inception angle of 120 ). (a) The 100% and (b) 120% of magnetizing fluxes
The ratio is calculated for all three phases, and from a trial
and error process the indices for the separation among the internal
fault condition, the external fault condition, and inrush condition
are defined. The most appropriate form of the decision algorithm
from the case studies of the system under the investigations can
be given as a flowchart illustrated in Fig. 16.
By considering Fig. 16, in the first step [19,20], the coefficients
of the signals obtained from the DWT are used to discriminate

between the normal condition and the fault condition including

the inception of fault (time at which the fault occurs). It can be
seen that if the first step can detect the fault condition, the next
step will be carried out to discriminate between the internal fault
condition and the external fault condition, as shown in Fig. 16.
In the final step, Phmax and Phsum are performed to distinguish
between a fault that occurs at high voltage and that at low voltage.
By observing Fig. 16, if Phmax is greater than 40 and Phsum is
309

IEEJ Trans 9: 302314 (2014)

A. NGAOPITAKKUL AND C. JETTANASEN

Fault detection decision algorithm and

fault inception angle [19, 20]

(Phmax > 15)

And
(Phsum > 35)

(Phmax > 40)

And
(Phsum > 80)

No

Zchk > 5

Yes

No

Yes

Yes
Zchk < 0.5

Winding to ground
fault at high
voltage winding

Yes

No

Interturn fault at
high voltage
winding

No

Interturn fault at low

voltage winding

(Phmax > 15)

And
(Phsum > 15)

Zchk > 0.2

No

Protective relay must be operated as

tripping

(Phmax > 0.3)

And
(Phsum > 0.5)

No

Yes

Zchk > 1

Yes

Winding to ground
fault at low
voltage winding

Yes

No

External fault at
high voltage
side

External fault at
low voltage
side

No

Protective relay must be operated as block or

untripping

Inrush current
condition

Fig. 16. Flowchart for decision algorithm

(a)
0

200

10

0.01

0.02

0.03

0.04

0.05
Detail

0.06

0.07

0.08

0.09

0.1

4
Scale 1

Scale 1

0.01
0.005
0

0.01

0.02

0.03

0.04

0.05
0.06
Approximation

0.07

0.08

0.09

Scale 5

Scale 5

0.01

0.02

0.03

0.04

0.05

0.06

0.07

0.08

0.09

0.1

0.04

0.01

0.02

0.03

0.04

0.01

0.02

0.03

0.04

0
5

0.04

0.045

0.05
Detail

0.06

0.07

0.08

0.09

0.1

0.05
0.06
Approximation

0.07

0.08

0.09

0.1

0.07

0.08

0.09

0.1

0.05

0.06

(x= -58.64
y=0.0450)

(x= 51.82
y= 0.0423)

200
0.0350

0.03

0
10

0.02

10

200

0
0.01
x 104

2
0

0.1

200

200

Differential

(b) 10

Differential
200

0.05

0.035

(x= 0.065
y= 0.0400)

(x= 4.72
y=0.0417)
0.04

0.045

0.05

Time (s)

Time (s)

Fig. 17. Responses to the winding-to-ground fault condition. Fault occurring at (a) high- and (b) low-voltage winding
greater than 80, then internal fault can occur at the high-voltage
winding. The differential relay must activate the trip circuits of
circuit breakers for isolating the faulted components from the rest
of the power system. However, the decision algorithm cannot

identify the type of fault (winding-to-ground fault or interturn

fault), so Zchk is performed to classify the type of fault (windingto-ground fault or interturn fault). This indicates a significant
difference between winding-to-ground fault and interturn fault. If
310

IEEJ Trans 9: 302314 (2014)

DISCRIMINATING AMONG INRUSH CURRENT, EXTERNAL FAULT, AND INTERNAL FAULT IN POWER TRANSFORMERS

(a)

Differential
0

200

50

0.01

0.02

0.03

0.04

0.06

0.07

0.08

0.09

0.1

0.02
0

0.01

0.02

0.03

0.04
0.05
0.06
Approximation

0.07

0.08

0.09

0.1

0.01

0.02

0.03

0.04

0.05

0.06

0.07

0.08

200

0.04

0.03

0.04

0.09

0.1

0.045

0.05
Detail

0.06

0.07

0.08

0.09

0.1

0.01

0.02

0.03

0.04
0.05
0.06
Approximation

0.07

0.08

0.09

0.1

0.01

0.02

0.03

0.04

0.07

0.08

0.09

0.1

0
50

0.05

0.06

40

(x= 137.57
y= 0.042)

0 (x= -0.85
y= 0.040)
200
0.035

0.02

50

0
200

0
0.01
x 103

0.5

Scale 5

200

Scale 5

0.05
Detail

Scale 1

Scale 1

0.04

Differential

(b) 50

200

(x= 27.27
y=0.0449)

0 (x= 0.034
y= 0.0400)
40
0.035

0.05

Time (s)

0.04

0.045

0.05

Time (s)

Fig. 18. Responses to the interturn fault condition. (a) Fault occurring at high- and (b) low-voltage winding
Differential

(a) 10

10

0.5

0.01

0.02

0.03

0.04

0.06

0.07

0.08

0.09

0.1

0.01

0.02

0.03

0.04
0.05
0.06
Approximation

0.07

0.08

0.09

0.01

0.02

0.03

0.04

0.05

0.06

0.07

0.08

0.09

(x= -0.169
y= 0.0400)

0.03

0.04

0.06

0.07

0.08

0.09

0.1

0.01

0.02

0.03

0.04
0.05
0.06
Approximation

0.07

0.08

0.09

0.1

0.04

0.045

0.05

0.05
Detail

0.01

0.02

0.03

0.04

0.07

0.08

0.09

0.1

0.2

0.2
(x= 0.11
0.1
y= 0.0417)
0
0.1
0.035

(x= 0.2017
y= 0.0401)

0.5
0.035

0.1

0.5
0

0.02

0.2

0
0.5

0.01

0.05

0.1
Scale 5

0.5

0.1

Scale 1

Scale 1

0.05
Detail

0.5
0

Differential

0.5

Scale 5

(b)

0.05

0.06

(x= 0.01
y=0.0443)
0.04

0.045

0.05

Time (s)

Time (s)

Fig. 19. Responses to the external fault condition. Fault occurring at (a) high- and (b) low-voltage sides
(a)

(b)

Differential
10

0
10

10
0

0.01

0.02

0.03

0.04

0.06

0.07

0.08

0.09

0.1

0.02
0

0.01

0.02

0.03

0.04
0.05
0.06
Approximation

0.07

0.08

0.09

0.1

0
10

0.01

0.02

0.03

0.04

0.05

0.06

0.07

0.08

0.09

0.01

0.02

0.03

0.04

0.06

0.07

0.08

0.09

0.1

0.05
Detail

0.01

0.02

0.03

0.04
0.05
0.06
Approximation

0.07

0.08

0.09

0.1

0.01

0.02

0.03

0.04

0.07

0.08

0.09

0.1

0.02
0
10
0
10

0.05

0.06

0.1

(x= 0.054
0 y= 0.0405)
1
0.035

0.04

Scale 5

10

Scale 5

0.05
Detail

Scale 1

Scale 1

0.04

Differential

10

(x= 0.55
0 y= 0.0408)

(x= -0.045
y=0.0406)
0.04

Time (s)

0.045

(x= -0.476
y=0.0406)

0.035

0.050

0.04

0.045

0.050

Time (s)

Fig. 20. Responses to the magnetizing inrush current condition. (a) The 100% and (b) 120% of magnetizing fluxes
considered, it can be observed that after the fault occurrence at
0.04 s, both the coefficient of detail (high-frequency components)
and approximation (low-frequency components) have a sudden
change compared to those before the occurrence of a fault, so
the inception of fault can be detected. It is found, according
the data presented in Table V, that Phmax , Phsum , and Zchk
are calculated and the all comparison indicators (Phmax , Phsum ,
and Zchk ) have a value greater than 80. This indicates that
Phmax , Phsum , and Zchk play an important role in the decision
algorithm.
Based on the data in Table VI, considering the interturn fault
at the high-voltage winding, it can be clearly seen that Phmax and
Phsum have still a value greater than 80 but Zchk has a value lower
than 5; this is due to the effect of the interturn fault. Based on a

Zchk is greater than 5, then it can be identified as a winding-toground fault at the high-voltage winding. On the other hand, an
interturn fault will be identified if Zchk is less than 5. Similarly, for
the internal fault occurring at low voltage, the decision algorithm
has a similar process.
To gain more understanding, one case of each type of faults
is considered as an example in Figs 1720. The proportion of
the spectral differential current signal, calculated between the
maximum and minimum value in a quarter cycle period of
analyzed data and the time deviation are shown in Figs 17 20.
The results illustrated in Tables VVIII are obtained from one
case of each type of faults as shown in Figs 17 20.
By observing Fig. 17(a) and the data in Table V, when the fault
occurring at 20% of the length of the high-voltage winding is
311

IEEJ Trans 9: 302314 (2014)

A. NGAOPITAKKUL AND C. JETTANASEN

Table V. Result of identification for case of winding-to-ground fault

High-voltage winding
Coefficient of DWT
Xmax diff
Xmin diff
tmax diff_x
tmin diff_x
Xchkdiff
Phsum
Phmax
Zchk
Result

51.8227
58.6477
0.0412
0.0450
118.945

Low-voltage winding

0.0592
12.8956
0.0722
14.7771
0.0449
0.0412
0.0400
0.0450
0.1059
29.7956
148.8469
118.945

Tripped

Zero

37.3422
42.3332
0.0412
0.0450
85.7877

85.7877

4.7263
0.0651
0.0417
0.0400
10.5936

0.0611
0.0075
0.0609
4.6632
0.0449
0.0400
0.0417
0.0419
0.1525
9.7306
20.4767
10.5936

Tripped

Zero
0.0134
0.0076
0.0400
0.0406
0.1312

0.1312

Table VI. Result of identification for case of interturn fault

High-voltage winding
Coefficient of DWT
Xmax diff
Xmin diff
tmax diff_x
tmin diff_x
diff
Xchk
Phsum
Phmax
Zchk
Result

Low-voltage winding

Zero

Zero

137.57
0.856
42.25
40.01
247.19

0.0633
0.0524
44.17
40.17
0.12
494.61
247.31

0.8846
137.60
40.01
42.25
247.31

0.00220
0.000861
40.33
40.65
0.04

0.04

27.28
0.034
44.49
40.01
24.33

0.0632
0.0699
44.97
42.09
0.18
48.96
24.45

0.0441
27.34
40.01
44.49
24.45

0.00160
0.00670
40.01
40.65
0.14

0.14

Tripped

Tripped

Table VII. Result of identification for case of external fault

High-voltage winding
Coefficient of DWT
Xmax diff
Xmin diff
tmax diff_x
tmin diff_x
diff
Xchk
Phsum
Phmax
Zchk
Result

A
0.2017
0.169
0.0401
0.0400
10.5914

Low-voltage winding

0.0645
0.012
0.0609
0.1111
0.0444
0.0403
0.0400
0.0406
0.1124
1.2387
11.9426
10.5914

Untripped

Zero

0.0749
0.1755
0.0403
0.0400
3.3386

3.3386

0.1166
0.016
0.0417
0.0443
0.1571

0.0651
0.0255
0.0622
0.0842
0.0449
0.0400
0.0417
0.0449
0.1591
0.100
0.4971
0.1591

Untripped

Zero
0.0099
0.0078
0.0400
0.0406
0.0916

0.0916

Table VIII. Result of identification for case of inrush current

Inrush current 100%
Coefficient of DWT
Xmax diff
Xmin diff
tmax diff_x
tmin diff_x
diff
Xchk
Phsum
Phmax
Zchk
Result

A
0.5388
0.456
0.0404
0.0406
24.88

Inrush current 120%

1.3926
0.4249
0.0038
1.4191
0.0450
0.0406
0.0400
0.0450
1.118
1.693
27.692
24.88

Untripped

Zero

0.0006
0.0059
0.0419
0.0448
0.0090

0.0090

0.555
0.0476
0.0408
0.0406
25.76

0.928
0.4428
0.0044
0.9499
0.0450
0.0406
0.0400
0.0450
0.7467
1.279
27.79
25.76

Untripped

Zero
0.0008
0.0046
0.0409
0.0448
0.0097

0.0097

Table IX. Summary of results from all simulations

Internal fault
Fault types
Number of case studies
Detection accuracy

External faults

HV winding

LV winding

HV side

LV side

Inrush current

1620
100.00%

1620
97.90%

192
98.96%

192
97.92%

60
96.67%

312

IEEJ Trans 9: 302314 (2014)

DISCRIMINATING AMONG INRUSH CURRENT, EXTERNAL FAULT, AND INTERNAL FAULT IN POWER TRANSFORMERS

further analysis of Table VI, it can be concluded that the interturn

fault can be detected and the differential relay must be activated.
When all conditions as stated in Section 3 are applied, the total
number of case studies is 3240 for internal faults condition, 384 for
external faults condition, and 60 for inrush condition. The accuracy
of the proposed decision algorithm for all case studies is shown in
Table IX.

(10) Miyazaki S, Mizutani Y. Diagnosis of abnormality of transformer

winding by frequency response analysis. IEEJ Transactions on Power
and Energy 2010; 130(6):547550.
(11) Mokryani G, Haghifam MR, Latafat H, Aliparast P, Abdollahy A.
Detection of inrush current based on wavelet transform and LVQ
neural network. IEEE PES Transmission and Distribution Conference
and Exposition 2010; 1:15.
(12) Perera N, Rajapakse AD. Development and Hardware Implementation of a fault transients recognition system. IEEE Transactions on
Power Delivery 2012; 27(1):400412.
(13) Oliveira MO, Ferreira GD, Garcia FH, Bretas AS, Perrone OE,
Reversat JH. Adaptive differential protection for power transformer
based on transient signal analysis. IEEE Power and Energy Society
General Meeting 2012; 1:17.
(14) Jamali M, Mirzaie M, Asghar Gholamian S, Mahmodi Cherati S.
A wavelet-based technique for discrimination of inrush currents
from faults in transformers coupled with finite element method.
IEEE Applied Power Electronics Colloquium (IAPEC2011) 2011;
1:138142.
(15) Gong M, Zhang X, Gong Z, Xia W, Wu J, Lv C. Study on a new
method to identify inrush current of transformer based on wavelet
neural network. International Conference on Electrical and Control
Engineering (ICECE2011) 2011; 1:848852.
(16) Zheng T, Gu J, Huang SF, Guo F, Terzija V. A new algorithm to
avoid maloperation of transformer differential protection in substations with an inner bridge connection. IEEE Transactions on Power
Delivery 2012; 27(3):11781185.
(17) Bing-da Z, Shuo Z. Recognizing the transformer inrush currents by
FPGA based on energy components. Spring Congress on Engineering
and Technology (S-CET2012) 2012; 1:14.
(18) Kunakorn A. Applications of discrete wavelet transform for transformer inrush current detection in protective control scheme. IEEE
International Symposium on Communications and Information Technology 2004; 2:871874.
(19) Ngaopitakkul A, Kunakorn A, Ngamroo I. Discrimination between
external short circuits and internal faults in transformer windings
using discrete wavelet transforms. 40th Annual Conference of the
IEEE Industries Application Society, 2005; 448452.
(20) Jettanasen C, Ngaopitakkul A. The spectrum comparison technique
of DWT for discrimination between external fault and internal faults
in power transformer. Journal of International Council on Electrical
Engineering 2012; 2(3):302308.
(21) Long Y, Jingdong N. A wavelet transform based discrimination
between internal faults and inrush currents in power transformer.
International Conference on Electric Information and Control Engineering (ICEICE2011), 2011; 11271129.
(22) Yang C, Yongchao P, Li P. The application of wavelet analysis in identifying transformer magnetizing inrush current. International Conference on Electronics Communications and Control
(ICECC2011), 2011; 39883991.
(23) Abniki H, Monsef H, Nabavi-Razavi MT, Ebrahimi Gardeshi M.
Differential protection improvement of power transformers using
wavelet analysis of power system frequency. 16th International
Conference on Intelligent System Application to Power Systems
(ISAP2011), 2011; 16.
(24) Hooshyar A, Sanaye-Pasand M, Afsharnia S, Davarpanah M, Mahdi
Ebrahimi B. Time-domain analysis of differential power signal to
detect magnetizing inrush in power transformers. IEEE Trans. on
Power Delivery 2012; 27(3):13941404.
(25) Saleh SA, Aktaibi A, Ahshan R, Rahman MA. The development of a
d-q axis WPT-based digital protection for power transformers. IEEE
Transaction on Power Delivery 2012; 27(4):22552269.
(26) Aktaibi A, Rahman MA. A novel technique for differential protection for power transformers. International Conference on Electrical
Machines (ICEM), 2012; 22932298.
(27) Aktaibi A, Rahman MA, Razali AM. An experimental implementation of dq axis wavelet packet transform hybrid technique for
3 power transformer protection, IEEE Industry Application Society
Annual Meeting (IAS), 2012; 22932298.
(28) Saleh SA, Scaplen Benjamin, Azizur Rahman M. A new implementation method of wavelet packet transform differential protection
for power transformers. IEEE Transactions on Industry Applications
2011; 47(2):10031012.

5. Conclusions
This paper proposed a technique for discriminating between
inrush current, external fault, and internal fault. The DWT algorithm was employed in the protection scheme. The simulations, analysis, and diagnosis were performed using ATP/EMTP
and MATLAB/Simulink. The current waveforms obtained from
ATP/EMTP were extracted to several scales with the DWT.
Daubechies4 (db4) was employed as the mother wavelet in order
to decompose the low-frequency components from fault signals.
The ratio between per-unit differential current and per-unit time
was calculated and used as comparative indicator in order to discriminate between inrush current, external fault, and internal fault.
The results obtained from the algorithm proposed in this paper can
detect and indicate the fault condition with an accuracy better than
95% as seen from Table IX. In addition, the proposed technique
uses data of the differential current with a time of a quarter cycle
for the analysis, which is less than that employed in a conventional
protection scheme.

Acknowledgments
The work presented in this paper is part of a research project sponsored
by the King Mongkuts Institute of Technology Ladkrabang Research Fund.
The authors would like to thank the sponsor for financial support.

References
(1) Huang S-R, Chen H-T, Wu C-C, Guan C-Y, Cheng C. Distinguishing internal winding faults from inrush current in power transformers using Jiles-Atherton model parameters based on correlation coefficient. IEEE Transactions on Power Delivery 2012; 27(2):
548553.
(2) Kang YC, Lee BE, Zheng TY, Kim YH, Crossley PA. Protection,
faulted phase and winding identification for the three-winding transformer using the increments of flux linkages. IET Generation, Transmission & Distribution 2010; 4(9):10601068.
(3) Lu Z, Tang WH, Ji TY, Wu QH. A morphological scheme for inrush
identification in transformer protection. IEEE Transactions on Power
Delivery 2009; 24(2):560568.
(4) Ma J, Wang Z, Yang Q, Liu Y. Identifying transformer inrush current
based on normalized grille curve. IEEE Transactions on Power
Delivery 2011; 26(2):588595.
(5) Lin JH, Zeng L, Bo ZQ. Analysis of electromagnetic transient and
adaptability of second-harmonic restraint based differential protection
of UHV power transformer. IEEE Transactions on Power Delivery
2010; 25(4):22992307.
(6) Perera N, Rajapakse AD. Recognition of fault transients using a
probabilistic neural-network classifier. IEEE Transactions on Power
Delivery 2011; 26(1):410419.
(7) Ashrafian A, Rostami M, Gharehpetian GB. Hyperbolic S-transformbased method for classification of external faults, incipient faults,
inrush currents and internal faults in power transformers. IET
Generation, Transmission & Distribution 2012; 6(10):940950.
(8) Kitayama M. A wavelet-based fast discrimination of transformer
magnetizing inrush current. IEEJ Transactions on Power and Energy
2006; 126(1):513.
(9) Sano T, Miyagi K. Experimental investigation on FRA diagnosis of
transformer faults. IEEJ Transactions on Power and Energy 2007;
127(7):791796.
313

IEEJ Trans 9: 302314 (2014)

A. NGAOPITAKKUL AND C. JETTANASEN

(41) Shah AM, Bhalja BR. Discrimination between internal faults and
other disturbances in transformer using the support vector machinebased protection scheme. IEEE Transactions on Power Delivery
2013; 28(3):15081515.
(42) Hamilton R. Analysis of transformer inrush current and comparison
of harmonic restraint methods in transformer protection. IEEE
Transactions on Industry Applications 2013; 49(4):18901899.
(43) Bastard P, Bertrand P, Meunier M. A transformer model for winding fault studies. IEEE Transactions on Power Delivery 1994;
9(2):690699.
(44) Islam SM, Ledwich G. Locating transformer faults through sensitivity analysis of high frequency modeling using transfer function
approach. IEEE International Symposium on Electrical Insulation,
1996; 3841.
(45) IEEE Working Group 15.08.09. Modeling and Analysis of System
Transients using Digital Programs, IEEE PES special publication.
(46) ABB Thailand, Test Report No. 56039.
(47) Transmission System Operation Planning Department. Switching and
Transmission Line Diagram, Electricity Generation Authorisation
Thailand (EGAT), 2010.

(29) Tsourakis G, Milis I, Prousalidis J. Transformer sympathetic inrush:

a case study. 8th Mediterranean Conference on Power Generation,
Transmission, Distribution and Energy Conversion (MEDPOWER
2012), 2012; 15.
(30) Elsamahy M, Babiy M. An intelligent approach using SVM to
enhance turn-to-turn fault detection in power transformer. IEEE Electrical Power and Energy Conference (EPEC2012), 2012; 255260.
(31) Moscoso M, Lloyd GJ, Liu K, Wang Z. Improvements to transformer
differential protectiondesign and test experience. 47th International Universities Power Engineering Conference (UPEC), 2012;
16.
(32) Adel Aktaibi, Rahman MA. A new hybrid technique of control
and protection for three phase power transformer. 38 th Annual
Conference on IEEE Industrial Electronics Society (IECON2012),
2012; 11861191.
(33) Oliveira LMR, Cardoso AJM. Extended Parks vector approachbased differential protection of three-phase power transformer. IET
Electric Power Applications 2012; 6(8):463472.
(34) de Leon Francisco, Farazmand Ashkan, Joseph Pekir. Comparing
the T and equivalent circuits for the calculation of transformer inrush currents. IEEE Transactions on Power Delivery 2012;
27(4):23902398.
(35) Bejmert D, Rebizant W, Schiel L, Staszewski L. A new multi-criteria
fuzzy logic transformer inrush restraint algorithm. 11th International
Conference on Developments in Power System Protection (DPSP),
2012; 16.
(36) Verney LH, Rolim JG. Assessment of transformer relay algorithms
and settings in laboratory using transient simulation data. 11th International Conference on Developments in Power System Protection
(DPSP), 2012; 16.
(37) Behjat V, Vahedi A, Setayeshmehr A, Borsi H, Gockenbach E.
Diagosing shorted turns on the windings of power transformers based
upon online FRA using capacitive and inductive couplings. IEEE
Transactions on Power Delivery 2011; 26(4):21232133.
(38) Sridevi T, Ramesh Reddy K, Leela Jaya Syamala N. Harmonic
analysis of inrush current using fast Fourier transform. International
Conference on Power, Energy and Control (ICPEC2013), 2013;
520524.
(39) Zhang Wenkui, Tan Qian, Miao Shihong, Zhou Liangsong, Liu Pei.
Self-adaptive transformer differential protection. IET Generation,
Transmission & Distribution 2013; 7(1):6168.
(40) Cunziang Y, Hao B. The fault diagnosis of transformer Based on
the SOM neural network. 5th International Conference on Measuring
Technology and Mechatronics Automation (ICMTMA2013), 2013;
11781180.

Atthapol Ngaopitakkul (Non-member) is currently an Assistant Professor with the Department of Electrical
Engineering, King Mongkuts Institute of Technology Laddrabang, Bangkok, Thailand. His
research interests include protection relays and
energy management.

Chaiyan Jettanasen (Non-member) received the B.Eng. and

M.Eng. degrees from the Institut National des
Sciences Appliquees (INSA) de Lyon, France,
in 2005, and the Ph.D. degree from Ecole Centrale de Lyon, France, in 2008. He is currently an Assistant Professor with the Electrical Engineering Department, King Mongkuts
Institute of Technology Laddrabang, Bangkok,
Thailand. His research interests include EMC in power electronic
systems.

314

IEEJ Trans 9: 302314 (2014)