Presentation on

Series Compensated
Transmission Line Protection
Presented By:
Arindam Dev
Student ID: 0417062259
Course Name: EEE 6903 – Advanced Protective Relays
Session: April-2020
Topics to Be Covered

▪ What is series compensated transmission line

▪ Effects of series compensator
▪ Series compensated transmission line protection challenges
▪ Approaches for series compensated transmission line protection
▪ Conclusion
What Is Series Compensated
Transmission Line
What Is Series Compensated Transmission Line

Series compensation is the method of improving the system voltage

by connecting a capacitor in series with the transmission line.
Series capacitors are applied to negate a percentage of and hence
reduce the overall inductive reactance of a transmission line. The
benefits of applying series capacitors on a transmission line include
increased power transferring capacity, improved stability margins,
better load division on parallel paths, ability to adjust line load levels,
reduced transmission losses, and reduced voltage drop on the
system during severe disturbances.
What Is Series Compensated Transmission Line

Typically, series capacitors are applied to compensate for 25 to 75

percent of the inductive reactance of the transmission line.
Effects of Series Compensator
Effects of Series Capacitor

▪ Change in line impedance seen by relay

▪ Over voltage protection of series capacitor
▪ Voltage Inversion
▪ Current Inversion
▪ Series compensator generated transient issues
▪ Unbalanced line impedance
▪ Parallel Lines
Effects of Series Capacitor

▪ Fault in capacitor bank

▪ Problems regarding controlled series compensation
– Capacitive boost mode without MOV conduction
– Capacitive boost mode with MOV conduction
– Blocking mode
– TCSC bypass operation
– Circuit breaker bypass
 Change in Line Impedance Seen by Relay

The distance relay resolves the

impedance calculations from just
finished measurements of voltages and
currents. The impedance calculation
gets effected due to inclusion of the
series compensation, and leads the
impedance relay for exaggerate
conclusion. In ideal conditions, the
apparent impedance seen by a distance
relay for an uncompensated line at the
relaying end can be seen by a dashed
line in first figure and on R–X plane in
second figure.
 Change in Line Impedance Seen by Relay

With inclusion of series

compensation, the characteristic got
modified at the point of
compensation, as seen by solid lines
in the figures. It is clear from second
figure that the distance relay
overreaches if Series Capacitor (SC) is
included in the fault circuit.
 Change in Line Impedance Seen by Relay

The directional integrity of the

distance relay can be lost in the case
of a fault just after compensator for
end line compensation as seen in
figure
Over Voltage Protection of
Series Capacitor
Over Voltage Protection of Series Capacitor

▪ Normally modification in the relay

settings to accommodate series
compensation is made only when it is
established that the capacitor is
invariably going to be part of the fault
circuit for a fault after compensator .
▪ But over-voltage protection of the
series capacitor could bypass the
capacitor from the faulted circuit. As a
normal practice, a Spark Gap (SG) or
Metal Oxide Varistor (MOV) or both
with a bypass circuit breaker protects
the capacitor against over-voltage as
shown in the figure.
Over Voltage Protection of Series Capacitor

This phenomenon leads to two different impedance conditions during fault:

1. In high-current fault condition, voltage across the capacitor increases to a truly
high value, which triggers MOV conduction to bypass the capacitor. In this case,
SC–MOV combination impedance will be reduced to the impedance of MOV only.
2. During low-current fault condition; the MOV remained in its high impedance
state. The SC–MOV combination offers impedance equal to parallel combination
of the pair.
Two impedance conditions increases difficulty in relay setting. A relay setting
without consideration of MOV conduction could over-reach and easily lost its
directional integrity. If the settings are made with sustained consideration of
MOV, the relay may under-reach in low fault conditions.
Voltage Inversion
 Voltage Inversion

Due to the presence of the series compensation in the fault circuit, different
electrical quantities like fault current, phase currents, sequence currents and
voltages (phase, sequence) may exhibit significant phase shifts as compared
with their natural positions. A change of 90 degree or more in voltage phase
angle is known as voltage inversion.
This type of condition occurs in a fault that occurs after the series
compensator for which, the total source impedance (Zs) is much higher than
the fault line impedance (ZF)
Voltage Inversion

Any fault on line CD with

inclusion of short circuit in
the fault circuit will create
angular difference to the
line current on bus C and D.
This difference is
approximately equal to the
load angle of the line. This
will force the relay on bus C
to see the fault in reverse
direction.
Current Inversion
 Current Inversion

Current inversion is a
phenomenon in which the
current phase angle is changed
by more than 90◦. A condition of
current inversion occurs on a
series-compensated line when,
for an internal fault just after the
compensating device, the
equivalent system on one side of
the fault is capacitive, and
another side of the fault is
inductive as shown in the figure.
It is worth noticing here that, as current inversion and voltage inversion
occurs at total distinct system conditions, therefore, the voltage and
current inversion cannot occur simultaneously. Hence, in case of any one
of these inversions, the protective relay gets affected.
Series Compensator Generated
Transient Issues
Series Compensator Generated Transient Issues

▪Addition
  of series compensation in the transmission line adds up
several transients in the faulted system. These transients affect
estimation of the voltage and current phasors. A transient on a non-
compensated transmission line has a major portion of the decaying DC.
However, on lines with series compensation, the primary transient is an
AC signal with a frequency determined by the series capacitance and
the system inductance and can be given by (neglecting resistance)
=f
where XC represents reactance of SC, XL is the line reactance up to fault
point and f indicates the power frequency.
Series Compensator Generated Transient Issues

It is evident from the equation that a high-frequency component is

introduced into the system when a fault occurs just after the capacitor
or in other words, Xc > XL. However, all the modern relays are equipped
with digital filters; this makes effect of high frequency negligible in
relays. For the fault at the remote end of the line (XC < XL), a low
frequency transient will be introduced into the system.
Due to low fault current, MOV offers high impedance, and this
frequency will prevail in the system for long time. This frequency will
lead to a spirally oscillation of impedance characteristic, when it shifts
from load impedance to fault impedance during fault.
 Series Compensator Generated Transient
Issues

Assuming the transient frequency lesser

than the fundamental, the voltage drop
across the line inductor is much less for
the transient current compared to the
fundamental frequency current.
Conversely, the transient voltage drop
across the capacitor is considerably larger
due to the lower frequency of the
transient current. Thus, during low
frequency transient condition, the line
appears to have much higher percentage
compensation.

Fig: Impedance oscillation due to subharmonic frequency

Unbalanced Line Impedance
 Unbalanced Line Impedance

A non-transposed or partially transposed transmission line offers

unbalanced impedance. This unbalance is significantly magnified with
inclusion of series compensation in of total line impedance. That
means, increases in zero and negative sequence components of the
currents and needs change in protection algorithms, especially for
earth sensing relays.
Parallel Lines
 Parallel Lines

Problems with protection of the series compensated lines amplify with

a parallel line configuration. The series compensator gets rid of a
portion of the self-reactance of the transmission line. However, its
effect is negligible on the mutual impedance of the parallel lines.
Moreover, the outage of any line and grounding on either end changes
the impedance equation. Therefore, all these conditions should also be
considered during the protection design of parallel series compensated
transmission line protection.
Fault in Capacitor Bank
 Fault in Capacitor Bank

The series compensators are normally installed as a capacitor bank on a

platform. An internal over-voltage protection is always necessary for
this capacitor bank. The compensator necessitates shorting the whole
capacitor platform in case of any over-voltage (for example, group
over-voltage because of fuse blowing). For a single-phase fault,
capacitor bank protection system needs to bypass other two-phase
capacitor platforms also. A delay in bypass of other two platforms
provides a window of opportunity for a high-speed sensitive ground
and direction detection schemes to operate. This is due to the fact that
this capacitor unbalance appears as internal fault of the protected
system
Problems Regarding Controlled
Series Compensator
 Problems Regarding Controlled Series
Compensator

Electronically Controlled Series Compensation (CSC) provides better

control over power flow of a transmission line than fixed series
compensation. The CSC increases transient stability of the system as
well. However, it introduces additional harmonics, rapid changes
associated to its control actions in primary system parameters such as
line impedances and load currents. Being the most preferred
controllable series compensator and covering all possible conditions for
other controllable devices, Thyristor Controlled Series Compensator
(TCSC) has been chosen in this review to identify effects of controllable
compensation on the transmission line protection.
Problems Regarding Controlled Series
Compensator

This figure shows a typical

TCSC module with different
protective elements. It
comprises a series capacitor,
in parallel with a thyristor-
controlled reactor (TCR) with a
reactor (Ls) and a MOV
 Problems Regarding Controlled Series
Compensator

The MOV will help the ideal tripping characteristic to be nearer to its original
than the non MOV operation. The circuit breaker bypasses the TCSC module
in case of a severe fault or equipment malfunction. A current limiting inductor
(Ld) is incorporated into the circuit to restrain both the magnitude and the
frequency of the capacitor current during the capacitor bypass operation.
During the fault, TCSC’s control system would react swiftly to take some
protective measures. TCSC control circuit changes thyristor firing angles to
take it in inductive mode, which reduces the fault current. If the short circuit
current were large enough, MOV would be in reduces resistance mode, and
will produce equivalent impedance of TCSC in bypassed mode (inductive).
Therefore, TCSC’s capacitive reactance would decrease and gradually change
into inductive reactance. This affects adversely to the transmission line
protection settings.
 Problems Regarding Controlled Series
Compensator

If the fault current is not large enough, TCSC’s control system will not
send its commands to bypass the thyristor, in this case TCSC’s
impedance characteristic becomes complicated. Moreover, with a fault
where transient voltage and current’s waveforms are not symmetrical,
the TCSC firing becomes confusing. In this condition, TCSC’s harmonics
become very complicated.
While investigating the effect on distance protection measurement due
to the residual compensation when an earth fault occurs on a series
compensated line, it has been observed that the directional integrity of
the normal impedance relay will be compromised with TCSC
installation.
 Problems Regarding Controlled Series
Compensator

To understand the TCSC operation during a fault period, let we

understand the normal operating conditions of TCSC. The TCSC
normally operates with anyone out of following operating modes:
1. Capacitive Boost Mode
2. Inductive Boost Mode
3. Blocking Mode
4. Bypass Mode
During fault, TCSC gives different operations depending upon the type
of fault.
 Capacitive Boost Mode
1. Without MOV Conduction
2. With MOV Conduction
Capacitive Boost Mode Without MOV
Conduction

In a high impedance fault

condition, a low fault current
exists in the system. The lower
fault current exerts less voltage
across the compensator than the
protective voltage level of TCSC.
Therefore, the MOV remains in
high impedance mode, and fault
will persist through TCSC
continuously. Due to existence of
compensation, the relay can
overreach considerably, and also
can lose its directional integrity. Fig : Capacitive Boost Mode Without MOV Conduction
Capacitive Boost Mode With MOV
Conduction

For a high-current fault case, the MOV

conducts to decrease the voltage across
the SC. However, the MOV is fast enough
to conduct and reset within a half-cycle. In
this case, neither the MOV nor the circuit
breaker would be short out the capacitor
continuously. This short-duration
condition usually repeated several times
during the fault period. During this
condition, the TCSC impedance would be
the parallel combination of the capacitor
and the MOV in a lower resistance mode.
The relay would overreach but differently
from the previous case without MOV
operation.
Fig : Capacitive Boost Mode Without MOV Conduction
Blocking Mode
 Blocking Mode

During fault transient time, phase

angle of the voltage across capacitor
changes swiftly, which changes the
firing angle of the TCSC rapidly in
some cases. To avoid overcurrent
situation for this case, the thyristors
are blocked by firing mechanism. In
this condition, the TCSC acts like a
fixed series capacitor only. The relay
would overreach as in case of fixed
SC with MOV. However, this
overreach is less than the case, when
the TCSC is in capacitive boost mode.
Fig : Blocking Mode
 Bypass Mode
1. TCSC Bypass Operation
2. Circuit Breaker Bypass
 TCSC Bypass Operation

For a very heavy fault current

condition, the MOV operation is
not enough to decrease the
capacitor voltage. This leads
toward total thyristor
conduction. In this case, the
distance relay would underreach
due to the presence of the
reactor in circuit.

Fig : TCSC Bypass Operation

 Circuit Breaker Bypass Operation

If the fault is not cleared within a

pre-specified time period, the
TCSC transits to circuit breaker
bypass mode. Since the series
reactor in the circuit breaker
circuit is very small, the relay
experiences the normal situation.
This condition is used only for
back-up protection.

Fig : Circuit Breaker Bypass Operation

 Series Compensated
Transmission Line Protection
Challenges
 Series compensated transmission line
protection challenges
Effects of Series
Compensation Consequences Problem with Distance Relay
1. Relay overreach
Abrupt change in line 2. In a close in fault after
impedance at the point of compensation, relay can see fault
series compensation in reverse direction
1. Produces low frequency transients
Capacitor bypassed by gap for 2. Capacitor will be bypassed before its Relay overreach
overvoltage protection voltage limit
Capacitor is bypassed by
series reactor for overvoltage Will produce high frequency transients Relay underreach
protection of capacitor
1. Low fault current: impedance = 1. •Change of impedance with
MOV is used for overvoltage capacitive reactance MOV conduction
protection of capacitor 2. High fault current: impedance = 2. Relay overreach
Parallel of capacitive reactant and
MOV resistance
Series compensated transmission line
protection challenges
Effects of Series
Compensation Consequences Problem with Distance Relay
1. Overreach if fault include
capacitor
Voltage inversion Voltage phase angle shifts more than 2. A potential underreach if fault
90◦ occurs just after capacitor
(reverse direction)
Current phase angle shifts more than Relay fails to operate for some
Current inversion 90◦ portion of line
Produced due to combination of SC and Can conduct MOV at low fault
Sub harmonic frequency line inductive reactance currents
Increase in already available unbalance
Unbalanced line impedance due unbalanced loading or lack of Affect the ground relay
transposition
Faults in capacitor bank Will produce imbalance in the system Affect the ground relay
Series compensated transmission line
protection challenges

Effect of different modes of TCSC operations on distance protection:

Mode of operation of TCSC in faulty
condition Effect on Protection System

Capacitive boost mode without MOV 1. Overreach of relay

conduction 2. Directional integrity will be lost

Capacitive boost mode with MOV Relay will overreach, but less than the case without
MOV operation
Blocking mode 1. TCSC will work like SC
2. Relay will overreach as in SC–MOV
TCSC bypass operation Distance relay will underreach
1. Normal operation of relay
Circuit breaker bypass 2. Used only as backup protection
 Approaches for Series
Compensated Transmission Line
Protection
Series Compensated Transmission Line
Protection

Addition of series compensation force changes into design of the

transmission line protection system. These changes are according to
alteration in system parameters as per size, location and over voltage
protection of the compensator as discussed earlier. Besides the
detection of faulty condition, fault type classification is an essential
protective relaying aspect for transmission line protection. In a series
compensated transmission line, the end line fault currents may be of
the order of the load currents. Moreover, identical impedance
conditions exist on either side of the series compensation. This
necessitates information about location of fault with respect to the
compensator (fault zone). Therefore, the final impedance calculations
of the distance relay always use fault type and zone information.
Series Compensated Transmission Line
Protection

In recent times, considerable development efforts have been registered

for development of protective relaying schemes for series compensated
transmission lines. The available approaches can broadly be classified in
following categories:
1. Application of multiple/dynamic impedance characteristics according
to the presence of the series compensator in the fault circuit, or to use
memory polarization to eliminate voltage alteration across
compensation.
2. Development of a mathematical model of the compensation device to
estimate the impedance level during a course of fault. The voltage across
the compensator can be estimated to correct the measured voltage.
Series Compensated Transmission Line
Protection

3. Protection of the series compensated line with a digital protection system that employs
an advanced digital signal processing tool and/or artificial intelligence technique. Normally,
these schemes operate into two stages:
▪ Signal pre-processing and feature extraction

In this stage, features of the measured electrical quantities are extracted and enhanced
for further analysis. Signal processing tools like Fourier Transform (FT), Discrete Fourier
Transform (DFT), Fast Fourier Transform (FFT), Wavelet Transform (WT) and Discrete
Wavelet Transform (DWT) are normally used for this purpose.
▪ Artificial intelligence
In second stage, the extracted features are used with artificial intelligence techniques
for decision making. Normally used classifiers are Artificial Neural Network (ANN), Fuzzy
Set Theory, Support Vector Machine (SVM), Extreme Learning Machine (ELM) etc.
Series Compensated Transmission Line
Protection

In elementary implementation of the distance relay for series compensated

transmission line protection; the reach settings have been reduced to accommodate
effects of series compensation. The first zone used to be set far below its factual reach;
around 30% of the line length instead of 80–90%. This cautious approach takes care of
disproportionally between actual and measured fault impedances. However, it will
make relay underreach during MOV operation. One more solution to avoid relay over
reaching is to use line side voltage measurement with respect to the compensator
instead of normally used bus side voltage. However, this solution can be applied easily
in case of end line compensation, where the compensator is available inside the sub-
station only. This solution is impracticable for a mid-line compensation scheme.
 Series Compensated Transmission Line
Protection

▪ Multiple Relay Characteristics

A fault zone detector sub-algorithm is a prime requirement in this type of relaying, which
investigate the presence and percentage reactance of series compensation in the fault circuit.
The relay alters the impedance characteristic according to the size, shape and reactance of the
compensator. In an initial approach for adaption of distance relay characteristic to series
compensation, Mandour and El-Alaily utilizes the existence of sub-harmonics in fault current to
investigate participation of the compensator in the fault circuit. According to the fault zone,
appropriate impedance characteristic has been chosen for the relaying task. However, the
compensator model employed in this work, utilizes spark-gap for capacitor overvoltage
protection. With introduction of the MOV, the proposed scheme in this work became erstwhile
and necessitates changes in computation. In an effort, Srivani and Vittal presented an
algorithm for generation of varying impedance characteristic according to the level of fixed
compensation. The dynamic characteristic has been developed using offline network details
and Full Cycle Fourier Algorithm (FCF). However, the scheme developed for this work has been
authenticated for three fixed levels of compensations only. Moreover, a requirement of more
than one cycle post fault data proves it slow compare to recent requirements.
Multiple Relay Characteristics

A protective algorithm divides the

transmission line into two sub-
systems of linear and nonlinear
components to modify the
characteristic according to
compensation level. The prime
advantage gained through this type
of schemes is that the relay
characteristic gets modified/selected
adaptively according to the level of
the compensation. However, these
algorithms usually proved slower as
modification/adoption necessitates
knowledge of the fault zone before
proceed to the final characteristic.
 Series Compensated Transmission Line
Protection

▪ Compensator Modelling
As already mentioned, the distance protection scheme for transmission line protection might calculate
the fault impedance fallaciously due to compensator voltage drop. This problem can be eliminated by
calculating this voltage drop with real-time measured quantities. One of the methods is to develop a
compensator equivalent impedance model for calculations. Goldsworthy presented an applicable linearize
SC–MOV model for system parameter calculation in fault conditions as shown in figure. This model
provides an approximation of the actual value only. However, this model has been preferred by many
researchers as fairly simple in application. Based on Goldsworthy, Coursol et al. reported a quasi-linear
model of SC–MOV combination for use in simulation studies. These models are used by numbers of the
researchers for series compensated line protection applications. These models are found fairly useful in SC–
MOV impedance calculations and hence for voltage estimation across compensator.
 Compensator Modelling

▪ A more recent non-linear model of parallel combination of SC and MOV has been
developed by Rosolowski and Saha. The model utilizes compensator current measurement
as an input parameter. This necessitates a dedicated communication channel with real-
time measurements in case of mid-line compensation. Usage of interpolation during
calculation in this model reduces the prediction accuracy. Moreover, increased
computational complexity encourages the researchers to use the previous linear models.
Saha et al. presented a SC–MOV impedance prevision algorithm by calculating two
different impedances across compensator. The first realized with measured voltage and
current and the other with measured current and calculated compensation voltage with
help of 2nd order gear differentiation rule. In this method, the final impedance has been
estimated by checking position of these two calculated impedances on three specially
shaped regions on the impedance plane. The compensator modeling provides a great
advantage in compensated line protection, as voltage alteration by the compensator can
be estimated. However, it is difficult to replicate exact non-linearity of the compensator in
mathematical modeling and these models marginally differ to the actual measured values.
This indicates requirement of more considerable efforts in its developments.
 Series Compensated Transmission Line
Protection
▪ Travelling Wave Based Approach
Traveling wave is multi-frequency transient
wave generated from the fault point due to sudden
change in system parameters with the inception of
fault. These traveling waves propagate on either side
of line and reflected back continuously until dies
down as shown in figure. Traveling wave equations
are sufficient for development of a transmission line
protection system. However, with inclusion of the
non-linear component of the series compensator,
these equations need modification to adopt
changes. A modified traveling wave equation,
including non-linearity because of compensation has
been developed by Thomas et al.
 Travelling Wave Based Approach

The algorithm estimates voltage on either side of the compensator

instead of modeling the compensator device. This makes the algorithm
independent of compensator type, mode of operation and its
parameters. However, this two-ended scheme requires a dedicated
communication channel and a high amount of calculations, to sum up
the final result that makes it slow. It is worth to note that traveling wave
is almost absent when the voltage fault inception angle is near to zero.
Moreover, involvement of the series compensator can produce
frequencies of a very high range. These frequencies can be too high for
a Capacitive Voltage Transformer (CVT) for a close fault after
compensating device, as it can be out of its measurable bandwidth.
 Series Compensated Transmission Line
Protection

▪ Signal processing tool and artificial intelligence

Digital Signal Processing (DSP)is the method to describe a signal
into another form that makes certain features of the original signal
more amenable for study. DSP is able to describe the electrical signal
more completely for power system stability and protection analysis.
 Signal processing tool and artificial
intelligence

▪ Advancement in Filtering
The basic signal processing tool of Discrete Fourier Transform (DFT) decomposes the continuous
time signal into an infinite sum of sinusoids. The FT, in other words, is based on the principle of
dilation of the subjected waveform in to reference sinusoidal waves. The DFT separates out all
spectral details of the subjected time domain signal in the frequency domain as shown in the figure
below. In research articles by Yang and Liu and Gu and Yu, DFT based algorithms have been proposed
with completely removal of decaying DC component of fault currents for non-compensated
transmission line. Based on these works, Yu and Gu presented Fourier filter based series compensated
transmission line protection scheme for simultaneously removal of the sub synchronous frequency
components to develop a clear fundamental frequency component in a quick time. With series
compensation, this process can take a long time to extract features due to long time decaying sub-
harmonic resonance.
 Advancement in Filtering

The DFT is one of the most preferred tools for protection industry. The
computational cost of the recursive DFT filter is lower. Moreover, a very good
harmonic immunity can be achieved by DFT filtration, which improves the fault
information. However, during approximation of a discontinuous function (fault
induces current/voltage); the function is defined in the whole interval of interest
by a global basis set. Therefore, Gibb’s phenomenon will restrict the final
resolution. With exploration of recent signal processing tools like Gabor
Transform (GT) and Wavelet Transforms (WT); the implementation of DFT has
reduced drastically, as they can provide more meaningful information for
protective requirements
 Signal processing tool and artificial
intelligence

▪ Higher Order Statistic

Spectral analysis is a good measure for judgment concerning the
predictability of the signal. Second order measures like DFT generate non-
zero values in its output due to Gaussian noise. Higher-Order Statistic (HOS)
eliminates this Gaussian noise and can produce more accurate tri-spectrum of
higher frequency, which is advantageous for fault analysis with series
compensation. The use of HOS with fuzzy classifier was investigated by
Pradhan et al. for fault classification. However, this method requires a heavy
computation that makes the system sluggish.
 Signal processing tool and artificial
intelligence

▪ Wavelet Transform
Both time and frequency resolution of a given signal is achieved by time localization of
different frequency components using Wavelet Transform (WT). In WT, the inputs are
manipulated through a process of translation (i.e. movements along the time axis) and
dilation (i.e. spreading out of the wavelet) to transform the signal into another form which
‘unfolds’ it in time and scale. The wavelet transform measures the correlation between the
input signal and scaled and translated version of the ‘mother wavelet’ which is of limited
duration and has zero average value. Application of WT for series compensated line
protection for fault-zone identification using ‘db4’ mother wavelet and fault classification
with ‘Haar’ as mother wavelet was proposed by Megahed et al. In this article, the use of
sampling frequency of 240 kHz proved high for practical implementation and enforces a
large amount of calculations. Dash and Samantray presented an algorithm for fault type
and zone classification using WT analysis for a controllable compensated transmission line.
The same authors proposed fault type and zone identification scheme with an improved
signal processing method of wavelet packet transform.
 Wavelet Transform

The entropy defined in terms of its probability distribution can be a good measure of randomness or
uncertainty. The entropy provides improved information about the system condition in a state of fault.
Wavelet entropy based fault classification, zone identification and location scheme has been proposed
by El-Zonkoly and Desouki. Samantaray and Dash presented an approach based on s-transform and
Probabilistic Neural Network (PNN) for fault analysis. The s-transform is an extension of Gabor
transform and wavelet transform, and is based on moving and scalable localizing Gaussian window.
However, application of the PNN requires an extensive training and memory as PNN stores all the
training information with its network. Another application of s-transform in unit type TCSC
compensated transmission line protection can be seen in [46]. The scheme in this article works on s-
transform separated differential energies between sending and receiving end. The scheme necessitates
synchronization of measurement with data from the remote end. Therefore, it costs dedicated
communication channel between line ends. Moreover, necessity of two-cycle post fault data proves it
slow in view of recent available literature. An advancement in s-transform, hyperbolic s-transform is
applied for fault classification, zone identification and fault location estimation with SVMs, SVCs, and
SVRs, respectively, in [47]. Choice of the support vector parameters is a major concern for the scheme.
Moreover, requirement and selection of a large input vector make it slow in implementation.
 Wavelet Transform

In recent developments the ability of wavelet transform has been widely appreciated for
transmission line protection. Being a non-stationary phenomenon, fault analysis information
about particular spectral components occurring at the time of fault occurrence is very
important. As WT expands a signal in terms of a wavelet, generated using translation and
dilation of a fixed wavelet function, it extracts time and frequency features simultaneously
(Fig. 13). By the help of WT, the changes introduced by the compensating device can be
identified clearly to enhance fault information. However, most of the wavelet based
techniques uses multi-level decomposition that requires multi-level filtering and a higher
amount of calculations. Researchers are expected to work for reduction in filtering
requirement and calculations by improvisation of the methodology.
 Signal processing tool and artificial
intelligence

▪ Artificial intelligence techniques with/without digital signal processing

tools
 Neural Network
An Artificial Neural Network (ANN) is a computational model based on the structure
and functions of biological neurons. The ANN helps to estimate the computing function
or distributions in the most efficient way. In a way, ANN is a better random function
approximation tool. Many contributions are registered in the literature for series
compensated line with ANN for fault analysis with or without help of a signal-processing
tool. The ANN is considered as one of the better methods for voltage and current pattern
classification. By investigation of its performance with Deterministic Differential
Approach (DDA) and Linear Model (LM) methods for online calculation of the voltage
across a non-linear SC–MOV model, itis found that the performance of ANN is superior
than DDA and LM.
 Neural Network

Xuan et al. applied ANN for pattern recognition based adaptive relaying
for a series compensated transmission line. The delta-bar-delta (DBD)
training algorithm used in this effort improves time response by
accelerating the convergence. However, authors investigated first zone
performance for the approach only with fixed series end-line
configuration. Radial Base Function based Neural Network (RBFNN) is a
three-layer NN with the middle layer carries radial basis as activation
function. An approach for protection of a controllable compensated
transmission line has been developed with use of RBFNN by Song et al.
[51]. However, to fit in today’s protection requirement, the system should
not be accurate only, but fast also. RBFN requires comparatively large
processing time, as its activation function is non-monotonic compare to
BPNN.
 Neural Network

Limitations of ANN or any AI technique as pattern recognition classifier can be overcome by inclusion of a
pre-processing signal analysis tool. Different protection schemes with two stages (i) signal processing and
(ii) classification are available. A combination of DFT as a signal-processing tool with ANN has been
proposed by Song et al, with special emphasis on the zone-1 performance. However, a two-cycle window
after fault inception required in this scheme is a rare possibility in today’s fast protection environment.
Moreover, scheme claims to produce improvement in results with data of two consecutive windows. In the
first stage of a two-stage algorithm presented by Ibrahim et al,the model information from the measured
signal has been extracted by the Total Least Square Estimation of Signal Parameters via Rotational
Invariance Technique (TLS-ESPRIT) with non-fundamental portion of the measured transient signal. This
model information is used for high-resolution signal parameter estimation. In the second stage, an artificial
neural network (ANN) is designed to estimate the faulted phase based upon the features extracted from
the first stage. The TLS-ESPRIT deals with a very high amount of calculations. Testing of the algorithm
under system parameter variation could authenticate the presented algorithm in this work. In an approach,
Abdelaziz et al. examined a use of a model transform techniques of ‘Wedepohl’ transformation for fault
detection and ‘Karrenbauer’ transformation for fault classification with ANN. Extreme Learning Machine
(ELM) is a learning algorithm for ANN proposed by Huang et al. The ELM provides faster training speed by
eliminating issues like local minima, improper learning rate and over-fitting. The ELM trained ANN has been
used with DWT for fault analysis.
 Neural Network

The ANN proved to be a simple, robust and efficient classifier for protection applications.
The efforts also can be seen toward making the ANN scheme more practicable by
increasing its training speed and accuracy with various training methods. Following
advantages of ANN can be separated out over other classifiers:
1. It is able to acquire complex, non-linear relationships
2. Its generalization capabilities are good so can be used for different applications
3. It is quick in response
4. t is very easy to implement in a digital system.
However, it requires an extensive training that necessitates a large training data set and
time. This leads to generation of a large number of example simulations. A new ANN
training is required for every line where the relay is used. ANN is sensitive to system
parameter variation also, like frequency
Artificial intelligence techniques with/without
digital signal processing tools

Support Vector Machine

Support vector machine is a computational learning method based on statistical
learning theory. In recent years, SVM has emerged as a powerful tool for
classification and regression problems. In SVM, the input features are mapped into
higher dimensional dot product space for better classification. This high
dimensional space is called the feature space. In this feature space, the SVM finds
out separating hyperplane according to the training data that contains ‘target
value’ (class type) and ‘attributes’ (features). The training is made to find out
‘support vectors’ on, and around this hyper plane in a way that the separation
between the classes is maximum. This optimal hyper plane is found by utilizing the
optimization theory with application of the statistical learning theory.
 Support Vector Machine

SVMs emerged as a potential tool for protective classification. The SVM

has a capability to handle very large feature spaces, so it is very efficient
to handle large classification problems. Moreover, by defining the
support vectors it can classify classes very close to each other.
However, SVMs are very sensitive to their classification parameters. In
the absence of this parameter selection mechanism, the SVM
parameters of cost (C), gamma and kernel function are chosen by trial-
and-error methods in all these schemes.
Artificial intelligence techniques with/without
digital signal processing tools

Fuzzy Logic Based Scheme

A fuzzy logic system represents the nonlinear data mapping of input into a
scalar value with degrees of truthfulness (‘1’) and falsehood (‘0’). Pradhan et al.
[60] proposed a fuzzy logic based system aided with DWT signal processing for
fault analysis. The scheme applies fourteen rules based ‘min–max’ type fuzzy
system for fault classification. The fault zone identification is presented with a
nine rule based fuzzy system with two separate ratios generated from DWT. The
approach cannot guarantee same accuracy with fuzzy systems as not tested with
variety of fault resistances and fault inception angles.
The application of fuzzy logic for series compensated line protection is limited due
to large variation in system parameters. The formation of fuzzy rules under such
variation makes it difficult and slow in nature. However, other classifier with a help
of a fuzzy system can make the system more accurate and easy.
 Conclusion

It is clear from the discussion that the choice and application of protective relay for a
series compensated transmission line needs a very careful evaluation. With addition of
series compensation, the accuracy and reliability of distance relay are in question.
Neither of the solution replicates non-linearity of the compensator in mathematical
modeling. An accurate compensator model that takes care of entire non-linearity of
the compensator is still due from researchers. The AI techniques, like ANN are fast but
lack in accuracy and take more efforts for training. On the other side, the Fuzzy
systems are difficult to design to accommodate variations. Moreover, their size
increases exponentially with an increase in input feature space. Their ability to
separate two classes very close is limited. SVMs can overcome these problems;
however, SVMs are sensitive to their classification parameter. Therefore, it requires
experimental investigation for proper values in the absence of proper parameter
adjustment mechanism. This indicates need to identification of a faster and accurate
artificial intelligence technique that can handle more nonlinearity and a large feature
space with reduced mathematical burden.