www.ietdl.

org
Published in IET Generation, Transmission & Distribution
Received on 11th July 2011
Revised on 28th October 2011
doi: 10.1049/iet-gtd.2011.0514

ISSN 1751-8687

Out-of-step blocking function in distance relay using

mathematical morphology
S. Gautam S.M. Brahma
Klipsch School of Electrical and Computer Engineering, New Mexico State University, Las Cruces, NM 88003-8001, USA
E-mail: gautam@nmsu.edu

Abstract: Out-of-step condition arising from an unstable power swing requires quick detection of the power swing followed by a
controlled separation of different areas of power system at predetermined locations. However, in order to perform such separation,
local operation of distance relays needs to be blocked using out-of-step blocking (OSB) function. Conventional blinder-based
method to implement the OSB function cannot detect a symmetrical fault once the relay is blocked during a power swing.
Different schemes are proposed to overcome this drawback. This study presents detection methods for power swing and
symmetrical faults, based on mathematical morphology. These methods offer distinct advantages over other schemes
proposed in the literature. The methods are combined and rigorously tested on waveforms generated using PSCAD/
EMTDCw. These methods are then used to propose a reliable and fast OSB tool that imposes a very low computation burden
at low sampling rate.

1 Introduction Researchers have proposed different methods to

implement the OSB function to overcome the inherent
United States Department of Energy describes its vision for problem with blinder-based schemes. Hou et al. [3] track
smart grid in a white paper [1]. The paper describes seven the swing centre voltage (SCV) for the purpose, but do not
characteristics that will bring this vision to fruition. One of account for the frequency deviations during the power
the characteristics requires the grid to ‘heal itself by swing. Mechraoui and Thomas [5] build a time-invariant
performing continuous self-assessments to detect and model and track the load angle to distinguish a fault from a
analyse issues, take corrective action to mitigate them and, power swing. They also report detection of a high-
if needed, rapidly restore grid components or network resistance earth fault during a power swing [6]. However
sections.’ Such requirements can arise during a power Mechraoui and Thomas [5, 6] do not deal with
swing, which is characterised by large fluctuations in power symmetrical faults, which are the primary source of
interchange over two areas of a power system. In case of an confusion. Su et al. [7] propose a scheme using SCV, but
unstable power swing, instead of taking local action, a their approach takes longer to detect faults. Brahma [4]
controlled separation of areas at pre-determined locations is reports an OSB tool using wavelet transforms. Despite
implemented using out-of-step tripping function [2]. This is being quite comprehensive, the main drawback of this
obviously a wide-area protection application. However, in approach is the high computational burden inherent with
order to implement such separation, local relays must not wavelets, and the requirement of a high sampling rate to
be allowed to operate. Unfortunately, distance relays tend to implement the tool. Lin et al. [8] report a scheme to detect
interpret the power swing as a symmetrical fault, and symmetrical faults during a power swing. The scheme is
therefore need to be blocked during a power swing. Out-of- able to detect a power swing, but the detection time for
step blocking (OSB) function is implemented in distance fault is very long; the fastest time reported is 14 ms.
relays to achieve this objective [2]. Lotfifard et al. [9] also report a scheme to detect
The problem with OSB is that once it is blocked, if not a symmetrical fault using Prony method, which is
properly designed, it becomes inoperative for symmetrical computationally intensive. Neuro-Fuzzy methods [10, 11]
faults. This is undesirable. Therefore, the OSB function are also proposed for the out-of-step blocking function, but
must be able to distinguish between a power swing and a these methods require training data, and may need to be
symmetrical fault, so the relay can be blocked during a retrained for application in a different system configuration.
power swing, but be allowed to operate for any symmetrical None of the reported schemes analyse cases when the
fault that may occur during the power swing [3]. Blinder- impedance calculated by the distance relay before the fault
based schemes are popularly used to implement the OSB and during the fault are very close, as may very well
function, but these schemes cannot recognise a symmetrical happen. Moreover, these papers do not model current
fault once the operating point enters the operating zones of transformers and voltage transformers while testing the
a distance relay [4]. proposed schemes. This paper aims to construct an

IET Gener. Transm. Distrib., 2012, Vol. 6, Iss. 4, pp. 313– 319 313
doi: 10.1049/iet-gtd.2011.0514 & The Institution of Engineering and Technology 2012
www.ietdl.org
improved and effective OSB function for distance relays implementation of OSB function is described in Section 6,
using mathematical morphology (MM). and conclusions are drawn in Section 7.
Morphological filters belong to a class of non-linear signal
processing tools that transform the shape of signals [12, 13], 2 Definition
thus deriving their name. These tools are based on set theory
and integral geometry. They are collectively called MM As stated earlier in Section 1, MM comprises two basic
and were originally developed by Matheron [14] and operations – dilation and erosion. Based on these two basic
Serra [15]. The terms ‘Morphological Filters’ and operations, several other operations are defined [17].
‘Mathematical Morphology’ are generally found to be used Dilation, erosion, opening and closing are the most
interchangeably, at least for power-related applications. In commonly used operations for the one-dimensional signals
contrast to tools like Fourier Transform and Wavelet typically seen in power systems. These operations are
Transform, MM entirely works in the time domain [16, 17], defined now.
and it has been extensively used in removing noise from Let f (n) be a signal, defined in domain Df ¼ {x0 , x1 , . . . ,
signals and images. xn}, and g(m) be the SE, defined in domain Dg ¼ {y0 , y1 ,
Morphological filters use a processing function called . . ., ym}, where n and m are integers, such that n . m.
the structuring element (SE) [18]. Dilation and erosion are Dilation of f (n) by g(m), ( f W g)
two basic operations in MM. These operations with their
different combinations and derivatives are used to extract yd (n) = (f ⊕ g)(n) = max{f (n − m) + g(m)}
one or more desired sub-signals from a given signal. These
sub-signals are then innovatively used for various 0 ≤ (n − m) ≤ n, m ≥ 0 (1)
applications. Choice of SE plays a crucial role in MM
operations. This choice is often based on past experience, Erosion of f (n) by g(m), (f ⊖ g) is
and trial-and-error. Gautam and Brahma [19 – 21] explore
the performance and capabilities of MM operators for real- ye (n) = (f ⊖g)(n) = min{f (n + m) − g(m)}
time applications in a power system. Gautam and Brahma 0 ≤ (n + m) ≤ n, m ≥ 0 (2)
[19, 20] discuss the effect of the SE on real-time
performance of MM. They also show the effectiveness of
Based on dilation and erosion, two composite operations
MM operations in detecting and extracting high-frequency
known as opening and closing are defined. The opening of
noise, for example, transients in the voltage waveforms at
f (n) by g(m), denoted by f ⊖ g, is defined as dilation of the
the instant of occurrence of fault. Gautam and Brahma [21]
eroded signal (f ⊖ g) by g
analyse the use of MM-based operations to effectively
detect and extract low-frequency signals, such as those
observed during power swings. yo (n) = (f W g)(n) = (ye ⊕ g)(n) = ((f ⊖g) ⊕ g)(n) (3)
As mentioned earlier, MM is a time-domain tool; therefore
it does not suffer from any frequency-related issues like Similarly, closing of f (n) by g(m), denoted by f † g, is defined
frequency drifts or aliasing. It is also computationally very as erosion of the dilated signal ( f ⊕ g) by g
efficient and fast, and does not require a high sampling rate,
thus overcoming the drawbacks of a wavelet-based yc (n) = (f † g)(n) = (yd ⊖ g)(n) = ((f ⊕ g) ⊖ g)(n) (4)
approach. Speed of detection is a major advantage of MM-
based techniques, the detection is often made in the sub- These operations simply involve addition and comparison,
cycle range. thus imposing extremely light computation burden.
This paper shows how MM can be applied in different Interested readers can refer to [19] for a detailed
ways to detect a power swing, and to detect a symmetrical explanation of these definitions and their real-time
fault during a power swing. In order to illustrate this, the interpretation.
paper first documents how the length of the SE can be The SE is the basis of all morphological operations, and is
adjusted to detect a power swing by processing the current used for the extraction of features and reduction of noise in a
signal. The associated time delay is also discussed. Then signal [22]. Depending on the signal type and application, the
the paper describes how the voltage waveform and a choice of the SE varies widely. It can be linear, square, disk-
different SE can be used to detect a symmetrical fault or ball-shaped, and even sinusoidal [23]. An appropriate
during a power swing. The two methods are then tested on choice of SE for a particular application depends on various
waveforms simulated using PSCAD/EMTDCw. The results factors including but not limited to the type and frequency
are analysed. Based on the analysis, a relay blocking logic of signal, signal spectrum and the sampling rate [24]. The
is proposed to implement a fast and effective OSB function frequency of interest determines the length of the SE, and
that imposes very low computation burden at a low this length in turn defines the size of the filter window.
sampling rate. Some trial and error using previous experience is also
The paper is divided into seven sections. Sections 1 and 2 needed to arrive at a final choice for the SE [19].
deal with the literature review and the definition of MM filters
and operators. Proposed MM operations to detect a power 3 Proposed MM operations to detect power
swing and a symmetrical fault during a power swing are swing and fault
outlined and discussed in Section 3. Section 4 describes the
set-up used to perform simulations, and documents some Based on the discussions in Section 1, the first objective of
results that help capture the sub-signal characteristics. Based this paper is to design an MM operation with appropriate
on these characteristics, Section 5 illustrates formulation SE to detect a power swing. A power swing sometimes
and comprehensive testing of a rule-based detection occurs when a power system is recovering from a severe
methodology to detect a power swing as well as a disturbance. Typical wave-shapes of system voltage and
symmetrical fault during a power swing. The current during a power swing are shown in Fig. 1, where it

314 IET Gener. Transm. Distrib., 2012, Vol. 6, Iss. 4, pp. 313 –319
& The Institution of Engineering and Technology 2012 doi: 10.1049/iet-gtd.2011.0514
 www.ietdl.org
abrupt, a short length of the SE is justified. The final
choices of height and length of the SE are made as 0.01
and 2, respectively. Both SE-C and SE-V are chosen to be
linear in nature. As mentioned in Section 1, this paper
further proposes a complete OSB tool, and a block diagram
to integrate this tool with a numerical relay.

4 Implementation of the proposed MM

operations with simulation results
Voltage and current waveforms were generated using
Fig. 1 Typical current and voltage waveforms during a power PSCAD/EMTDCw to test the performance of the operations
swing proposed in Section 3. A 400 kV transmission system
shown in Fig. 2 was used to generate the required
waveforms. The system details are provided in the
is easy to observe that a low-frequency envelope is appendix. One end of the two parallel transmission lines
superimposed on a power frequency signal. The frequency was modelled as an equivalent machine and the other end
of the envelope is called the swing frequency and may vary as an infinite bus. The transmission lines were modelled
between 1 – 3 Hz during slow swings and 4 – 7 Hz during with distributed parameters. The relay is assumed to be
fast swings [4]. It is crucial to detect such swings as located at breaker B1, fed through a 600/5 A multi-ratio
quickly as possible in order to take corrective action. Such class C100 CT, with tap set at 500/5 A, and a capacitor-
action may comprise controlled separation of parts of a coupled voltage transformer (CCVT) with the phase voltage
power system implemented through a special protection ratio of 230 kV/115 V. Assuming mho characteristics for
scheme. the relay, the zones were defined as 90, 150 and 200% of
Gautam and Brahma [21] have described an MM-based the length of in line-1. The CT model from PSCAD/
operation called closing – opening difference operation EMTDCw with Silectron 53 core was used for the
(CODO), and its use on current waveforms to detect a simulation. The saturation characteristics of Silectron 53
power swing, with successful simulation results. For the were modelled using the Lucas model [25 – 27]. Equation
sake of continuity, this operation is described below. (6) was used for the calculation of the cross-section area of
The CODO is the difference of closing and opening core
operation, as described mathematically by (5)
V = 4.44 × f × B × A × N volts (6)
yCODO (n) = yc (n) − yo (n) = (f †g)(n) − (f W g)(n) (5)
For the class of CT chosen, the induced voltage V in
The choice of the structuring element (SE-C) for this the secondary is 70 V at knee point, frequency f is 60 Hz,
operation has been made based on a frequency envelope of the magnetic flux density B is assumed to be 1.5 T and the
1 – 10 Hz to cover probable swing frequencies. The length number of secondary turns N is 100. These values yield
of the SE is a trade-off between security, influenced by the the cross-sectional area (A) of the core as 1.75 × 1023 m2.
length of buffer, and the detection time, influenced by The length of the magnetic path was taken as 0.5 m.
delay. An SE of shorter length provides less buffer between Similarly, the CCVT model available in PSCAD/
the output of MM during healthy state and the output EMTDCw was used for the simulation. The saturation
during a disturbance, whereas a longer SE unnecessarily characteristics of the CCVT were modelled using the Lucas
increases the detection time. Therefore slow and steady model [25 – 27]. The description of the CCVT model, and
changes in a waveform (like power swing) require longer calculation of its parameters are shown in the appendix.
SEs, whereas fast and abrupt changes (like faults) can be The calculation of parameters is based on [27, 28].
detected using shorter SEs. A low sampling rate of 32 spc (1.92 kHz) was chosen to
It can be seen from Fig. 1 that the low-frequency envelope suit the real-time application. All the figures in this paper
characterising a power swing visibly changes in one cycle. show the current and voltage waveforms referred to the
Therefore an SE with at least one-cycle length is required to phase A current sampled through the secondary side of CT
reliably detect the swing. With this in mind, an iterative and phase A voltage sampled through the secondary side of
process was implemented to optimise the length of the SE. CCVT, unless stated otherwise.
Lengths from 32 (one cycle) to 48 (1.5 cycle) were tried, A load angle of d ¼ 608 was chosen, causing a large power
and the length of 40 (with a default height of 0.01) flow that stresses the transmission lines. In order to simulate a
provided the best performance in terms of security and the power swing, a three-phase fault was applied at the mid-point
detection time for all cases considered. of line-2 at 0.5097 s, and breakers B5 and B6 were operated
The second objective of the paper is to detect a symmetrical
fault during a power swing. An attempt was made by Gautam
and Brahma [21] to use the CODO as described in the
previous paragraph to achieve this objective. The results
were not satisfactory, as the method yielded unacceptably
long detection times – more than 10 cycles in some cases.
Therefore this paper proposes to exploit the high-frequency
noise content of the voltage waveform at the occurrence of
fault, using a different choice of SE (SE-V). Same sampling
rate of 32 spc is assumed while making this choice. Since
the change in the voltage waveform owing to fault is very Fig. 2 System considered for simulation

IET Gener. Transm. Distrib., 2012, Vol. 6, Iss. 4, pp. 313– 319 315
doi: 10.1049/iet-gtd.2011.0514 & The Institution of Engineering and Technology 2012
www.ietdl.org

Fig. 4 Fault outside zone-3 during a power swing and the

corresponding change in the output of the CODO
Fig. 3 Current during a power swing and the corresponding
absolute change in the output of CODO

after 0.1 s to clear the fault. This delay in fault clearance

creates a power swing on line-1. The CT secondary current
was treated with CODO described in (5), using SE-C
described in Section 3.
Fig. 3a shows the current waveform and the CODO output.
A close-up of the part of the waveform covering the fault
period is also inserted. Power swing can be clearly
identified from the nature of the current waveform. It is
clear that the CODO output is capable of tracking the
current peak, both before and after the swing. The output of
CODO is a horizontal line before and during fault, but its
slope changes as it tracks the swing. This behaviour of
CODO lends itself to be exploited for the detection of the
power swing. The output of CODO, however, lags the
current waveform by a time decided by the length of the SE
and the sampling interval [19]. This time-lag constitutes the
filter delay. The filter delay for this case is 20.31 ms, which
is calculated as follows
Fig. 5 Fault inside zone-1 during a power swing and the
corresponding change in the output of the CODO
filter delay = (length of SE − 1) × (sampling interval (Dt))
= (40 − 1) × (16.667/32 ms)
This ensures further immunity from minor irregularities in
= 20.31 ms the CODO output. It can be clearly observed from Fig. 3b
that the change is practically zero consistently during
Fig. 3b shows the change in the absolute value of the CODO normal conditions. Close observation of the figure revels
output. This change is calculated each time a new sample is abrupt changes corresponding to the fault inception and
received, and is displayed as a bar-chart. Before calculating clearance time, with a lag time of 20.31 ms. After the swing
the change, the CODO output is first smoothed using a starts, the change in the CODO output increases
Moving Average operation described by (7). The average is substantially. The value reaches zero several times but only
taken over the length of the SE – LSE-C . This operation is for a brief period, and never lies below a certain threshold
necessary to remove some very minor irregularities in the for more than a few cycles. This change in the CODO
CODO output waveform output faithfully captures the characteristics of swing, and
provides a distinct signature that can be detected by a rule-
L
SE-C based method.
1
y′CODO (n) = yCODO (n − m + 1) (7) As mentioned in Section 3, in order to detect a symmetrical
LSE-C m=1 fault during a power swing, it is proposed that the transients
in the voltage waveforms be captured using a different
Dy′CODO (n) = |y′CODO (n) − y′CODO (n − 4 × spc)| (8) structuring element (SE-V). Figs. 4a and b show the
voltage waveform and the corresponding CODO output
Change in the absolute value of the smoothened CODO with the structuring element SE-V, when a three-phase fault
output is calculated using (8). It can be seen that the values during swing occurs on line-1 at 1.1362 s into the
used to calculate this change are spaced by four cycles. simulation at a distance of 5 km from bus-1. The chosen

316 IET Gener. Transm. Distrib., 2012, Vol. 6, Iss. 4, pp. 313 –319
& The Institution of Engineering and Technology 2012 doi: 10.1049/iet-gtd.2011.0514
 www.ietdl.org
fault instant corresponds to the time when the swing current is also shown. It can be clearly seen that the detection of swing
around its minimum value and the apparent impedance seen is very fast, and it easily outperforms the blinder-based
by relay is outside zone-3. schemes conventionally used. The detection time in Fig. 6
Figs. 5a and b show similar plots for the case where the translates to 67.4 ms.
three-phase fault occurs at 1.3011 s, which corresponds to Using the observations made in Section 4, another rule was
the time when the swing current is around its maximum formulated to detect a symmetrical fault during a swing: If the
value and the apparent impedance seen by relay is within CODO output corresponding to a voltage waveform in any of
zone-1. Careful observation of Figs. 4 and 5 shows that a the three phases is above a certain threshold value, a fault is
distinct spike is generated after the inception of fault. This detected. The same procedure as in the case of the previous
property is exploited for detection by a rule-based method. rule was implemented to determine the threshold value for
this rule. The threshold value in this case is 0.05 V, which
is approximately 200% of the value during the normal
5 Formulation of method to detect a power condition. This rule was used on voltage waveforms
swing and fault generated by simulating symmetrical faults during the
power swing at different instants and different locations.
A rule-based detection method was formulated to detect a Table 1 records the results, including the detection time.
power swing based on the observations documented in The fault inception times are chosen in pairs to simulate the
Section 4. The rule is: If the absolute change in the CODO fault in three different areas of the R – X plane viz outside
output corresponding to the current waveform in any of the zone-3 (FIT-1 and FIT-2), within zone-2 (FIT-3 and FIT-4)
three phases is above a certain threshold value for four and within zone-1 (FIT-5 and FIT-6). The fastest and the
consecutive cycles, a swing is detected. The threshold value slowest detection times are 0.6 and 1.5 ms, respectively.
for the rule was chosen as 0.05 A. The value is more than For all the cases in Table 1, different fault resistances were
200% of the CODO output during the normal condition, simulated within a range of 1– 10 V. The change in the fault
thus providing a very good buffer zone. It is important to resistance did not show any change in the fault detection time.
mention here that inception or clearing of faults or other The proposed method works even when the measured
disturbances in the system affect the current waveform only impedance before and after a fault are similar, as is the case
for a short period of time. The period of four cycles for FIT-5 and FIT-6 in Table 1, which correspond to faults
(67 ms) is therefore chosen to avoid such disturbances from in zone-1. The cases were also simulated for other load
being confused with a swing. Simulation results described angles, in order to test the proposed algorithm for different
later in this section show that this rule is never violated swing frequencies as shown in Fig. 7. The implementation
when the power system is operating in normal condition, or of the proposed rules for these cases resulted in successful
when there is a fault in the system, indicating that the
method is robust.
Fig. 6 shows the performance of this rule. The swing locus
is shown on a R– X plane, where the relay characteristics are

Fig. 7 Power swings at different swing frequencies corresponding

Fig. 6 R –X diagram showing the swing and its detection to different load angles

Table 1 Results from testing the detection method with corresponding detection times

SN Distance, km FIT-1 (1.1362 s) FIT-2 (1.1402 s) FIT-3 (1.2830 s) FIT-4 (1.2867 s) FIT-5 (1.2978 s) FIT-6 (1.3011 s)

FDT, s Delay, ms FDT, s Delay, ms FDT, s Delay, ms FDT, s Delay, ms FDT, s Delay, ms FDT, s Delay, ms

1 5 1.1370 0.8 1.1411 0.9 1.2839 0.9 1.2875 0.8 1.2984 0.6 1.3021 1.0
2 40 1.1370 0.8 1.1411 0.9 1.2839 0.9 1.2875 0.8 1.2990 1.2 1.3021 1.0
3 80 1.1375 1.3 1.1411 0.9 1.2844 1.4 1.2880 1.3 1.2990 1.2 1.3021 1.0
4 105 1.1375 1.3 1.1417 1.5 1.2844 1.4 1.2880 1.3 1.2990 1.2 1.3026 1.5
5 125 1.1375 1.3 1.1417 1.5 1.2844 1.4 1.2880 1.3 1.2990 1.2 1.3026 1.5

SN ¼ serial number; FIT ¼ fault inception time; FDT ¼ fault detection time

IET Gener. Transm. Distrib., 2012, Vol. 6, Iss. 4, pp. 313– 319 317
doi: 10.1049/iet-gtd.2011.0514 & The Institution of Engineering and Technology 2012
www.ietdl.org
detection of the swings as well as the faults. Thus, the swing
and the fault detection methods were comprehensively and
successfully tested.

6 Integration of OSB function with distance Fig. 9 Implementation of Truth Table in Table 2 with XOR gate
relay
7 Conclusion
Fig. 8 shows the block diagram describing the
implementation logic of the OSB function to be integrated This paper shows that MM operations can be designed for
with a distance relay. The current waveforms are sampled quick and efficient detection of a power swing, as well as a
continuously at 1.92 kHz, and fed to the Swing Detection symmetrical fault during a power swing. Current waveforms
Module. As soon as swing is detected, the swing flag (SF) are used to detect a swing, and voltage waveforms are used
is enabled and the distance relay is blocked from to detect a fault. The main difference in the design for the
responding to a symmetrical fault. The control is then two operations is the length of the SE. The operations are
transferred to the Fault Detection Module. The Fault used to create a comprehensive OSB tool that can be
Detection Module gets sampled voltage waveforms from integrated with a distance relay. The tool imposes very low
the Sampling Module, and performs the CODO operation computation burden, offers extremely quick detection of
described in Sections 3 and 4. When the Fault Detection swing as well as fault, is not affected by frequency
Module detects a symmetrical fault, the fault flag (FF) is variations during swing and does not require a high
enabled and the distance relay is released to operate for the sampling rate. The tool is rigorously and successfully tested
fault. Table 2 shows when the relay operation has to be at different swing frequencies, with different fault locations
blocked (indicated by 1), and when it has to be released at different inception angles and different fault resistances
(indicated by 0). The relay is only blocked when the swing using waveforms generated by PSCAD/EMTDCw. This tool
exists (SF enabled), but there is no symmetrical fault (FF will help achieve a reliable and secure separation of
disabled). The Fault Detection Module is not active unless different areas of a power system during an unstable power
SF is high, so the input combination (0, 1) is not possible swing, thus improving the wide-area protection and control
and hence declared not applicable in the truth table. The of grid.
Swing Detection Module continuously monitors the current
waveforms and disables the Fault Detection Module if
the swing subsides. This perfectly implements the OSB 8 References
function. Fig. 9 shows the implementation of the the truth 1 ‘A vision for the smart grid’. White Paper, US Department of Energy,
table formed by Table 2 using an XOR gate. Office of Electricity Delivery and Reliability, June 2009. Available from:
www.netl.doe.gov/smartgrid/referenceshelf/whitepapers/Whitepaper_The
Modern Grid Vision_APPROVED_2009_06_18.pdf =0pt
2 Tziouvaras, D., Hou, D.: ‘Out-of-step protection fundamentals and
advancements’, Available from: http://www.selinc.com/WorkArea/
DownloadAsset.aspx?id=2811=0pt
3 Hou, D., Benmouyal, G., Tziouvaras, D.: ‘Zero-setting power-swing
blocking protection’. IEE Conf. Publications, 2005, vol. 2005,
No. CP508, pp. 249–254
4 Brahma, S.: ‘Distance relay with out-of-step blocking function using
wavelet transform’, IEEE Trans. Power Deliv., 2007, 22, (3),
pp. 1360– 1366
5 Mechraoui, A., Thomas, D.: ‘A new blocking principle with phase and
earth fault detection during fast power swings for distance protection’,
IEEE Trans. Power Deliv., 1995, 10, (3), pp. 1242–1248
6 Mechraoui, A., Thomas, D.: ‘A new principle for high resistance earth
fault detection during fast power swings for distance protection’, IEEE
Trans. Power Deliv., 1997, 12, (4), pp. 1452–1457
7 Su, B., Dong, X., Bo, Z., et al.: ‘Fast detector of symmetrical fault
during power swing for distance relay’. IEEE Power Engineering
Society General Meeting, 2005, June 2005, vol. 2, pp. 1836–1841
8 Lin, X., Gao, Y., Liu, P.: ‘A novel scheme to identify symmetrical faults
occurring during power swings’, IEEE Trans. Power Deliv., 2008, 23,
(1), pp. 73–78
9 Lotfifard, S., Faiz, J., Kezunovic, M.: ‘Detection of symmetrical faults
by distance relays during power swings’, IEEE Trans. Power Deliv.,
2010, 25, (1), pp. 81–87
Fig. 8 Block diagram to implement the OSB function
10 Reddy, M., Mohanta, D.: ‘Adaptive-neuro-fuzzy inference system
approach for transmission line fault classification and location
incorporating effects of power swings’, IET Gener. Transm. Distrib.,
Table 2 Truth table 2008, 2, (2), pp. 235 –244
11 Zadeh, H.K., Li, Z.: ‘A novel power swing blocking scheme using
SF FF Relay logic adaptive neuro-fuzzy inference system’, Electr. Power Syst. Res.,
2008, 78, (7), pp. 1138– 1146
0 0 0 12 Maragos, P., Schafer, R.: ‘Morphological filters – part I: their set-
0 1 NAa theoretic analysis and relations to linear shift-invariant filters’, IEEE
1 0 1 Trans. Acoust. Speech, Signal Process, 1987, 35, (8), pp. 1153– 1169
1 1 0 13 Maragos, P., Schafer, R.: ‘Morphological filters – part II: their relations
to median, order-statistic, and stack filters’, IEEE Trans. Acoust. Speech
a Signal Process., 1987, 35, (8), pp. 1170– 1184
Fault detection module will not be enabled unless SF flag is 14 Matheron, G.: ‘Random sets and integral geometry’ (Wiley, New York,
active high 1975)

318 IET Gener. Transm. Distrib., 2012, Vol. 6, Iss. 4, pp. 313 –319
& The Institution of Engineering and Technology 2012 doi: 10.1049/iet-gtd.2011.0514
 www.ietdl.org
15 Serra, J.: ‘Image analysis and mathematical morphology’ (Academic,
New York, 1982)
16 Lu, Z., Turner, D., Wu, Q., Fitch, J., Mann, S.: ‘Morphological
transform for detection of power quality disturbances’. Int. Conf.
Power System Technology, 2004, PowerCon 2004, November 2004,
vol. 2, pp. 1644–1649
17 Li, G., Zhou, M., Luo, Y., Ni, Y.: ‘Power quality disturbance detection
based on mathematical morphology and fractal technique’. IEEE/PES
Transmission and Distribution Conf. and Exhibition: Asia and Pacific,
2005, 2005, pp. 1 –6
18 Zhang, C., Lin, X., Chen, Z., Bo, Z., Redfern, M.: ‘A novel
mathematical morphology based anti-CT saturation criterion for
generator differential protection’. 39th Int. Universities Power
Engineering Conf., 2004, UPEC 2004, September 2004, vol. 2,
pp. 763–767 Fig. 10 Circuit model of CCVT used for simulation
19 Gautam, S., Brahma, S.M.: ‘Overview of mathematical morphology in
power systems a tutorial approach’. IEEE Power and Energy Society
General Meeting, 2009, PES’09, July 2009, pp. 1– 7 Transmission lines
20 Gautam, S., Brahma, S.M.: ‘Properties of mathematical morphology
based filter for online filtering of power system signals’. 41st North
American Power Symp., 2009, NAPS’09, October 2009, pp. 1– 5 Z1 = Z2 = 0.12 + j0.88 V/km, Z0 = 0.309 + j1.297 V/km,
21 Gautam, S., Brahma, S.M.: ‘Application of mathematical morphology
based filters to detect a power swing’. IEEE Power and Energy C1 = 1.0876 × 10−8 F/km
Society General Meeting, 2010, PES’10, July 2010, pp. 1– 6
22 Jing, M., Zengping, W., Yan, X., Lei, M.: ‘Single-ended transient 9.2 Model of CCVT
positional protection of transmission lines using mathematical
morphology’. Seventh Int. Power Eng. Conf., 2005, IPEC 2005, 29
November 2005–2 December
The model of CCVT consists of four blocks of circuit.
23 Wu, Q., Lu, Z., Ji, T.: ‘Protective relaying of power systems using Initially the high voltage is stepped down using the
mathematical morphology’ (ser. Power Systems, Springer, 2009) capacitor voltage divider, which is then connected to the
24 Tingfang, Y., Pei, L., Xiangjun, Z., Li, K.: ‘Application of adaptive primary of a conventional voltage transformer (VT). A
generalized morphological filter in disturbance identification for power ferroresonance suppression circuit (FSC) and measuring/
system signatures’. Int. Conf. Power System Technology, 2006
PowerCon 2006, October 2006, pp. 1– 7 relaying instrument burden are connected to the secondary
25 Lucas, J.: ‘Representation of magnetization curves over a wide region of the VT. The model of CCVT used for the simulation is
using a non-integer power series’, Int. J. Electr. Eng. Educ., 1088, 25, shown in Fig. 10, and its parameters are given
(4), pp. 335– 340
26 Lucas, J., McLaren, P., Keerthipala, W., Jayasinghe, R.: ‘Improved √
simulation models for current and voltage transformers in relay Vp = 400/ 3 kV, Vm = 5 kV and Vs = 115 V
studies’, IEEE Trans. Power Deliv., 1992, 7, (1), pp. 152–159
27 Tziouvaras, D., McLaren, P., Alexander, G., et al.: ‘Mathematical Lseq = 0.94 mH and Rseq = 0.23 V
models for current, voltage, and coupling capacitor voltage
transformers’, IEEE Trans. Power Deliv., 2000, 15, (1), pp. 62– 72 Lbur = 2.4 H, Rbur = 301 V and Rpbur = 785 V
28 Fernandes, D. Jr., Neves, W., Vasconcelos, J.: ‘Coupling capacitor
voltage transformer: a model for electromagnetic transient studies’,
Electr. Power Syst. Res., 2007, 77, (2), pp. 125–134 Rf 1 = 5.5 V, Rf 2 = 3.9 V and Rf 3 = 40 V
Lf 1 = 0.01 H, Lf 2 = 0.394 H and Cf 2 = 8.0 mF
9 Appendix Core loss in normal condition, W0 = 7.5 W
9.1 Details of test system Using voltage divider rule for the capacitor C1 and C2
The details for the system shown in Fig. 2 are given below:
C1 + C2 Vp
=
C1 Vm
Equivalent generator G, 600 MVA, 22 kV, 60 Hz,
Vp
H = 4.4 MW s/MVA = 46.188 ⇒ C2 = 45.188 × C1
Vm

Xd = 1.81 pu, Xd′ = 0.3 pu, Xd′′ = 0.23 pu, Assuming C1 ¼ 9660 pF we get C2 ≃ 436 500 pF. The
compensating inductor LComp is then calculated as follows.
Xq = 1.76 pu, Xq′′ = 0.25 pu For full compensation

′
Td0 ′′
= 8 s, Td0 ′′
= 0.03 s, Tq0 = 0.03 s, XL = XC
wL = 1/wC
Ra = 0.003 pu, XP (Potier reactance) = 0.15 pu
where C ¼ C1 + C2 and L ¼ LComp + LTrans
Transformer, 600 MVA, 22/400 kV, 60 Hz, D/Y LTrans ¼ transformer reactance referred to primary side

⇒ L = 1/(w2 C) = 1/(3772 × (C1 + C2 )) = 15.77H

X = 0.163 pu, Xcore = 0.33 pu, Rcore = 0,
[ LComp = 15.77 H − 1.78 H = 13.99 H
Pcopper = 0.00177 pu

IET Gener. Transm. Distrib., 2012, Vol. 6, Iss. 4, pp. 313– 319 319
doi: 10.1049/iet-gtd.2011.0514 & The Institution of Engineering and Technology 2012