Electric Power Systems Research 155 (2018) 340–349

Contents lists available at ScienceDirect

Electric Power Systems Research

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

Waveform asymmetry of instantaneous current signal based

symmetrical fault detection during power swing
C. Lazaro a,∗ , J.P. Marques a , G. Marchesan b , G. Cardoso a
a
Federal University of Santa Maria, Av. Roraima, 1000 Santa Maria, RS, Brazil
b
Federal University of Pampa, Bagé, RS, Brazil

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

Article history: This paper proposes a methodology for detection of short-circuit during power swing with easy adjust-
Received 21 July 2017 ment, in order to unblock the distance relay and allow its operation. If a fault occurs during power swing,
Received in revised form 22 October 2017 it must be immediately detected and removed. One of the major challenges in these detectors is to dis-
Accepted 5 November 2017
tinguish power swing from symmetrical fault near to the electrical center (ı = 180◦ ), as both are balanced
phenomena. This work aims to propose a detector of fault which is reliable and capable of detecting a
Keywords:
short-circuit occurring in the moment that apparent impedance crosses near to electrical center. The
Fault detection during power swing
methodology proposed presents as main advantages, beside the ability to identify the three-phase faults
Power swing blocking
Unblock distance protection relay
during power swing near to the electrical center up to ı ≤ 120◦ , to work in low sampling frequency and
Transmission line protection obtain the shortest response time compared to other techniques. In order to unblock the relay in the case
of a short-circuit during power swing, the asymmetry coefficient of the current signal is calculated in
a window of 1 cycle (60 Hz). If the asymmetry coefficient surpasses the predefined threshold, the relay
is unblocked (reset of PSB function – Power Swing Blocking), allowing its operation. The methodology
uses predefined thresholds through the analysis of signals of different power swing frequencies. In this
work, the defined threshold is able to cover a great number of faults and to be tolerant to power swing
frequencies lower than 6 Hz. The predefined threshold avoids the need for extensive studies on stability
which estimate the behavior of power electrical system.
© 2017 Elsevier B.V. All rights reserved.

1. Introduction of major disturbances which caused blackout involve undue trip-

In steady state, a power balance between generation and load
is maintained. However, events such as short-circuits, switch of
transmission lines, disconnection of the generator, and loss or
application of large block of load may compromise this balance and
cause the phenomenon known as power swing, in electrical sys-
tems. A power swing may cause undue operations of distance relays
[1,2]; resulting in the shutdown of generators and consequent risk
of insufficient supply of the generated power, causing loss of bal-
ance and blackout. The stability of Electric Power System (EPS) has
become a main subject in the operation of a country’s intercon-
nected system. The North American Electric Reliability Corporation
(NERC) reported an alarming statistics where an important portion
∗ Corresponding author.
E-mail addresses: carlos.lazaro@mail.ufsm.br (C. Lazaro),
jefersonjpm@mail.ufsm.br (J.P. Marques), gustavomarchesan@unipampa.edu.br
(G. Marchesan), ghendy@ufsm.br (G. Cardoso).
ping of distance relays [3–8]. In order to overcome this, researchers
have developed a wide range of blocking techniques in recent years.
Techniques for blocking against power swing have been developed
(PSB – Power Swing Blocking) in order to avoid these undue tripping
during stable swings and to unblock short-circuits during power
swing. The detection of symmetrical fault during power swing has
been a challenge, due to the fact that power swing and symmetrical
fault are balanced phenomena, and therefore, block in this case.
The PSB techniques are divided into conventional and non-
conventional methods [8–10]. The conventional methods work
through concentric characteristics and/or through blinders plus an
adjustment of time in order to measure time in which the appar-
ent impedance goes through these characteristics. If the apparent
impedance crosses in a lower time than the time set, it is inter-
preted that the event is a fault. Otherwise, it is interpreted that the
event is a power swing, and then the PSB inhibits the tripping of
the distance relay. However, this technique has two disadvantages:
(i) it is incapable of detecting new events after the relay is inhib-
ited; (ii) the measurement of the time set and values of blinders

https://doi.org/10.1016/j.epsr.2017.11.005
0378-7796/© 2017 Elsevier B.V. All rights reserved.
 C. Lazaro et al. / Electric Power Systems Research 155 (2018) 340–349
Fig. 1. Signal processing.
are obtained through complex studies of the behavior of local grid,
which demand professionals in the subject [11,12].
The method based in the Swing Voltage Centre (SVC), which
uses the expression Vcos , where V is the magnitude of local voltage
measured and  is the angle between the voltage and current signals
locally measured. During the power swing, the SVC changes contin-
uously, however when a short-circuit happens the SVC remains the
same. This criterium is used to distinguish between a power swing
and a short-circuit [8,11,13]. Due to the fact that fault generates
electromagnetic transients, the Wavelet Transform (WT) demon-
strated to be efficient to the analysis of this transients in power
systems. Brahma [14] studied different sampling rate used by the
WT to detect power swing and symmetrical fault during the power
swing. The author claimed that the sampling rate of 40.96 kHz is
ideal for the detection of power swings and symmetrical fault in
any place of the electrical power system. The main disadvantage
of the WT is that most of the relays do not present sampling rate
so elevated. Despite the fact that techniques for detection of sym-
metrical fault during the power swing already exist [14–19]; the
reliability in the cases of power swing need improvements.
This work presents a method based on the identification of
short-circuit during the power swing through the monitoring of
asymmetry coefficient of current signal seen by the relay. The
advantages of the proposed method in relation to the mentioned
are: (i) it does not require to be set, avoiding the need for intensive
analysis on the electrical power system; (ii) it is able to distinguish
fault during stable power swing through a threshold (iii) it does not
need elevated sampling rate.
Section 2 presents a review of the concepts approached in this
work and explains the working of methodology proposed. Section
3 shows studies of case and results and, in Section 4 the conclusion
is presented.
2. Proposed methodology
This work is based on the assumption that the occurrence of
a short-circuit on a transmission line results in sudden change of
frequency components of current waveform. Here, a filter which
measures the asymmetry of data distribution in a window with
fixed length is used in order to detect a short-circuit and unblock
the relay during power swing condition. Fig. 1. shows the diagram of
signal processing through this filter, which has as input as the cur-
rent instantaneous signal, then, the absolute value is determined
and the asymmetry coefficient is calculated to the data group inside
the window (1 cycle of 60 Hz). Thus, through this filter of asymme-
try, the value of Fisher Asymmetry Coefficient (FAC) is obtained
and used for the detection of a short-circuit during power swing.
The filter of asymmetry proposed has the function of detecting a
short-circuit through a one threshold, based on the value of the
asymmetry coefficient of the current waveform.
2.1. Coefficient of asymmetry
The asymmetry coefficient is a measurement that indicates
symmetry of a distribution in relation to the mean data without
341
Fig. 2. Types of asymmetry for a distribution of data.
requiring a graphic representation. In this work, Fisher Asymmetry
Coefficient (FAC) is used, which indicates the change in the amount
of elements to the right and to the left in relation to the mean data.
There are three types of distribution curve according to its asymme-
try: Negative asymmetry, symmetry and positive asymmetry. Fig. 2
shows distribution behavior for each type. Among the conditions
to classify the type of asymmetry are: If FAC < 0, distribution has
negative asymmetry and flattens to higher values in relation to the
mean. If FAC = 0, distribution is symmetrical, in other words, data
distribution fulfills the condition of normal distribution. If FAC > 0,
distribution has positive asymmetry and flattens to lower values in
relation to the mean. FAC evaluates proximity of data to its mean.

N
3
When higher the sum (xi − x̄) , greater the asymmetry will
i=1
be. Considering the distribution x = (x1 , x2 , . . ., xN ), Fisher Asymme-
try equation is:
N 3
i=1 (xi
− x)
FAC = (1)
N 3
where
FAC: fisher asymmetry coefficient;
x̄: mean of data distribution;
: standard deviation of data distribution;
N: length of data distribution.
In this work, the asymmetry coefficient is employed applied to
a data window of 1 cycle (60 Hz) sampled to 64 samples per cycle
(spc); this variable is represented in this work by the abbreviation
FAC. This sampling rate is subject to characteristics of the protection
relay analyzed.
2.2. Decision criterion
The voltage and current signals measured for each phase and the
output of the asymmetry filter are shown in Fig. 3(a)–(c), respec-
tively. A differentiated increase in the amplitude value is evidenced
in the output of the asymmetry filter when short-circuit is applied
in 3 s, as shown in Fig. 3(c). The filter is able to provide a response
with a differentiated value in the occurrence of short-circuit com-
pared to a power swing. As for faults during low and/or high power
swing frequencies, the distinction of short-circuit through the filter
is evident, providing well-established decision making and poste-
rior reset of blocking against power swing in protection relays. The
proposed methodology aims to help the protection relay to reset
PSB function, in the occurrence of fault during power swing. For
the output signal is applied a 2nd order low-pass filter with cutoff
frequency of 35 Hz in order to obtain a cleaner response. This value
was also moved to 0.5 in vertical axis considering as reference the
FAC value in normal operation.
342 C. Lazaro et al. / Electric Power Systems Research 155 (2018) 340–349

Fig. 3. (a) Voltages waveform during power swing; (b) current signals for symmetrical three-phase fault during power swing; (c) asymmetry filter response.

Table 1
Maximum values of FAC registered for different frequencies.

Load angle (◦ ) fslip (Hz) FAC

60 0 0.0047
60 0.5 0.0065
60 1 0.0673
60 2 0.1866
60 3 0.3059
60 4 0.4125
60 5 0.4791
60 6 0.5497
60 7 0.6671
60 8 0.6876
60 9 0.7215

Fig. 4. (a) Measured current by distance relay; (b) asymmetry filter response.

In steady state, FAC has values near to −0.5, referring to a nega-

tive asymmetry coefficient. This is due to the fact the filter’s input
signal has a different data distribution from a normal distribution
condition, in this case, the signal has a similar distribution to a sine
waveform. Fig. 4(a) and (b) show filter input and output proposed
respectively. Finally, an improved view of the FAC values after low-
pass filter is represented by line blue shown in Fig. 4(b). It is possible
to notice that FAC has values near from to zero until t < 1 s (steady
state). In t > 1 s, FAC expresses significant change by obtaining val-
ues higher than zero, accusing the presence of short-circuit. The
short-circuit turns off in t = 1.2 s, presenting slight variation in FAC
values due to the electromagnetic transients. As for t > 1.6 s, the EPS
presents low swing frequencies resulting in FAC slight variation,
FAC value will depend of swing frequency present in EPS.
Table 1 presents the maximum values of FAC registered to dif-
ferent power swing frequencies. In order to obtain these values, a
test system was implemented with two sources connected through
a transmission line [8], where its frequencies were varied and a
load angle of 60◦ was defined. Fig. 5 shows FAC behavior for power
swings of 0 Hz, 1 Hz, 3 Hz, 6 Hz, and 7 Hz. It is possible to observe
that the FAC reaches values lower than 0.0673 for frequencies of
power swing lower than 1 Hz. In the cases of 3 Hz to 6 Hz, the FAC
reaches values lower than 0.5497. For 7 Hz, FAC reaches values of Fig. 5. (a) FAC for fslip of 0 Hz; (b) FAC for fslip of 1 Hz; (c) FAC for fslip of 3 Hz; (d)
FAC for fslip of 6 Hz; (e) FAC for fslip of 7 Hz.
0.6671.
 C. Lazaro et al. / Electric Power Systems Research 155 (2018) 340–349 343

Start

IA Sampling
IB @ 1
IC 64apc

N
Is PSB active? 2

Y
Proposed Methodology

Compute asymmetry
3 Fig. 7. Western Systems Coordinating Council (WSCC) 9-bus test system.
filter for IA, IB and IC

N
FACA or FACB or 4
FACC > FAC_set

Y
Fault detected 5
Keep PSB reset for a
defined time

End
Fig. 8. Stable power swing – asymmetry filter response.

Fig. 6. Flowchart of proposed methodology.

In order to define the threshold, it was considered a simple two-
source system (equivalent parameters from the IEEE-9 Bus System)
connected through of the line of transmission studied (lines 6–9
with 0.20631 + j0.8993 Ohm/km and a 100 km length), where the
Electrical Center (EC) crosses on the half of the LT when the phase
angle (␦) between the sources is 180◦ [8]. When ␦ is equal to 180◦ ,
the distance relay measure an impedance equal when a three-phase
short-circuit occurs. The threshold is based in the obtained FAC
when the angle ı with values of 30◦ < ı <180◦ changes for an angle
␦ equal to 180◦ . Thus, the threshold chosen must be tolerant to
swing frequencies up to 6 Hz and detect three-phase fault in the
most of cases. Hence, a threshold equal to 0.6 detects three-phase
fault when applied on an angle ı ≤ 120◦ and it is tolerant to the
presence of swing frequencies until 6 Hz.
Fig. 6 shows a flowchart used by methodology to achieve its
objective of unblock the distance relay in presence of a short-circuit
during power swing. In block 1, the protection relay samples the
voltage and current signals to 64 spc; if there is power swing, the
relay activates its PSB function (blocking the relay). In block 2, the
methodology applies the asymmetry filter over the current sig-
nal for phases A, B and C (block 3). When the FAC for each phase
is obtained, the values are compared to the FAC set (block 4). If
the sentence of block 4 is satisfied, the method will maintain the
block by PSB on reset for a defined time (block 5); this time, should
guarantee the operation of main and backup protection over the
transmission line in presence of fault. Otherwise, the relay will
remain blocked.
3. Results and discussions
In order to evaluate the proposed method, IEEE9-Bus system
shown in Fig. 7 is used. The settings used for IEEE9-Bus system
were obtained in [20]. For the simulations, the voltage and speed
regulators were implemented for the generators involved. The pro-
gram DigSilent Power Factory was used to simulate the electrical
power system and MATLAB to analyze the obtained data. The cur-
rent and voltage instantaneous signals are sampled to 64 spc by the
distance relay studied, R96, as shown in Fig. 7.
A Current Transformer (CT) is used with relation of 500/5A and
Voltage Transformer (VT) with relation of 230/0.12 kV. The set-
tings of areas 1, 2 and 3 are 0.8*ZL69, 1.2*ZL69 and ZL69 + 0.5*ZL46,
respectively. The distance relay R96 is allocated on the transmis-
sion line connected between the bars 6 and 9. The R96 relay is a
great choice to be analyzed during power swing phenomena, as it is
located near to the electrical center of the power system, where the
methods are not usually able to distinguish between short-circuits
and power swing. For validating the performance of methodology
proposed, the following cases were analyzed.
3.1. Stable power swing
The event of stable power swing is caused by the application
of a short-circuit at the middle of the transmission line connected
between bus 4 and 5, in t = 1 s. The duration of the fault is 150 ms,
isolated by the switches CB1 and CB2. Fig. 8 shows the current
waveform and the resulting asymmetry coefficient under stable
344 C. Lazaro et al. / Electric Power Systems Research 155 (2018) 340–349
Fig. 9. R–X diagram for stable power swing.
Fig. 10. Unstable power swing – asymmetry filter response.
power swing conditions. It is possible to observe that in the moment
of short-circuit, the FAC magnitude surpasses the set predefined
of 0.6 detecting a short-circuit. This threshold is represented by
the non-continuous line. In this case study, the detection of short-
circuit is efficient, even being applied in the neighbor line. After the
fault is isolated, the electrical system swings until it is localized in
a new operation point. It is possible to observe that, by the begin-
ning of the simulation, the methodology inserted in the R96 relay
identifies the fault (fault occurs in 1 s) in adjacent line (line 4–5) in
6 ms. On the other hand, the opening of the switches causes asym-
metry however it does not surpasses the threshold. For the rest of
the simulation, the FAC value does not surpass the threshold; the
power swing softens with time, causing low swing frequencies.
Fig. 9 shows the trajectory of apparent impedance seen by the
R96 relay on bus 9 during the simulation. It is possible to observe
that the trajectory of impedance do not cross the operation areas
of relay and softens until it is located in a new point of operation,
characterizing, then, a stable power swing. Also, moments in which
the fault is applied and isolated are shown (line 4–5). In this case,
methodology detects effectively a short-circuit.
3.2. Unstable power swing
A short circuit at the middle of the transmission line connected
between buses 4 and 5 is applied in t = 1 s. The duration of the
fault is 210 ms, isolated by CB1 and CB2. Fig. 10 shows the current
waveform and the resulting asymmetry coefficient under unstable
Fig. 11. R–X diagram for unstable power swing.
Fig. 12. Short-circuit during stable power swing – asymmetry filter response.
power swing conditions. It is possible to observe that in the moment
of the short-circuit, the FAC magnitude surpasses the threshold,
detecting this event. After the fault is isolated, the electrical system
is unable to maintain the stability, losing synchronism. The current
waveform is submitted to an increased in the swing frequency caus-
ing its deformation. In the simulation, between 1.210 s and 2.4 s, the
FAC values do not surpass the threshold, being tolerant to a swing
frequency range, in this case, lower than 6 Hz.
Fig. 11 shows the trajectory of apparent impedance seen by the
R96 relay on bus 9. It is possible to observe that the trajectory of
impedance goes into the electrical center of power system, crossing
the operation areas of relay analyzed. The trajectory of apparent
impedance crosses the relay operation area until the left semiplane
of R axis, resulting in a reverse power flow. In this case, the electrical
system loses synchronism and does not return to a new operation
point, characterizing power swing as unstable.
3.3. Short-circuit during stable power swing
A short circuit at the middle of the transmission line connected
between buses 4 and 5 is applied in t = 1 s. The duration of the fault
is 100 ms, isolated by CB1 and CB 2. In t = 1.9 s, a short-circuit is
applied in 75% of the length of the line 6–9 under protection of
the R96 relay. The duration of the fault is 100 ms, next, the fault
is extinguished without opening of breakers. Fig. 12 shows the
current waveform and the resulting asymmetry coefficient for the
 C. Lazaro et al. / Electric Power Systems Research 155 (2018) 340–349
Fig. 13. R–X diagram.
case study. It is possible to observe that, the methodology identi-
fies the fault (fault occurs in 1 s) in line 4–5 in 7 ms. It is also is
detected a fault (fault occurs in 1.9 s) over the line 6–9 in 5.43 ms.
In 2 s, methodology perceives the moment in which the fault dis-
appears, due to an important asymmetry in current waveform in
that moment. In this case study, methodology detects effectively
the short-circuit during power swing.
3.4. Short-circuit near to electrical center during power swing
In this simulation, the performance of the methodology to detect
faults near to the electrical center is evaluated by analyzing of
few points located over the apparent impedance trajectory. Power
swing is created through a short-circuit at the middle of the line
4–5 applied in t = 1s. The duration of the fault is 209.9 ms, iso-
lated by CB1 and CB2. Fig. 13 shows the trajectory of apparent
impedance goes into the electrical center of power system, crossing
the operation areas of relay analyzed. The performance is evaluated
by the application of the symmetrical short-circuit in points ‘a’, ‘b’,
‘c’, ‘d’ and ‘e’ independently. The procedure used is the following:
the power swing begins in 1.2101 s, a short-circuit is applied on
the selected point and the FAC is monitored through its thresh-
old to detect fault. The same procedure is applied for each point
separately.
For point ‘a’, the maximum value of FAC when the fault is applied
(fault occurs in 1.31s) was 0.811, 0.399 and 0.621, for phases A, B
and C, respectively. The current waveform and FAC values obtained
for point ‘a’ are shown on Fig. 14(a) and (b) respectively. It is pos-
sible to observe that for phase B, the FAC value do not surpass its
Fig. 14. (a) Instantaneous current signal; (b) asymmetry filter response for symmetrical three-phase fault.
345
threshold due to slight variation in the form. For phases A and C,
FAC surpasses the threshold, detecting a fault. The fault is detected
in 22 ms.
Regarding point ‘b’, the maximum value of FAC registered when
the fault is applied (fault occurs in 1.41 s) was 0.426, 0.138 and
0.621, for phases A, B and C, respectively. The current waveform
and FAC values obtained for point ‘b’ are shown in Fig. 15(a) and
(b), respectively. It is possible to observe that phases A and B do
not surpass the threshold due to its slight change in form between
events. However, phase C is able to surpass the threshold, detecting
a fault. The fault is detected in 22.75 ms.
Concerning point ‘c’, the FAC values when the fault is applied
(fault occurs in 1.51 s) were values lower than 0.062 as shown in
Fig. 16(b). In Fig. 16(a) it is observed how the short-circuit hides
itself inside the power swing. Due to the slight change in current
waveform between events, the fault was not detected through the
given threshold. This point is considered as critical in scenarios of
fault during power swing, this due to the fact it could not differen-
tiate a symmetrical fault from power swing.
As for point ‘d’, the maximum value of FAC registered when the
fault is applied (fault occurs in 1.5775 s) was 0.7006, 0.274 and
0.314, for phases A, B and C, respectively. The current waveform
and FAC values obtained for point ‘d’ are shown on Fig. 17(a) and
(b) respectively. It is possible to observe that phases B and C do
not surpass the set due to its small change in form between events.
However, phase A can surpass the threshold detecting a fault. The
fault is detected in 17.5 ms.
In relation to point ‘e’, the maximum value of FAC registered
when the fault is applied (fault occurs in 1.6002 s) was 0.903, 0.872
and 0.116, for phases A, B and C respectively. Fig. 18(a) and (b)
present the current waveform and FAC values obtained for point
‘e’ respectively. It is possible to observe that for phase C, the FAC
value do not surpass the value of its threshold due to the slight
change in form. For phases A and B, FAC surpasses the threshold,
detecting a fault. The fault occurs is detected in 14.8 ms. Hence, with
the threshold proposed it is guaranteed that the major number of
faults will be detected in points near to the electrical center (critical
scenario).
3.5. Discussion of results
As shown on Table 2 several study cases were simulated for the
evaluation of the methodology proposed. The results show satis-
factory performance for detecting fault during power swing under
many simulation conditions. Among these conditions: (i) the dura-
tion of the short-circuit over the line 4–5 for generating power
swing was between 50 ms and 210 ms; (ii) the distances used for
the application of the fault over the line 6–9 were 5%, 25%, 47.5%,
50%, 75% and 95%; (iii) the duration of the symmetrical fault applied
over the line 6–9 for the cases from 1 to 20 was 100 ms, for the cases
from 21 to 25 it was 1s, as shown on Table 2. The results present
346 C. Lazaro et al. / Electric Power Systems Research 155 (2018) 340–349

Fig. 15. (a) Instantaneous current signal; (b) asymmetry filter response for symmetrical three-phase fault.

Fig. 16. (a) Instantaneous current signal; (b) asymmetry filter response for symmetrical three-phase fault.

Fig. 17. (a) Instantaneous current signal; (b) asymmetry filter response for symmetrical three-phase fault.

Fig. 18. (a) Instantaneous current signal; (b) asymmetry filter response for symmetrical three-phase fault.

detection times on the range of tenth of millisecond. The minimum
time employed on the detection was 3.18 ms, as the case study 1
presents, where the fault was applied in 1.9 s to 5% of the length of
line 6–9. For case study 5, where the fault was applied to 95% of the
length of line 6–9, the time detection was of 4.91 ms.
For the mentioned cases it is possible to observe a difference
of 1.73 ms in its times of detection, showing a minimum effect on
the detection time of any distance of fault on the line. As for the
cases covered, this difference of time was lower than 2.6 ms. On
the other hand, the maximum time employed on the detection was
of 22.75 ms as the case study 22 presents, where the fault is applied
in 1.41 s with duration of 1 s. This case study is part of the critical
scenarios studied in this work (fault occurs near to electrical cen-
ter). The times of detection result to be satisfactory for the objective
proposed. The methodology proposed may be employed as an algo-
rithm which helps to PSB function in allowing the operation or not
of the protection.
 C. Lazaro et al. / Electric Power Systems Research 155 (2018) 340–349 347

Table 2
Study cases.

Study case Line 4–5 fault Line 6–9 fault Fault type Fault resistance () Fault start time Fault final time Method detection Fault detected
duration (ms) location (%) (s) (s) time (ms) (Yes/No)

1 50 5 ABC 0.0 1.9 2 3.18 Yes

2 50 25 ABC 0.0 1.9 2 4.05 Yes
3 50 50 ABC 0.0 1.9 2 4.39 Yes
4 50 75 ABC 0.0 1.9 2 4.74 Yes
5 50 95 ABC 0.0 1.9 2 4.91 Yes
6 100 5 ABC 0.01 1.9 2 3.69 Yes
7 100 25 AB 0.0 1.9 2 4.57 Yes
8 100 50 AG 0.01 1.9 2 4.91 Yes
9 100 75 ABC 0.01 1.9 2 5.43 Yes
10 100 95 ABC 0.01 1.9 2 5.78 Yes
11 150 5 AC 0.01 1.9 2 3.87 Yes
12 150 25 ABG 0.01 1.9 2 4.74 Yes
13 150 50 AC 0.1 1.9 2 5.43 Yes
14 150 75 BC 0.1 1.9 2 5.95 Yes
15 150 95 ABC 0.1 1.9 2 6.48 Yes
16 200 5 CG 0.01 1.9 2 3.52 Yes
17 200 25 BC 0.1 1.9 2 4.39 Yes
18 200 50 ABC 0.01 1.9 2 4.74 Yes
19 200 75 BG 0.1 1.9 2 5.09 Yes
20 200 95 AC 0.1 1.9 2 5.61 Yes
21 210 47.5 ABC 0.0 1.31 2.31 22.00 Yes
22 210 47.5 ABC 0.0 1.41 2.41 22.75 Yes
23 210 47.5 ABC 0.0 1.51 2.51 X No
24 210 47.5 ABC 0.0 1.5775 2.5775 17.50 Yes
25 210 47.5 ABC 0.0 1.6002 2.6002 14.80 Yes

In order to validate the proposed method several simulations

were made, where parameters of fault location, fault resistance and
sampling rate were varied, as shown in Table 3. The power swing
phenomena was created by a three-phase fault applied on 50% of
the line 4–5 with a duration of 50 ms and 100 ms. Fault resistances
of 0.1 Ohm to 200 Ohm were used. Fault locations of 5%, 50% and 95%
were applied on line 6–9. The results show similar detection times
for three-phase short-circuits with fault resistance of 0.1 Ohm to
5 Ohm. In cases with fault resistance of 10 Ohm and 20 Ohm were
observed similar detection times. Analyzing the cases 41–47, we
can observe that the proposed method detects three-phase fault
during power swing with detection times of 3.64–6.51 ms for fault
resistance of 0.1–200 Ohm. The difference in detection times is
lower than 2.88 ms for those cases. In cases with single and double
phase to ground, the method presents satisfactory results with a
maximum detection time of 11.45 ms in all cases tested. Fig. 19. Response of detection average time of all techniques.
The proposed method was investigated for different sampling
rate. In this paper, the tested cases of 26 to 69 were analyzed for 64
Table 4 presents the main characteristics of recent techniques
spc, 32 spc and 16 spc, as shown in Table 3. The detection time dif-
[14–19]. The proposed method has simple functions without hard-
ference among then was below to 4 ms. Comparing the FAC values
to-design parameters and, it results in a low computational burden.
in 64 spc (reference) with 32 spc and 16 spc for all tested cases, the
Still, it does not suffer any influence fault resistance and it works in
obtained average error was 4.32% and 19.15%, respectively. Hence,
a sampling rate used by commercial protection devices.
it is recommended use the proposed method with 32 spc, 64 spc or
higher sampling rate.
4. Conclusions
In order to validate the strength of the proposed method and the
selection of the threshold were tested over the transmission line
This paper presents a new method of fault detection during
(LT) values of X/R ratio from 4 to 20 (typical values of X/R ratio in
power swing based on FAC. The proposed analytical method is able
LT’s connected in voltages of 110–500 kV), voltages of the EPS from
to detect three-phase fault during power swing even those faults
110 kV to 500 kV and lengths of 50 km, 100 km and 300 km over
to close to electrical center. The simulations results show that the
the line 6–9. The tests showed that the threshold selected works
FAC threshold equal to 0.6 is suitable for transmission lines with X/R
for all the simulations in the characteristics previously mentioned,
range of 4 to 20. Unlike the other techniques, a high sampling rate is
validating the adjustment criterion of the proposed method.
not necessary in the proposed method. The main characteristics of
the method are: (i) is able to detect three-phase (3 ph) and double-
3.6. Qualitative analysis phase (2 ph) faults during power swing even when the transmission
line is operating with heavy loading (ı ≤ 120◦ for 3 ph and ı ≤ 80◦ for
Analyzing Fig. 19, it is possible to observe that, in general, the 2 ph); (ii) it works for high swing frequencies (up to 6 Hz). Finally,
proposed method provides a faster detection time of fault during the results show that the proposed method is efficient and can help
power swing when compared with other techniques. Moreover, in the improvement of PSB function built-in the protection relays
348 C. Lazaro et al. / Electric Power Systems Research 155 (2018) 340–349

Table 3
Study cases analyzed for different sampling rate.

Study case Line 4–5 fault Line 6–9 fault Fault type Fault resistance () Fault start Fault final Fault detected Method detection time (ms)
duration (ms) location (%) time (s) time (s) (Yes/No)

64 spc 32 spc 16 spc

26 50 5 ABC 0.1 1.9 2 Yes 3.12 4.68 4.16

27 50 5 ABC 1 1.9 2 Yes 3.12 4.68 4.16
28 50 5 ABC 5 1.9 2 Yes 3.12 4.16 4.16
29 50 5 ABC 10 1.9 2 Yes 3.38 4.16 4.16
30 50 5 ABC 20 1.9 2 Yes 3.38 4.68 4.16
31 50 50 ABC 0.1 1.9 2 Yes 4.42 4.68 4.16
32 50 50 ABC 1 1.9 2 Yes 4.42 4.68 4.16
33 50 50 ABC 5 1.9 2 Yes 4.42 4.68 4.16
34 50 50 ABC 10 1.9 2 Yes 4.68 4.68 4.16
35 50 50 ABC 20 1.9 2 Yes 4.68 4.68 4.16
36 50 95 ABC 0.1 1.9 2 Yes 4.94 5.20 5.20
37 50 95 ABC 1 1.9 2 Yes 5.20 5.20 5.20
38 50 95 ABC 5 1.9 2 Yes 5.20 5.20 5.20
39 50 95 ABC 10 1.9 2 Yes 5.20 5.20 5.20
40 50 95 ABC 20 1.9 2 Yes 5.20 5.20 5.20
41 100 5 ABC 0.1 1.9 2 Yes 3.64 5.20 5.20
42 100 5 ABC 1 1.9 2 Yes 3.64 4.68 5.20
43 100 5 ABC 5 1.9 2 Yes 3.64 4.68 5.20
44 100 5 ABC 10 1.9 2 Yes 3.90 4.68 5.20
45 100 5 ABC 20 1.9 2 Yes 3.90 4.68 5.20
46 100 5 ABC 100 1.9 2 Yes 5.20 5.72 6.25
47 100 5 ABC 200 1.9 2 Yes 6.51 6.77 7.29
48 100 50 ABC 1 1.9 2 Yes 4.94 5.20 5.20
49 100 50 ABC 5 1.9 2 Yes 5.20 5.20 5.20
50 100 50 ABC 10 1.9 2 Yes 5.20 5.20 6.25
51 100 50 ABC 20 1.9 2 Yes 5.20 5.20 6.25
52 100 50 ABC 100 1.9 2 Yes 6.77 6.77 7.29
53 100 50 ABC 200 1.9 2 Yes 8.85 8.85 9.37
54 100 95 ABC 0.1 1.9 2 Yes 5.72 5.72 6.25
55 100 95 ABC 1 1.9 2 Yes 5.98 5.72 6.25
56 100 95 ABC 5 1.9 2 Yes 5.98 6.25 6.25
57 100 95 ABC 10 1.9 2 Yes 5.98 6.25 6.25
58 100 95 ABC 20 1.9 2 Yes 6.25 6.25 7.29
59 100 95 ABC 100 1.9 2 Yes 8.85 8.85 10.41
60 100 5 AB 0.1 1.9 2 Yes 7.55 7.81 7.29
61 100 5 ABG 1 1.9 2 Yes 3.90 4.16 4.16
62 100 5 AG 5 1.9 2 Yes 4.68 5.72 6.25
63 100 5 BG 10 1.9 2 Yes 5.20 5.20 5.20
64 100 5 CG 20 1.9 2 Yes 3.90 4.16 5.20
65 100 95 AB 0.1 1.9 2 Yes 9.89 9.89 9.37
66 100 95 ABG 1 1.9 2 Yes 7.29 7.29 7.29
67 100 95 AG 5 1.9 2 Yes 11.45 11.45 11.45
68 100 95 BG 10 1.9 2 Yes 7.55 7.81 7.29
69 100 95 CG 20 1.9 2 Yes 5.98 6.25 6.25

Table 4 References
Comparison of techniques.

Method [14] [15] [16] [17] [18] [19] PM [1] A. Esmaeilian, M. Kezunovic, Impact of electromechanical wave oscillations
propagation on protection schemes, Electr. Power Syst. Res. (2016) 7.
SFreq 40.9 2 20 2 1.92 1 3.84 [2] A.P. Morais, G. Cardoso, L. Mariotto, G.D. Ferreira, Numerical distance relaying
RAT 1.5 1.5 1.5 1.75 2 1 0.43 algorithm based on mathematical morphology and least-squares curve fitting
CBurden High Low High Low Low Low Low method, Electr. Power Syst. Res. 81 (5) (2011) 1144–1150.
Rf NA NI NA NI NA NA NA [3] R. Baldick, B. Chowdhury, I. Dobson, Z. Dong, B. Gou, D. Hawkins, Z. Huang, M.
Joung, J. Kim, D. Kirschen, S. Lee, F. Li, J. Li, Z. Li, C.-C. Liu, X. Luo, L. Mili, S.
SFreq: sampling frequency (kHz); RAT: response of detection average time (cycle of Miller, M. Nakayama, M. Papic, R. Podmore, J. Rossmaier, K. Schneider, H. Sun,
60 Hz); CBurden: burden calculation; Rf: fault resistance; NA: not affected; NI: not K. Sun, D. Wang, Z. Wu, L. Yao, P. Zhang, W. Zhang, X. Zhang, Vulnerability
investigated; PM: proposed methodology. assessment for cascading failures in electric power systems, 2009 IEEE/PES
Power Systems Conference and Exposition (2009) 1–9.
[4] J.P. Conti, The day the samba stopped [power blackouts], Eng. Technol. 5 (4)
(2010) 46–47.
[5] C. Liu, J. Li, Patterns of cascaded events in blackouts, Power Energy Soc. Gen.
installed over the transmission line. A fast and reliable detector of Meet. – Convers. Deliv. Electr. Energy 21st Century (2008) 1–2.
[6] Operador Nacional Do Sistema Elétrico, Informativo ONS – relatório final
faults during power swing benefits the electrical system stability,
sobre blecaute de 2009, 2009.
leaving it least likely to blackouts. [7] North Americam Electric Reliability Corporation (NERC), Protection System
Response to Power Swings, Syst. Prot. Control Subcomm., 2013, pp. 1–61.
[8] IEEE Power System Relaying Committee, Power Swing and Out-of-Step
Considerations on Transmission Lines, Work. Gr. D-6 Rep., 2005, pp. 1–59.
Acknowledgments [9] M.K. Gunasegaran, C. Tan, A.H.A. Bakar, H. Mokhlis, H.A. Illias, Progress on
power swing blocking schemes and the impact of renewable energy on power
swing characteristics: a review, Renew. Sustain. Energy Rev. – ELSEVIER
This research work was supported by the Coordination for the (2015) 280–288.
Improvement of Higher Education Personnel (CAPES) for Doctoral [10] IEEE Power System Relaying Committee, IEEE Guide for Protective Relay
Research. Applications to Transmission Lines, 2004.
 C. Lazaro et al. / Electric Power Systems Research 155 (2018) 340–349
[11] N. Fischer, G. Benmouyal, D. Hou, D. Tziouvaras, J. Byrne-finley, B. Smyth, Do
system impedances really affect power swings – applying power swing
protection elements without complex system studies, 2012 65th Annu. Conf.
Prot. Relay Eng. (2012) 108–119.
[12] P.E. Mooney, N. Fischer, Application guidelines for power swing detection on
transmission systems, 2006 Power Systems Conference: Advanced Metering,
Protection, Control, Communication, and Distributed Resources (2006)
159–168.
[13] P.K. Nayak, J. Ganeswara, P. Kundu, A.K. Pradhan, P. Bajpai, A comparative
assessement of power swing detection techniques, Drives Energy Syst. 2010
Power India, 2010 Jt. Int. Conf. Power Electron. (2010) 1–4.
[14] S.M. Brahma, Distance relay with out-of-step blocking function using wavelet
transform, IEEE Trans. Power Deliv. 22 (3) (2007) 1360–1366.
[15] X. Lin, Y. Gao, P. Liu, A novel scheme to identify symmetrical faults occurring
during power swings, IEEE Trans. Power Deliv. 23 (1) (2008) 73–78.
349
[16] R. Dubey, S.R. Samantaray, Wavelet singular entropy-based symmetrical
fault-detection and out-of-step protection during power swing, IET Gener.
Trans. Distrib. 7 (10) (2013) 1123–1134.
[17] J. Khodaparast, M. Khederzadeh, Three-phase fault detection during power
swing by transient monitor, IEEE Trans. Power Syst. 30 (5) (2015) 2558–2565.
[18] A.P. Morais, G. Cardoso, L. Mariotto, G. Marchesan, A morphological filtering
algorithm for fault detection in transmission lines during power swings,
Electr. Power Syst. Res. 122 (2015) 10–18.
[19] R. Dubey, S.R. Samantaray, B.K. Panigrahi, V.G. Venkoparao,
Phase-space-based symmetrical fault detection during power swing, IET
Gener. Trans. Distrib. 10 (8) (2016) 1947–1956.
[20] P.M. Anderson, A.A. Fouad, Power System Control and Stability, second ed.,
Jonh wiley & Sons, New Jersey, 2003.