Electric Power Components and Systems

ISSN: 1532-5008 (Print) 1532-5016 (Online) Journal homepage: http://www.tandfonline.com/loi/uemp20

A Novel Approach to Detect Faults Occurring

During Power Swings by Abrupt Change of
Impedance Trajectory

Jalal Khodaparast, Mojtaba Khederzadeh, Filipe Faria da Silva & Claus Leth
Bak

To cite this article: Jalal Khodaparast, Mojtaba Khederzadeh, Filipe Faria da Silva & Claus
Leth Bak (2017): A Novel Approach to Detect Faults Occurring During Power Swings by
Abrupt Change of Impedance Trajectory, Electric Power Components and Systems, DOI:
10.1080/15325008.2017.1378774

To link to this article: https://doi.org/10.1080/15325008.2017.1378774

Published online: 28 Dec 2017.

Submit your article to this journal

Article views: 17

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at

http://www.tandfonline.com/action/journalInformation?journalCode=uemp20

Download by: [Universitetbiblioteket I Trondheim NTNU] Date: 13 January 2018, At: 00:56
 Electric Power Components and Systems, 0(0):1–15, 2017
Copyright C Taylor & Francis Group, LLC
ISSN: 1532-5008 print / 1532-5016 online
DOI: 10.1080/15325008.2017.1378774

A Novel Approach to Detect Faults Occurring During

Power Swings by Abrupt Change of Impedance
Trajectory
Jalal Khodaparast,1 Mojtaba Khederzadeh,2 Filipe Faria da Silva,2 and Claus Leth Bak2
1
Electrical Engineering Department, Shahid Beheshti University, Tehran, Iran
2
Department of Energy Technology, Aalborg University, Aalborg, Denmark
Downloaded by [Universitetbiblioteket I Trondheim NTNU] at 00:56 13 January 2018

CONTENTS
Abstract—The main purpose of power swing blocking is to distin-
1. Introduction guish faults from power swings. However, faults occurred during a
2. Limitation of Concentric Power Swing Blocker (CPSB) power swing should still be detected and cleared promptly. This paper
proposes an index based on detecting abrupt jump of impedance
3. Performance of Commercial Distance Relay in Detecting
trajectory by utilization of the predicting capability of the Kalman
Fault During Power Swing Filter. The proposed index is calculated by assessing the difference
4. Kalman Filter between predicted and actual samples of impedance. The predicted
5. Dynamic Phasor impedance samples are obtained using the Kalman filter and Taylor
expansion, which are used in this paper to track the phasor precisely.
6. Proposed Method for Detecting Fault During Power Swing Second order of Taylor expansion is used to decrease the corrugation
7. Simulation Results effect of impedance estimation and increase the reliability of the pro-
8. Conclusion posed method. Furthermore, in order to increase the selectivity of the
proposed method, the proposed index is armed with phase compari-
References son logic to detect internal faults. The instantaneous estimation and
prediction capability of the Kalman filter are two reasons for propos-
ing the use of Kalman filter. Intensive studies have been performed
and the merit of the method is demonstrated by test simulations.

1. INTRODUCTION
Distance relay mal-operation has raised concerns about black-
outs in power systems. Distance relays make decisions based
on the entering of impedance trajectory in protected zones.
When a fault occurs in a protected line, the impedance trajec-
tory enters in the distance relay zones and the relay operates.
However, this event (distance relay’s zones penetration) may
occur during power swing condition. During a power swing,
the voltage and current fluctuate simultaneously, causing fluc-
tuation in the measured apparent impedance at the distance
relay, which, hence, may enter the relay tripping zones. This
Keywords: Kalman filter, power swing, dynamic phasor, symmetrical fault, condition causes relay mal-operation and may lead to consec-
impedance trajectory, fault during power swing
Received 14 July 2016; accepted 2 September 2017 utive events (cascading outages) and even a major blackout
Address correspondence to Mojtaba Khederzadeh, Faculty of Electrical eventually [1]–[3].
Engineering, Shahid Beheshti University, A.C., Tehran 1658953571, Iran. To avoid this malfunction, power swing blocker (PSB) has
E-mail: [email protected]
Color versions of one or more of the figures in the article can be found online
become as a vital part of modern distance relay [4]. The major
at www.tandfonline.com/uemp. duty of PSB is discriminating power swing from fault and

1
 2 Electric Power Components and Systems, Vol. 0 (2017), No. 0

FIGURE 1. Simple two-machine system.

blocking distance relay from operating during power swing.

However, a fault can occur during power swing and it is vital
that the relay manages to properly detect the fault during the
event. Further, due to the symmetric nature of voltages and
Downloaded by [Universitetbiblioteket I Trondheim NTNU] at 00:56 13 January 2018

currents during the power swing, symmetrical faults are diffi-

cult to detect. Therefore, this issue attracts attentions of many
researchers at the moment.
There are various suggestions in the literature as to how
to deal with this issue. The most traditional method is uti-
lizing the rate of change of impedance (concentric power
swing blocker (CPSB)) for power swing detection [4]. One
limitation of this method is the incapacity in detecting faults
during power swing when impedance trajectory crosses con-
centric characteristics (it is exemplified by using Figure 1
as the sample system to simulate a power swing with the per-
formance of its distance relay’s CPSB as shown in Figure 2).
New methods based on voltage phase angle are presented
in [5, 6], but without considering high resistance fault and
symmetrical fault. Differential power-based fault detection
technique is another proposed method, which makes use of
autoregression technique to predict samples in the future
[7]. However, the mentioned method needs lots of simu-
lations to select appropriate parameters for the regression
method. Time frequency transforms are other solutions for
this issue. Wavelet transform and S-transform are presented
in [8]–[9] and [10]–[11], respectively, to detect power swing,
FIGURE 2. Performance of CPSB in different cases.
but they require high sampling rate, a requirement of most
Time–Frequency transforms. Some other methods based on in this method. In [18], a method based on the maximum
adaptive neuro-fuzzy inference system and support vector change rate of three-phase active and reactive powers has been
machine have been proposed in [12]–[14]. These methods proposed. However, the analytical demonstration of this strat-
require many simulations under different conditions for egy is based on a somewhat impractical hypothesis, which
training process and even retraining in other power sys- considers impedance without resistive component. Combi-
tem. Mathematical morphology is presented in [15, 16] for nation of Park’s transformation and moving data window is
detection of symmetrical fault during power swing and it is presented in [19] to extract power coefficients during fault
based on the monitoring of the shape of signal. Although this and power swing. These coefficients are approximately zero
method uses time domain transformation, the selection of during power swing and significant during fault. Computa-
processing function and its length is difficult. Moving average tional burden of power coefficients’ calculation limits the
is a low-pass filter used in [17] to discriminate power swing application of this method. Reference [20] proposes a method
from fault. The moving average varies periodically during based on fundamental frequency component detection that is
power swing, while it becomes either positive or negative con- created on instantaneous three-phase active power after incep-
sistently during fault. Utilization of all three-phase currents tion of a symmetrical fault. However, it assumed that the fault
even in symmetrical fault increases computational burden resistance is negligible. Reference [21] proposes a technique
 Khodaparast et al.: A Novel Approach to Detect Faults Occurring during Power Swings by Abrupt Change of Impedance Trajectory 3

based on negative sequence component of current and cumu- 2. LIMITATION OF CONCENTRIC POWER SWING
lative sum (CUSUM) for detecting three-phase fault during BLOCKER (CPSB)
power swing in series compensated line. A new method based
In order to show the capability and incapability of CPSB in
on extracted transient monitor (TM) index of current signal by
discriminating between fault and power swing under different
least square dynamic phasor is proposed in [22]. Phase space
conditions, three different cases are analyzed in this section. It
algorithm is presented in [23] to discriminate three-phase
is worthy to note that Mho characteristic is considered in this
fault during power swing based on shape information. Phase
section but the limitation is raiseable of all kinds of concentric
space representation of signal can be obtained based on main
characteristics such as Blinder and Lens.
signal and lagged main signal. The proposed index uses the
Case1: First test is programmed to examine capability of
distance between mapped point and the origin of x–y plane.
CPSB in detecting power swing. Consider that the power
Dimension of phase space and lagging sample number are
system shown in Figure 1 is under power swing condition.
two important quantities, which should be selected carefully.
Downloaded by [Universitetbiblioteket I Trondheim NTNU] at 00:56 13 January 2018

The impedance trajectory of this case is shown in Figure 2

Reference [24] proposes a new method based on monitoring
(Case1). According to this figure, after power swing initiation,
the center of circular impedance trajectory. According to
impedance starts to change and approximates relay zones.
this method, the center of impedance trajectory during the
Before entering the relay’s outermost zone, CPSB is placed to
power swing is always outside of zone1 but it comes to zone1
time the duration of the impedance locus as it travels between
during fault. This method uses the filtered value of impedance
outer and inner zones. This recorded time is compared with
trajectory, so the design of an appropriate filter is critical for
the threshold value for detecting power swing. Therefore,
this method. A novel estimation and error calculation-based
CPSB can detect power swing accurately under this condition.
method is proposed in [25]. It is based on estimating a future
Case2: A second test is programmed to show capability
sample current and error calculation and then index value
of CPSB in detecting fault during power swing in a special
(DI) based on sum of the entire squares of errors in the
condition. As in the previous case, first, the power swing
data window. A method based on wide area measurement
is simulated in power system. However, a three-phase fault
systems (WAMS) is proposed in [26] to prevent malfunction
is simulated at the end of the protected line during power
of distance relay during power swing. This method extracts
swing. The impedance trajectory of this case is shown in
the data from phasor measurement unit (PMU) and classifies
Figure 2 (Case2). According to the figure, after power swing
the power system conditions based on fuzzy logic algorithm.
initiation, impedance moves toward distance zones so that it
In this paper, fault during power swing is detected based on
reaches to CPSB for the first time and then traveling time
abrupt deviation of impedance trajectory at the fault initiation
is recorded. Therefore, power swing is detected by compar-
during power swing. The proposed index in this paper uses
ing recorded time with threshold value. As a consequence
prediction part of the Kalman filter to predict a sample ahead.
of power swing, impedance moves back and gets far away
Deviation in impedance trajectory is obtained by subtract-
from the distance zones leaving the outer CPSB characteris-
ing the predicted sample from the measured one. Moreover,
tic. Next, a three-phase fault occurs during power swing. This
dynamic phasor is developed to decrease error of impedance
causes impedance to cross the CPSB characteristic again dur-
estimation and as a consequence the reliability of the proposed
ing fault and new traveling time is recorded, which can be used
method is increased. In addition, phase comparison logic is
for detecting fault. Hence, CPSB can detect both power swing
added to the proposed method to improve its selectivity in
and fault during power swing in this case.
order to discriminate internal fault from external fault. The
Case3: A third test is prepared to show a condition
rest of the paper is organized as follows: Limitations of CPSB
in which CPSB cannot detect a fault during power swing.
in symmetrical fault detection during power swing are pre-
Impedance trajectory of this case is shown in Figure 2
sented in Section 2 to clarify the motivation of proposing
(Case3). Power swing is programmed as in the two previous
new method. Moreover, some common methods in commer-
cases. According to Figure 2 (Case3), after power swing initi-
cial distance relay for fault detection during power swing are
ation, impedance trajectory crosses CPSB for the first time
presented in Section 3. Sections 4 and 5 briefly describe the
and the traveling time is recorded, which can be used for
concepts of the Kalman filter and dynamic phasor, respec-
detecting power swing. However, a fault occurs, while the
tively, as foundations of the proposed method. Section 6 deals
impedance trajectory is located in the inner characteristic of
with complete description of the proposed method, and finally
CPSB. Therefore, impedance trajectory does not cross the
Section 7 presents simulation results to verify the proposed
CPSB again during fault. This condition shows incapacity of
approach.
CPSB for fault detection during power swing.
 4 Electric Power Components and Systems, Vol. 0 (2017), No. 0

To sum up, in this section, we have tried to explain 4. KALMAN FILTER

the problem with CPSB. Conditions may occur in which
Kalman filter is one of the most important and common data
impedance trajectory during power swing comes inside inner
fusion algorithms in use today. This success is due to its small
zone of CPSB and then a fault occurs during power swing.
computational requirement, elegant recursive properties, and
Under this condition, there will not be an impedance cross
its status as an optimal estimator [30]. Generally, the pur-
of CPSB for fault, so there will not be a new recorded time.
pose of the Kalman filter is to provide the closed-form recur-
Therefore, it is necessary to augment CPSB with new index,
sive solution for estimation of dynamic systems, which is
which can detect this condition.
described as

xk = Ak−1 xk−1 + qk−1
(1)
3. PERFORMANCE OF COMMERCIAL DISTANCE yk = Hk xk + rk
RELAY IN DETECTING FAULT DURING POWER where Eq. (1) shows the space state model of system, xk and
Downloaded by [Universitetbiblioteket I Trondheim NTNU] at 00:56 13 January 2018

SWING xk−1 are states of the system at time steps k and k − 1, respec-
Although commercial distance relay used negative sequence tively, Ak−1 and Hk are transition and measurement matrices of
for detecting unsymmetrical fault during power swing, they the system, respectively, qk−1 and rk are process and measure-
use different solutions for detecting symmetrical fault during ments noise, respectively, and finally yk is the measurement at
power swing. According to manuals of some of the commer- step k. The Kalman filter algorithm involves two stages: the
cial relays, these solutions are given in the following. prediction step, where the next state of the system is predicted
given the previous measurements, and the update step, where
the current state of the system is estimated given the mea-
3.1. Detection Based on Communication Facilities
surement at that time step [31]. The standard Kalman filter
Some commercial distance relays are equipped with power equations for the prediction stage are given in Eq. (2):
swing logic (PSL), whose main purpose is to secure selec-  −
tive and reliable operation of distance protection for fault dur- mk = Ak−1 mk−1
(2)
ing power swing [27]. PSL detects all kinds of faults during Pk− = Ak−1 Pk−1 ATk−1 + Qk−1
power swing by communication facilities and extra distance
where m− −
k and Pk are the predicted mean and covariance of
zones at both ends of the protected line, and it is activated
state at the time step k before seeing the measurement. The
after power swing detection. The main issue of this method is
predicted mean is updated by measurement in updating step:
the delay problem, which is common in communication-based ⎧
methods. ⎪
⎪ v = yk − Hk m−
⎪ k
⎪
k
⎪ = − T
⎪
⎨ S k Hk Pk Hk + Rk
− T
3.2. Detection Based on Additional Set of Inner Blinders Kk = Pk Hk /Sk (3)
⎪
⎪
⎪ mk = m− + Kk v k
⎪
These relays include an additional set of inner blinders to pro- ⎪
⎪ k
⎩
vide proper detection of the internal symmetrical fault [28]. If Pk = Pk− − Kk Sk KkT
the positive sequence impedance resides between these blin-
where
ders for a specific duration, PSB unblocks. The challenge
faced by this method is its performance in high-resistance
mk and Pk estimated mean and covariance of the state at the
fault, which may cause locating impedance outer of inner
time step k after seeing the measurement, respectively;
blinders.
v k measurement residual;
Sk posterior measurement covariance; and
3.3. Detection Based on Magnitude of Current Kk Kalman filter gain, which indicates how much modification
In order to protect the network against a fault that may occur is needed for the state variables based on measurement.
during power swing, blocking signals can be stopped when
current thresholds are exceeded [29]. Fault is detected when These formulations can help estimating the impedance at
the absolute value of differential current becomes greater than the first step (necessary in distance relay) and can also help
10% of the nominal current for three cycles. The main issue of to predict next sample of impedance in fault detection during
this method is the low level of current in high-resistance fault power swing. Since voltage and current signals are modulated
or fault in heavily loaded line. during power swing, the ordinary Kalman filter for impedance
 Khodaparast et al.: A Novel Approach to Detect Faults Occurring during Power Swings by Abrupt Change of Impedance Trajectory 5

calculation leads to insufficient accuracy in phasor estimation. 6. PROPOSED METHOD FOR DETECTING FAULT
Therefore, combination of Kalman filter and second order of DURING POWER SWING
Taylor expansion is used in this paper to increase accuracy
In this section, a new method is proposed to solve the limi-
of phasor estimation during power swing. This aim can be
tation of CPSB for detecting fault during power swing. This
achieved by dynamic phasor concept described in the next
method is based on abrupt change of impedance trajectory,
section.
whose index can be extracted via phasor estimation process.
Based on the proposed method, utilization of the Kalman filter
5. DYNAMIC PHASOR for phasor estimation in distance relay provides foundation for
solving the problem faced by CPSB when attempting to detect
Dynamic phasor is a recent concept in phasor estimation area, fault during power swing. Suppose a power swing occurs,
which considers phasor as a time varying complex parameter by considering (|V1 |/|V2 |)= constant, where |V1 | and |V2 | are
(both amplitude and phase are time varying parameters). This
Downloaded by [Universitetbiblioteket I Trondheim NTNU] at 00:56 13 January 2018

amplitudes at both ends of the transmission line, impedance

concept has been presented in [32, 33], and it consists of a trajectory moves in circular swing characteristic as shown in
more precise estimation of phasor when amplitude and phase Figure 3. However, when a fault occurs during power swing,
of signal change during time. This accuracy can be obtained impedance trajectory leaves circular swing characteristic and
using the first three terms of Taylor expansion of phasor as moves toward line impedance characteristic. According to this
(more than three terms will not improve estimation accuracy figure, abrupt leaving of swing characteristic (abrupt change
considerably [20]) : of impedance trajectory) at the fault initiation can be consid-
   ered as an indication of fault happening during power swing.
p |t=t0
p(t ) = [p|t=t0 ] + [p |t=t0 (t − t0 )] + (t − t0 )2 (4) Therefore, the proposed index (Eq. (7)) is obtained by sub-
2!
tracting measured impedance from predicted one:
where t is the current time and t0 is the previous time, p |t=t0
Index = impedance|measured − impedance| predicted (7)
and p |t=t0 are the first and second derivatives of time-varying
phasor at time t0 . Therefore, sinusoidal signal with dynamic where predicted impedance is obtained by predicted voltage
phasor (voltage or current) will be (Eq. (5)) phasor divided by current phasor. Therefore, the proposed
   index can be rewritten as Eq. (8):
1 p |t=t0
S(t ) = [p|t=t0 ] + [p |t=t0 (t − t0 )] + (t − t0 )2 e j2π f1 t
2 2! phasor − V |measured phasor − V | predicted
  ∗ Index = − (8)
1  p |t=t0 phasor − I|measured phasor − I| predicted
[p|t=t0 ] + [p |t=t0 (t − t0 )] + (t − t0 )2
e− j2π f1 t
2 2!
(5) In order to predict a sample ahead (voltage/current phasor),
prediction part of the Kalman filter is used in the proposed
where ∗ represents conjugating operator. Since the main aim is method. In addition, in order to increase reliability of the pro-
the estimation of dynamic phasor by Kalman, Eqs. (4) and (5) posed method (decrease phasor estimation error during power
should be used to construct transition and measurement matri- swing), the precision of the Kalman filter is augmented using
ces, which the Kalman filter is based on. Transition matrix can
be constructed by state space Eq. (6):
⎧ p |t=t
⎪ 
⎨ p(t ) = [p|t=t0 ] + [p |t=t0 (t − t0 )] + 2! 0 (t − t0 )
2

⎪ p (t ) = [p |t=t0 ] + [p |t=t0 (t − t0 )] (6)

⎩  
p (t ) = [p |t=t0 ]

where p (t ) and p (t ) are the first and second derivatives

of p(t ) in time domain. Transition matrix and measurement
matrix can be obtained as presented in [24]. In the next step,
the Kalman filter is employed based on these transition and
measurement matrices, which leads to dynamic phasor esti-
mation. The final step is the calculation of the impedance FIGURE 3. Illustration of the concept of impedance trajec-
using the dynamic phasor of voltage and current. tory deviation.
 6 Electric Power Components and Systems, Vol. 0 (2017), No. 0
Downloaded by [Universitetbiblioteket I Trondheim NTNU] at 00:56 13 January 2018

FIGURE 4. Flowchart of the proposed complete CPSB (CPSB and proposed method).

dynamic phasor concept previously explained. In order to pre-

scheme is similar to that of a line differential system in that
dict future samples (voltage/current) using the Kalman filter, it
the relay at each terminal monitors the currents going into
is necessary that the state space equations are obtained. Based
and out of the line. If the currents monitored at the two ends
on the state Eq. (5), the transition matrix is
of the line are not in the pre-determined correct phase rela-
⎡ ⎤ ⎡ τ2
⎤⎡ ⎤ tion, a flag is raised, which shows that an internal fault occurs
p((n + 1)τ ) 1 τ 2!
p(nτ )
⎣ p ((n + 1)τ ) ⎦ = ⎣ 0 1 τ ⎦ ⎣ p (nτ ) ⎦ (9) in transmission line. It is worthy to note that the comparison
p ((n + 1)τ ) 0 0 1 p (nτ ) is performed separately for each of the three phase currents
in the proposed method. Figure 4 shows the flowchart of the
where the middle matrix is the state transition matrix, the left augmented CPSB. The complementary part of CPSB, which
matrix is the prediction vector of the future sample, the right is proposed in this paper, is shown inside the dashed line box
matrix is the current phasor, and τ is the sampling time. Based in Figure 4. In Figure 4, outside of dash line box (dotted line
on Eq. (9), a sample ahead of voltage phasor and current pha- box) presents common CPSB, which operates based on the
sor is obtained and the predicted impedance can be calculated. rate of change of impedance. The decision of CPSB is used
Finally, based on Eq. (8), the proposed index is obtained. not only for judging power swing occurrence but also for acti-
One can suppose that the proposed scheme does not pro- vating complementary block. Therefore, this decision is sent
vide selectivity. Indeed, there is no guarantee that only inter- to the complementary block to trigger this block. In addition,
nal faults are detected by using the proposed index. For exam- in the parallel process, the proposed index (impedance tra-
ple, the proposed index could overcome the threshold in case jectory deviation) and phase comparison (PC) between cur-
an external close-in fault in adjacent line takes place, lead- rents local and remote ends (the remote signal is transmitted
ing to a false trip command issuing. To overcome this draw- by communication link) of transmission line are calculated in
back, additional logic (phase comparison scheme) is added to every sample (inside of the dashed line box). Therefore, if
the proposed method. The operation of the phase comparison CPSB detects power swing, complementary block is activated
 Khodaparast et al.: A Novel Approach to Detect Faults Occurring during Power Swings by Abrupt Change of Impedance Trajectory 7

and compares calculated index with threshold value (selection

process of threshold will be explained in the next paragraphs).
If the value of the proposed index is higher than h2 , the last
conditional operator is activated and compares the phase com-
parison (PC) with threshold value (h3 ) and finally it detects
fault during power swing if proposed index (Index) and PC
are higher than their threshold values.
According to Figure 4, we explain the whole procedure as
follows. First, CPSB operates and makes decision based on the
rate of change of impedance. Two concentric characteristics
and a timer are used by CPSB that records the duration of
impedance traveling between outer and inner zones of CPSB.
Downloaded by [Universitetbiblioteket I Trondheim NTNU] at 00:56 13 January 2018

If the recorded time is lower than the threshold (h1 ), CPSB

would detect it as a fault and would detect it as power swing if
the recorded time is higher than the threshold value. It is worth FIGURE 5. Main signal.
noting that the choice of this threshold (h1 ) is not affected by
the method proposed in this paper.
7.1. Dynamic Phasor Estimation During Power Swing
Next, if the CPSB detects power swing, the proposed
method would be triggered. Based on proposed method, then, This section examines the Kalman filter’s capability for
the calculated index is compared with the threshold (h2 ). If dynamic phasor estimation during power swing. Consider the
the proposed index is higher than the threshold (h2 ), fault test case (which is common in power swing) as
occurrence is anticipated and then the comparison (PC) logic ⎧
is activated. If PC is higher than the threshold value (h3 ), ⎨ S(t ) = a(t ) cos(2π f1t + φ(t ))
a(t ) = a0 + a1 sin(2π fat ), φ(t ) = φ0 + φ1 sin(2π fφ t )
the proposed method detects internal fault during power ⎩
swing. To determine the threshold (h2 ) for detecting a fault a0 = 1, a1 = 0.1, fa = 5, φ0 = 1, φ1 = 0.1, fa = fφ = 5.
during power swing by the proposed index, the maximum (10)
index values during different power swing conditions with The main signal (Figure 5) is sampled at 1 kHz, so 20 sam-
different swing frequencies and during different non-fault ples are obtained over a window of 20 ms. Estimation results
conditions should be examined. Moreover, minimum index of non-dynamic (zero-order Taylor) and dynamic (second-
values during different fault conditions (fault in different order Taylor) phasor are shown in Figures 6 and 7, respec-
distances, in different time initiation, and power angles) tively. In these figures, the dashed and solid lines repre-
should also be examined. According to the maximum and sent ideal (real) components and the estimations, respectively.
minimum values, the threshold (h2 ) should be considered According to these two figures, the dynamic concept is a better
a value between maximum and minimum to achieve the choice for power swing condition. The corrugated estimation
balance between dependability and security. It is worthy to
note that the threshold h3 is also selected based on the same
strategy.

7. SIMULATION RESULTS
Three simulation cases are provided next. The first case is
the assessment of the augmented Kalman filter performance
for phasor estimation during power swing to ensure the accu-
racy of dynamic phasor estimation process that is used in this
paper. The second case is a simulation of single machine to
infinite bus (SMIB) and the third case consists of the simu-
lation of a large power system (IEEE 39 Bus) to ensure the
appropriate performance of the proposed method in a large
power system. FIGURE 6. Outputs of zero-order Taylor and Kalman filter.
 8 Electric Power Components and Systems, Vol. 0 (2017), No. 0

FIGURE 9. Impedance trajectory for fault during power

swing in SMIB.
Downloaded by [Universitetbiblioteket I Trondheim NTNU] at 00:56 13 January 2018

FIGURE 7. Outputs of second-order Taylor and Kalman filter. swing (t = 1.9 sec at 25% of the line) and the impedance tra-
jectory of this condition is shown in Figure 9. According to
of phasor, which is shown in Figure 6, is a result of unsuit- the figure, the power swing makes impedance trajectory to
able definition of the phasor during power swing. However, move toward protected zone 3, which could lead to malfunc-
this corrugation effect is eliminated by dynamic phasor con- tion of the distance relay. In order to prevent such malfunction,
cept (i.e., use of a higher order Taylor expansion) as shown in CPSB (with two blinder characteristics) is designed for a dis-
Figure 7, which helps reducing the error of phasor estimation tance relay. The recorded time by CPSB is about 0.036 sec and
during power swing and results in a more reliable detection since this value is lower than the threshold (h1 = 0.005 sec),
process. CPSB operates accurately and detects it as a power swing, so it
blocks distance relay. Moreover, CPSB triggers the proposed
7.2. Simulation Results of Proposed Method in SMIB method to be ready for detecting fault during power swing.
In order to validate the proposed method for detecting fault Another three-phase fault (F2) occurs during power swing,
during power swing, the power system shown in Figure 8 is which should be detected by the proposed index. Under this
considered and simulated in PSCAD (power system’s data are condition, the proposed index (Index) and PC are calculated
presented in [8]). In order to accurate modeling of the power at every sample, which are shown in Figure 10. The figure
system in PSCAD, digital waveforms are created by using a shows that there is a peak value (190 ) corresponding to a
sampling rate of 10 kHz, and then downsampled to 20 sam- fault during power swing at t = 1.9 sec. Since the index value
ple/cycle and finally processed by MATLAB. A distance relay is higher than the threshold (h2 =15 ) and PC is higher than
is at bus 1 in the upper line (Line-1 = 280 km, Z1 = 0.12 h3 = 5º, internal fault is detected by the proposed method.
+ j0.88 /km). A three-phase fault (F1) is simulated at the Output of the proposed method (trip command) under this
middle of the lower line, which occurs at t = 1 sec, and it is
cleared after 0.03 sec by opening the breakers at both ends
(CB1, CB2). This occurrence causes a power swing on the
line between buses 1 and 2, which is observed by the distance
relay R. Another three-phase fault (F2) occurs during power

FIGURE 8. SMIB power system with two parallel transmis- FIGURE 10. Output of the proposed method in SMIB for
sion lines. fault during power swing.
 Khodaparast et al.: A Novel Approach to Detect Faults Occurring during Power Swings by Abrupt Change of Impedance Trajectory 9

Case Swing Frequency (Hz) Fault Initiation (sec) Power Angle (º) Fault Location (%) Index Value (Swing) () Index Value (Fault) ()

1 0.66 2.35 169 85 0.1 27

2 1.00 1.70 150 75 0.2 28
3 1.90 1.35 124 75 0.5 31
4 4.20 1.14 102 50 1 90
5 6.20 1.10 85 50 1.5 100
6 9.00 1.07 45 25 1.9 225

TABLE 1. Simulation results of SMIB under different conditions.

condition is also shown in Figure 10. According to the fig- two-phase fault (Phase-B and Phase-C to ground). According
ure, the proposed method detects internal fault during power to the left column, the proposed method can detect Phase-A
Downloaded by [Universitetbiblioteket I Trondheim NTNU] at 00:56 13 January 2018

swing and issues trip command to distance relay accurately. It as internal faulted phase and the trip command is sent just for
is worth mentioning that the time requirement for the Kalman Phase-A. However, in the right column, the proposed method
filter performance in every sample is about 50 μsec using a detects that Phase-B and Phase-C are two internal faulted
processor (Intel (R) Core(TM) 2 Duo, CUP 2.66 GHz, RAM phases and issues trip command for both of these phases.
4GB, Dell Vostro, Texas, USA) and about 30 µsec using a pro- According to these results, we can conclude that the proposed
cessor (Intel (R) Core(TM) i7-4600U, CUP 2.7 GHz, RAM method is applicable for both symmetrical and unsymmetrical
8GB, Dell Latitude, Texas, USA). faults during power swing.

7.2.1. Procedure for Determination of Threshold Value. In

order to show the procedure of threshold (h2 ) determination
and analysis of the proposed method under different condi-
tions (six cases), SMIB is simulated in different frequencies,
fault initiations, power angles, and fault locations. The pro-
posed index values during power swing and fault initiation
under different conditions are tabulated in Table 1 and shown
in Figure 11. According to Table 1, the threshold value (h2 =15
) is considered between the maximum value during power
swing conditions (1.9  is the maximum value in the sixth
column of Table 1) and the minimum value at fault initia-
tions (27  is the minimum value in the seventh column of
Table 1). According to Figure 11, the faults during power
swings are anticipated by a high value (higher than threshold
h2 ) of the proposed index during power swing under different
conditions. Moreover, the same strategy is also employed for
determination of h3 , in which the maximum value for external
faults is obtained to be 0.7ºand the minimum value for inter-
nal fault is obtained to be 80º. According to these values, the
threshold is considered to be 5º.

7.2.2. Performance of Proposed Method for Unsymmetrical

Fault. The proposed method in this paper is applicable for
both symmetrical and unsymmetrical faults. In order to show
the capability of the proposed method for detecting unsym-
metrical fault during power swing, single-phase fault and two-
phase fault are programmed in SMIB at t = 1.9 sec and their
results are shown in Figure 12. There are two columns in
this figure: the left column is regarding to single-phase fault FIGURE 11. Index values under different conditions accord-
(Phase-A to ground) and the right column is regarding to ing to Table 1.
 10 Electric Power Components and Systems, Vol. 0 (2017), No. 0
Downloaded by [Universitetbiblioteket I Trondheim NTNU] at 00:56 13 January 2018

FIGURE 12. Performance of proposed method for unsymmetrical fault.

7.3. Simulation Results for IEEE 39 Bus (New England) at t = 2.2 sec, during power swing. Based on the proposed
Figure 13 shows the location of the distance relay and of hypo- method, future sample of impedance trajectory is predicted
thetical three-phase faults in the IEEE 39 Bus system (data of by the Kalman filter and it is compared with the measured
this power system is presented in [34]). This power system impedance one sample later.
is simulated in PSCAD and in order to accurate modeling of Figure 14 shows the output of the proposed method for
the power system in PSCAD, digital waveforms are created internal fault during power swing. According to this figure,
by using sampling rate of 10 kHz, and then downsampled to there is an increase in index value at the fault during power
20 sample/cycle and finally processed by MATLAB. A three- swing at t = 2.2 sec. When a fault happens during power
phase fault (F1) occurs at t = 1sec in the line between Bus swing, the proposed method compares the index value (Index)
10 and Bus 13 and it is cleared after 0.25 sec. Consequently, with the threshold (h2 = 15 ) and predicts fault occur-
a power swing is created, which is detected by CPSB (Mho rence during power swing. In the next step, PC is compared
characteristic) first and then trigger command is sent to the with its threshold value (h3 = 5º) and detects an internal
proposed method to be ready for detecting fault during power fault because PC is higher than the threshold. Finally, the trip
swing. Suddenly, another three-phase fault (F2) occurs at 50% command is sent to distance relay, which is also shown in
of the line between Bus 16 and Bus 19 (line under analysis) Figure 14.
 Khodaparast et al.: A Novel Approach to Detect Faults Occurring during Power Swings by Abrupt Change of Impedance Trajectory 11
Downloaded by [Universitetbiblioteket I Trondheim NTNU] at 00:56 13 January 2018

FIGURE 15. Output of the proposed method in IEEE 39 Bus

for external fault during power swing.

on the proposed method, phase angle of the current of local

end is calculated by the Kalman filter and the phase angle of
the current of remote end is transmitted to the local termi-
nal. The impedance trajectory deviation (Index), phase dif-
ference between two ends (PC), and the trip command are
FIGURE 13. New England power system. shown in Figure 15. According to Figure 15, close-in fault
creates high value of impedance trajectory deviation but since
PC value is very low (PC = 0.7º) as compared to the thresh-
7.3.1. Performance of Proposed Method for Close-in Fault
old value (h3=5º), the proposed method detects this event
in Adjacent Line. In this section, the performance of the pro-
as an external fault and keeps distance relay in blocked state
posed method subjected to external fault is examined. Accord-
(trip = 0) .
ing to the proposed method, the internal and external faults
can be discriminated selectively by phase comparison index
(PC). PC is calculated based on the phase difference between 7.3.2. Performance of Proposed Method under Non-fault
the phase angles of the two line ends, which the remote-end Conditions. The dependability of the proposed method is
signal is transmitted by a communication link. In order to examined by the previous case (fault during power swing).
analyze this part of the proposed method, a three-phase fault Based on the dependability definition, it is expected that the
(F3) is simulated at 10% of the line between Bus 16 and Bus proposed method issues a trip command in fault event dur-
24 (adjacent line) at t = 2.2 sec, during power swing. Based ing power swing. However, it is necessary to examine both
dependability and security of any proposed method for assur-
ing the appropriate performance of the method. Therefore, in
order to examine security of the proposed method and assur-
ing no malfunction of the proposed method, other case studies
under non-fault conditions during power swing (load chang-
ing (100% increase) and capacitor switching (connection of
capacitor with capacity C = 2500 µF) during power swing)
are examined in this section. Based on the security definition,
it is expected that the proposed method does not provide a
false trip command in non-fault event during power swing.
Therefore, six different cases are simulated at t = 2.2 sec dur-
ing power swing and the results are tabulated in Table 2.
According to this table, the proposed method can avoid false
command trip for non-fault condition during power swing
FIGURE 14. Output of proposed method in IEEE 39 Bus for because the index values of all cases are lower than the thresh-
internal fault during power swing. old (h2=10 ).
 12 Electric Power Components and Systems, Vol. 0 (2017), No. 0

Case Condition Bus Index Value ()

1 Load switching 16 3
2 Load switching 15 1.25
3 Load switching 24 1.75
4 Capacitor switching 16 4.7
5 Capacitor switching 24 3.5
6 Capacitor switching 17 2.1

TABLE 2. Simulation results of IEEE 39 Bus under different non-

fault conditions.

7.3.3. Comparison with Other Methods. In order to com-

Downloaded by [Universitetbiblioteket I Trondheim NTNU] at 00:56 13 January 2018

pare the proposed method (PM) with other methods, five dif-
ferent methods are programmed (none of these five methods FIGURE 16. Performance of the proposed method and five
have the limitation presented in Section 2). These five meth- other methods in IEEE 39 Bus.
ods are as follows:
(M1) The method based on frequency component of three-
phase active power [20]: This method is based on the index values are normalized (divided by the maximum value)
detection of frequency components of an instantaneous three- so the maximum values (1 p.u.) in all methods are related
phase active power. It is demonstrated that during a power to fault initiation. So, by analyzing the index value during
swing, the instantaneous three-phase active power has a fre- power swing, the reliability of methods is compared. The cor-
quency component of slip frequency whereas after the fault responding indices of all six methods (shown in Figure 16)
inception time, a damping power-frequency component will during power swing and fault are tabulated in Table 2. Since
be created on its profile. the index values become normalized, the index value of all
(M2) T he method based on superimposed current [35]: methods are equal to one during fault, and so it is possible
This method is based on extracting superimposed component, to compare different methods with their index values dur-
created in fault time. When a fault occurs in power system, ing power swing so that the lowest value during power swing
it transits from normal to fault condition. The change cre- demonstrates the largest possible region for threshold selec-
ated as a consequence of this transition is considered as an tion and so the most reliable method.
index named superimposed component. This component is Therefore, by comparing the normalized index values dur-
calculated based on the expected no-fault current sample and ing power swing (according to the second column of Table 2),
subtracting it from the latest measured sample. the lowest value is related to the proposed method, so this
(M3) T he method based on transient monitor[22]: This method provides the largest region for threshold selection.
method is based on the extracted TM index of current signal In order to make clear the performance of different methods
by least square dynamic phasor, which is low during power (Methods 1—5) and the superiority of the proposed method,
swing and high at fault initiation. comparative results are shown in Figure 17. This chart shows
(M4) T he method based on a sample estimation and error the region of the maximum (third column of Table 2) and
calculation[25]: This method is based on estimating a future the minimum index (second column of Table 2) and so the
sample current and error calculation and then index value allowable region for selecting the threshold. According to this
(DI) based on sum of the entire squares of errors in the data chart, the proposed method has the largest allowable region
window. for selecting the threshold, which shows the highest reliability
(M5) T he method based on negative component [21]: Due compared to other methods.
to transients in current signals in the initial period, a negative Computational burden is another criteria, which is exam-
sequence component is noticeable even for three-phase faults. ined in this section for comparison between different meth-
Therefore, this component is extracted for current signal for ods. All six methods are compared based on float point oper-
detecting fault during power swing. ations (FLOPs) and the their results are tabulated in Table 3
All of these methods are simulated under the presented (last column). FLOP is a count of operations carried out by a
condition (fault occurs at t = 2.2 sec during power swing) in given algorithm or computer program. It is of interest to com-
IEEE 39 Bus power system and their performances are shown pare the computational complexity of two algorithms. Instead
in Figure 16. In order to provide comparison foundation, the of runtime counting, it is decided to count the FLOPs by
 Khodaparast et al.: A Novel Approach to Detect Faults Occurring during Power Swings by Abrupt Change of Impedance Trajectory 13

Method Power Swing Fault FLOPs of the augmented Kalman filter performance for phasor esti-
mation during power swing. The second case is a simula-
Proposed method 2×10−4 1 268,596
Method 1 1.25×10−3 1 87,532
tion of SMIB and the third case consists of the simulation
Method 2 0.040 1 47,766 of a large power system (IEEE 39 Bus). Moreover, the pro-
Method 3 5×10−4 1 241,596 posed method is compared with some other methods and the
Method 4 0.007 1 67,532 simulation results show that it has the highest level of reli-
Method 5 4×10−4 1 14,922 ability and the highest CPU computation time compared to
other methods. To sum up, according to the simulation results,
TABLE 3. Normalized index values and computational burden for
all six methods. the proposed method demonstrates its capability to properly
unblock a distance relay under fault condition during power
swing.
Downloaded by [Universitetbiblioteket I Trondheim NTNU] at 00:56 13 January 2018

REFERENCES
[1] N. Moaddabi and G. B. Gharehpetian, “Wide-area method for
self-healing of smart grids in unstable oscillations,” Electr.
Power Compon. Syst., vol. 41, no. 4, pp. 365–382, Jan. 2013.
[2] M. Moazzami, R. A. Hooshmand, and A. Khodabakhshian,
“Blackout prevention in power system using flexible AC trans-
mission system devices and combined corrective actions,”
Electr. Power Compon. Syst., vol. 41, no. 15, pp. 1433–1455,
Oct. 2013.
[3] S. R. Samantaraya, R. K. Dubeyb, and B. Chitti Babub, “A
novel time frequency transform based spectral energy function
for fault detection during power swing,” Electr. Power Compon.
FIGURE 17. Comparison between methods according to Syst., vol. 40, no. 8, pp. 881–897, Apr. 2012.
allowable region for threshold. [4] IEEE PSRC WG D6, “Power swing and out of step con-
siderations on transmission lines,” A report to power system
relaying committee of the IEEE Power Engineering Society,
scanning and parsing each line of the MATLAB codes and 2005.
infer the floating point operations. The required mathemati- [5] A. Mechraoui and D. W. P. Thomas, “A new blocking principle
cal operations of any program can be analyzed by matrix sizes with phase and earth fault detection during fast power swings
[36]. for distance protection,” IEEE Trans. Power Del., vol. 10, no.
According to Table 3, the highest computational burden is 3, pp. 1242–1248, Jul. 1995.
[6] A. Mechraoui and D. W. P. Thomas, “A new principle for high
related to the proposed method that shows the difficulty of
resistance earth fault detection during fast power swings for
this method. It is obvious that this high computational burden distance protection,” IEEE Trans. Power Del., vol. 12, no. 4,
is related to dynamic phasor estimation and phasor calculation pp. 1452–1457, Oct. 1997.
of both voltage and current in the proposed method. [7] J. Ganeswara and A. Pradhan, “Differential power based sym-
metrical fault detection during power swing,” IEEE Trans.
Power Del., vol. 27, no. 3, pp. 1557–1564, Jul. 2012.
[8] S. Brahma, “Distance relay with out of step blocking function
8. CONCLUSION
using wavelet transform,” IEEE Trans. Power Del., vol. 22, no.
This paper presents a new method for detecting a fault dur- 3, pp. 1360–1366, Jul. 2007.
ing power swing using abrupt change of impedance trajec- [9] R. Dubey and S. R. Samantaray, “Wavelet singular entropy-
based symmetrical fault-detection and Out-of-Step protection
tory at the fault initiation. This abrupt change is detected by
during power swing,” IET Generat., Trans. Distribut., vol. 7,
the deviation between measured impedance and an impedance no. 10, pp. 1123–1134, Oct. 2013.
predicted using the Kalman filter. In addition, dynamic pha- [10] N. Mohamad, A. Abidin, and I. Musirin, “Intelligent power
sor concept is employed to reduce the phasor estimation error swing detection scheme to prevent false relay tripping using
during power swing and so increase the reliability of the pro- S-transform,” Int. J. Emerg. Electric. Power Syst., vol. 15, no.
3, pp. 195–311, Jun. 2014.
posed method. Finally, phase comparison logic is used in the
[11] Z. Moravej, J. D. Ashkezari, and M. Pazoki, “An effective
proposed method to increase its selectivity for discrimina- combined method for symmetrical faults identification during
tion of internal from external faults. Three simulation cases power swing,” Electric Power System. Res., vol. 64, no. 1, pp.
are provided in this paper. The first case is the assessment 24–34, Jan. 2015.
 14 Electric Power Components and Systems, Vol. 0 (2017), No. 0

[12] H. Zade and Z. Li, “A novel power swing blocking scheme [29] MiCOM P441/P442/P444 – Technical Manual of Numerical
using adaptive neuro-fuzzy inference system,” Electric Power Distance Protection. ALSTOM, 2011.
Syst. Res., vol. 78, no. 7, pp. 1138–1146, 2008. [30] R. Faragher, “Understanding the basis of the Kalman filter via
[13] M. J. Reddy and D. K. Monhanta, “Adaptive-neuro-fuzzy infer- a simple and intuitive derivation,” IEEE Signal Process. Mag.,
ence system approach for transmission line fault classifica- vol. 29, no. 5, pp. 128–132, Sep. 2012.
tion and location incorporating effects of power swings,” IET [31] Jouni Hartikainen, Arno Solin, and Simo Sarkka, Optimal Fil-
Generat., Trans. Distribut., vol. 2, no. 2, pp. 235–244, Mar. tering with Kalman Filters and Smoothers. Espoo, Finland:
2008. Aalto Univ. School of Science, 2011.
[14] N. G. Chothani, B. R. Bhalja, and U. B. Parikh, “New support [32] Jose Antonio de la o serna, “Dynamic phasor estimates for
vector machine-based digital relaying scheme for discrimina- power system oscillations,” IEEE Trans. Instrument. Measur.,
tion between power swing and fault,” IET Generat., Trans. Dis- vol. 56, no. 5, pp. 1648–1657, Oct. 2007.
tribut., vol. 1, no. 1, pp. 17–25, Jan. 2014. [33] J. Serna and J. Rodriguez-Maldonado, “Instantaneous oscillat-
[15] S. Gautam and S. Brahma, “Out-of-step blocking function in ing phasor estimates with Taylor–Kalman filters,” IEEE Trans.
distance relay using mathematical morphology,” IET Generat., Power Sys., vol. 26, no. 4, pp. 2336–2344, Nov. 2011.
Downloaded by [Universitetbiblioteket I Trondheim NTNU] at 00:56 13 January 2018

Trans. Distribut., vol. 6, no. 4, pp. 313–319, Apr. 2012. [34] A. Pai, Energy Function Analysis for Power System Stability.
[16] A. P. Morais, G. C. Junior, L. Mariotto, and G. Marchesan, “A Berlin, Germany: Springer, 1989.
Morphological filtering algorithm for fault detection in trans- [35] A. Apostolov, D. Tholomier, and S. Richard, “Superim-
mission lines during power swings,” Electric Power Syst. Res., posed components based sub-cycle protection of transmission
vol. 122, no. 1, pp. 10–18, May2015. lines,” in Proc. Power Sys. Conf. Exposit., pp. 592–597, Oct.
[17] J. Rao and K. Pradhan, “Power swing detection using moving 2004.
averaging of current signals,” IEEE Trans. Power Del., vol. 30, [36] H. Qian, “Counting the Floating Point Operations (FLOPS),”
no. 1, pp. 368–376, Feb. 2015. MATLAB Central File Exchange, No. 50608, Ver. 1.0, June
[18] X. Lin, Y., Gao, and P. Liu, “A novel scheme to identify sym- 30, 2015.
metrical fault occurring during power swings,” IEEE Trans.
Power Del., vol. 21, no. 1, pp. 73–78, Jan. 2008.
[19] K. Andanapalli and P. LiB. Varmau, “Parks transformation
based symmetrical fault detection during power swing,” in
Proc. 18th Natl. Power Syst. Conf., Dec. 2014, pp. 1–5. BIOGRAPHIES
[20] B. Mahamedi and J. Zhu, “A novel approach to detect symmet-
Jalal Khodaparast was born in Iran in 1987. He received
rical faults occurring during power swings by using frequency
components of instantaneous three phase active power,” IEEE the B.Sc. and M.Sc. degrees in electrical engineering from
Trans. Power Del., vol. 27, no. 3, pp. 1368–1376, Jul. 2012. Shahrood University of Technology, Shahrood, Iran, in 2008
[21] P. Nayak and P. Bajpai, “A fault detection technique for the and 2010, respectively, and is currently working toward
series-compensated line during power swing,” IEEE Trans. the Ph.D. degree in the Electrical Engineering Department,
Power Del., vol. 28, no. 2, pp. 714–722, Apr. 2013. Shahid Beheshti University, Tehran, Iran. He was a Guest
[22] J. Khodapaast and M. Khederzadeh, “Three-phase fault detec-
tion during power swing by transient monitor,” IEEE Trans.
Ph.D. Student (visiting period) in the Aalborg University,
Power Syst., vol. 30, no. 5, pp. 2558–2565, Sep. 2015. Denmark, 2015–2016. His research interests include wide
[23] R. Dubey, S. R. Samantaray, B. K. Panigrahi, and V. G. area protection, flexible AC transmission systems (FACTS),
Venkoparao, “Phase-space-based symmetrical fault detection power system dynamics, digital signal processing, and phasor
during power swing,” IET Generat., Transmiss. Distrib., estimation.
vol. 10, no. 8, pp. 1947–1956, Feb. 2016.
[24] R. Jafari, N. Moaddabi, M. Eskandari-Nasab, and G. B.
Gharehpetian, “A novel power swing detection scheme inde- Mojtaba Khederzadeh (SM’06) received the B.Sc. degree
pendent of the rate of change of power system parameters,” in electrical engineering from Sharif University of Technol-
IEEE Trans. Power Del., vol. 29, no. 3, pp. 1192–1202, Feb. ogy, Tehran, Iran, in 1980, the M.Sc. degree in electrical engi-
2014. neering from Tehran University, Tehran, Iran, in 1990, and the
[25] I. G. Tekdemir and B. Alboyaci, “A novel approach for Ph.D. degree in electrical engineering from Sharif University
improvement of power swing blocking and deblocking func-
tions in distance relays,” IEEE Trans. Power Del., vol. 32,
of Technology in 1996. He was a Postdoctoral Fellow in the
no. 4, pp. 1986–1994, Aug. 2017. University of Western Ontario, London, ON, Canada, from
[26] P. Kundu and K. Pradhan, “Wide area measurement based pro- 2004 to 2005. He is currently an Associate Professor and the
tection support during power swing,” Electric Power Sys. Res., Director of the Power System Protection and Control Research
vol. 63, no. 1, pp. 546–554, Dec. 2014. Center, Electrical Engineering Department, Shahid Beheshti
[27] REL670 IEC – Manual of Line Distance Protection. ABB
University, Tehran, Iran. His research interests include power
Group, 2016.
[28] SEL-421 Relay – Instruction Manual of Protection and system protection, control and monitoring, and power system
Automation System. SEL, 2011. dynamics.
 Khodaparast et al.: A Novel Approach to Detect Faults Occurring during Power Swings by Abrupt Change of Impedance Trajectory 15

Filipe Faria da Silva was born in Portugal in 1985. He Claus Leth Bak was born in Denmark, in 1965. He received
received the M.Sc. degree in electrical and computers engi- the B.Sc. degree in electrical power engineering in 1992, the
neering from Instituto Superior Técnico, Portugal in 2008 and M.Sc. degree in electrical power engineering, Aalborg Uni-
the Ph.D. degree in electric power systems from Aalborg Uni- versity (AAU), in 1994, and the Ph.D. degree in 2015. He
versity, Denmark, in 2011. He was with EDP-Labelec in 2008 was a Professional Engineer with Electric Power Transmis-
and with Energinet.dk from 2008 to 2011. He is currently an sion and Substations, NV Net Transmission Company, with
Associate Professor in the Department of Energy Technology, specializations within the area of power system protection. In
Aalborg University, where he is also the Semester Coordina- 1999, he was an Assistant Professor with ET-AAU. He cur-
tor of the Electrical Power System and High Voltage Engi- rently holds a Full Professor position with ET-AAU. He serves
neering’ master program and the Vice Leader of the Mod- as the Head of the Energy Technology Ph.D. Program and
ern Power Transmission Systems’ research program. He is an the Head of the Section of Electric Power Systems and High
Active Member of CIGRE and IEEE, being currently the Head Voltage with AAU and is a Member of the Ph.D. Board in
Downloaded by [Universitetbiblioteket I Trondheim NTNU] at 00:56 13 January 2018

of Denmark’s IEEE-PES and a regular Member of the Cigré the Faculty of Engineering and Science. His research interests
SC C4. He has authored more than 70 articles and a book. include corona phenomena on overhead lines, power system
His research interests include power cables, electromagnetic modeling and transient simulations, underground cable trans-
transients, system modeling, network stability, HVDC trans- mission, power system harmonics, power system protection,
mission, and HV phenomena. and HVDC-VSC offshore transmission networks.