See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/224244752

Synchrophasor Assisted Adaptive Reach Setting of Distance Relays in

Presence of UPFC

Article in IEEE Systems Journal · October 2011

DOI: 10.1109/JSYST.2011.2158694 · Source: IEEE Xplore

CITATIONS READS

37 175

3 authors, including:

Seethalekshmi K. Sn Singh
Institute of Engineering & Technology - Lucknow Indian Institute of Technology Kanpur
35 PUBLICATIONS 485 CITATIONS 389 PUBLICATIONS 9,631 CITATIONS

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Forecasting (Solar Power, Wind Power and Market Price ) View project

Effect of Electric Vehicles on Power Distribution System View project

All content following this page was uploaded by Seethalekshmi K. on 14 December 2018.

The user has requested enhancement of the downloaded file.

396
Synchrophasor Assisted Adaptive Reach Setting of
Distance Relays in Presence of UPFC
K. Seethalekshmi, Student Member, IEEE, S. N. Singh, Senior Member, IEEE, and
S. C. Srivastava, Senior Member, IEEE
Abstract—The operation of Flexible AC Transmission System
Controllers (FACTS) in the power transmission system pose a
challenge to the distance relaying scheme. This paper suggests an
adaptive scheme for estimating the trip boundaries of a distance
relay in presence of a Unified Power Flow Controller (UPFC),
utilizing a Generalized Regression Neural Network (GRNN).
Initially, the impact of the UPFC on the relay’s trip boundary
is studied for its automatic power flow control mode as well as
bypass mode of operation. The GRNN has been trained off-line
with the data generated from a detailed performance analysis
of the power system for various faults considering the effects
of the UPFC, fault resistance, and system loading conditions on
the trip boundaries. This work has also proposed a strategy that
computes the control parameters of the UPFC on-line, namely,
series voltage and reactive current injections, utilizing the syn-
chronized phasor measurements from Phasor Measurement Units
(PMUs). Pre-fault system states, including the control parameters
of the UPFC and the apparent impedance measured by the relay
unit have been utilized by the GRNN for predicting the trip
boundaries of the relay. The proposed scheme has considered
Single Line-to-Ground (SLG), Double Line-to-Ground (LLG) and
Three Phase-to-Ground (LLLG) faults and the effectiveness of the
scheme has been tested on 39-bus New England system and also
on a 17-bus system, a reduced equivalent of practical Northern
Regional Power Grid (NRPG) system in India.
Index Terms—Distance relay, generalized regression neural net-
work (GRNN), synchrophasor, trip boundary, unified power flow
controller (UPFC).
I. INTRODUCTION
T HE introduction of Flexible AC Transmission System
(FACTS) [1] controllers in the power system opens
up new challenges to the line protection as they change the
impedance of the lines dynamically. Consequently, distance
relays, in the associated transmission system, will have an
overreaching or underreaching effect depending on the control
modes of the FACTS controllers. Hence, determining the
boundaries of operation of a distance relay, adaptively in the
presence of FACTS controllers, is a challenging task.
An extensive analysis of the distance protection scheme, in
the presence of various FACTS controllers, has been carried
Manuscript received February 28, 2011; revised May 29, 2011; accepted May
29, 2011. Date of publication July 05, 2011; date of current version August 24,
2011.This work was supported by the Department of Science and Technology,
New Delhi, India, under Project DST/EE/20100258.
The authors are with the Department of Electrical Engineering, Indian Insti-
tute of Technology Kanpur, Kanpur 208016, India (e-mail: seethalk@iitk.ac.in;
snsingh@iitk.ac.in; scs@iitk.ac.in).
Digital Object Identifier 10.1109/JSYST.2011.2158694
1932-8184/$26.00 © 2011 IEEE
IEEE SYSTEMS JOURNAL, VOL. 5, NO. 3, SEPTEMBER 2011
out in the literature [2]–[8]. It deals with the change in the ap-
parent impedance measured by the relay in the presence of these
controllers. The performance analysis in [3] utilizes a steady
state voltage source model of the UPFC with fixed compensa-
tion levels for studying the effects of location of the UPFC and
fault resistance on the distance relay operation. A more exhaus-
tive analysis has been carried out in [4], [5] considering a de-
tailed model of the UPFC. However, [3]–[5] do not discuss the
strategies for mitigating the underreach/overreach issues in the
distance relays.
The challenge in the distance protection in the presence of
FACTS controllers such as UPFC, is due to the variation of
its control parameters namely, the series voltage and reactive
current injections. These protection issues can be addressed ei-
ther through the development of dedicated schemes for fault
location in transmission lines provided with UPFC [6], [7] or
through the adaptive prediction of trip boundaries of the conven-
tional distance relays. The development of dedicated schemes
[6], [7] has attracted much more attention as compared to the
adaptive trip boundary prediction. As a typical distance relay
consists of fault detection, classification, measurement and trip
region comparison units, replacing the fixed Mho/quadrilateral
trip region comparison unit with an adaptive trip characteris-
tics is an effective option for dealing with the protection is-
sues in the presence of an UPFC. One such scheme has been
proposed by the authors’ in [8] and was demonstrated only for
Single-Line-to-Ground (SLG) fault.
This paper proposes a scheme to predict the trip boundaries of
a conventional distance relay in the presence of UPFC through
the knowledge of the control parameters of the UPFC. It com-
putes the series voltage and reactive current injection by the
UPFC on-line with the help of synchronized phasor measure-
ments [9] and these parameters are utilized in the adaptive trip
boundary prediction. Additionally, the scheme also considers
the fact that depending on the magnitude of the fault current,
the UPFC may change its status to bypass operating mode [10],
where series voltage injection is zero.
The proposed scheme is based on Artificial Neural Network
(ANN), which has found applications in adaptive reach settings
of the conventional distance relays [11], [12]. To adapt to the
changes in the system, external to the protected line, an adap-
tive reach setting concept is presented using three Back Prop-
agation Neural Networks (BPNN) in [11] and by using three
Radial Basis Function Neural Networks (RBFNN) in [12], uti-
lizing local signals. These ANN structures suffer from the draw-
back that they require large training time for learning the several
possible combinations of the operating scenarios of the system.
SEETHALEKSHMI et al.: SYNCHROPHASOR ASSISTED ADAPTIVE REACH SETTING
For overcoming the above drawback, the proposed scheme uti-
lizes a single Generalized Regression Neural Network (GRNN)
[13]. Pre-fault system data along with the impedance measured
by the relay unit have been utilized by the GRNN for the es-
timation of the trip characteristics under various faults namely,
Single Line-to-Ground (SLG), Double Line-to-Ground (LLG)
and Three Phase-to-Ground (LLLG) faults.
The proposed scheme has been tested on 39-bus New Eng-
land system [14] and also on a reduced equivalent of Northern
Regional Power Grid (NRPG) system, India [15], [16]. Trip
boundaries generated by the GRNN for various types of faults
have been compared with the conventional Mho relay charac-
teristics and also the capability of the GRNN to approximate
the function has been compared with that of BPNN. Test results
indicate that the proposed GRNN architecture is able to effec-
tively track the relay trip boundary with the pre-fault operating
conditions along with the pre-fault control settings of the UPFC.
II. IMPEDANCE ANALYSIS WITH UPFC
A. Unified Power Flow Controller (UPFC)
An UPFC uses a combination of a shunt controller and a
series controller interconnected through a common dc bus. Both
series and shunt controllers use a Voltage Source Converter
(VSC) [17]. One VSC, connected in shunt to the transmission
line through a shunt transformer, is operated as a Static Syn-
chronous Compensator (STATCOM). It controls the ac voltage
at its terminals and also the dc bus voltage. Reactive current
injection is self-adjusted to maintain the voltage at the bus,
where the shunt branch is connected. Another VSC, connected
to the system through a series transformer, is used to control the
active and reactive power flow in the line. The series controller
can operate either in automatic power flow control mode or
manual voltage injection mode.
As installation of the UPFC requires a sub-station, in practice,
it may either be located at the sending end or the receiving end
of the line. In this study, the UPFC is located near to the sending
end of the line.
B. Impedance Measured by the Relay in the Presence of UPFC
The effect of the UPFC on apparent impedance seen by the
relay mainly depends on its location with respect to the fault
point and its operating mode under the fault. The most common
operating mode of the UPFC is automatic power flow control
mode. In this mode of operation, series injection voltage is dy-
namically adjusted to maintain the power flow in the line to the
reference settings. UPFC also assumes a self protection mode
called bypass mode, in which power flow can not be allowed
through the series controller beyond a certain level. Shunt con-
troller may or may not be connected, if it is intended to provide
a voltage regulation feature [10].
1) Automatic Power Flow Control Mode of UPFC: Fig. 1
shows an UPFC installed near a relay point in a line. In gen-
eral, where the UPFC is included in the fault loop, the measured
voltage at the relay consists of the voltage drop in the line, series
voltage injection in the line, line drop due to the shunt current
injection in the line and the voltage drop in the fault resistance.
397
Fig. 1. Transmission line - with UPFC located near the relay end.
For a SLG fault, voltage measured at the relay point will be as
follows:
(1)
where and denote the positive and zero sequence phase
currents and and stand for the positive and zero se-
quence impedance of the transmission line. Series voltage
source of the UPFC is represented by , ( ).
Voltage source of shunt branch is denoted by ,( )
and current injected by the shunt branch as respectively and
“ ” represents fraction of the line length to the fault location
from the relay point. Fault resistance and fault current are
denoted by and , respectively.
Apparent impedance in (1), is affected by the se-
ries voltage and the shunt current injections by the UPFC apart
from the fault resistance and shunt capacitance of the line. A
fixed level compensation, based on the initial series voltage in-
jection level and reactive current injection, is not feasible as
these parameters are continuously changing in order to fulfill
the control objectives. Hence, this work proposes to compute
the control parameters of the UPFC on-line for ensuring the
satisfactory operation of the conventional distance relay for the
faults in its primary and backup zones.
2) Bypass Mode of UPFC: In the bypass mode, (1) gets mod-
ified as series voltage injection by the UPFC, becomes zero.
Hence, the measured voltage, at the relay location, becomes
(2)
As this mode of operation causes apparent impedance to
change considerably, the proposed scheme has considered this
mode also while deriving the adaptive relaying strategy. The
distance relays in the studies are assumed to be self–polarized.
III. COMPUTATION OF CONTROL SETTINGS OF UPFC
The control settings of the UPFC, under a given operating
condition, are generally obtained using state estimation proce-
dures [18]. However, the state estimation procedures involve
nonlinear optimization techniques, which are time consuming.
With the aid of Phasor Measurement Unit (PMU) based Wide
Area Monitoring Systems (WAMS) [19], it is possible to get
the information exchange between the line terminals. So, UPFC
control settings namely, series voltage injection and reactive
398 IEEE SYSTEMS JOURNAL, VOL. 5, NO. 3, SEPTEMBER 2011

Equation (8) can be rewritten as

(11)
Define

Fig. 2. Equivalent circuit of line - with UPFC. (12)

(13)
current injection can be directly computed [20] from the con- (14)
verged load flow results. Equivalent circuit of a transmission
(15)
line, having a UPFC, is shown in Fig. 2.
Power flow received at bus can be obtained from the basic
Then (11) can be rewritten as
power flow equations as

(16)
(3)

Reactive power flow in line - is

where and are the half-line charging suseptance and
impedance of the line - , respectively:

Equation (3) can be rewritten as (17)

where is the phase angle difference between

(4) buses and . Reactive current injection, , can be computed
using (17) utilizing the transmission line terminal quantities.
or
IV. GRNN BASED ADAPTIVE DISTANCE RELAYING SCHEME
(5)
An ANN based adaptive scheme has been developed to deter-
where mine the ideal trip boundary settings of digital distance relays
that automatically takes care of variations in control parameters
of the UPFC, network conditions and fault resistance. The ar-
(6) chitecture of the ANN, selected in this work, is the Generalized
Regression Neural Network (GRNN). Generalized Regression
(7) Neural Network (GRNN)

Using (5), (6) and (7), series voltage injection of the UPFC can
be calculated with the knowledge of the receiving end power
flows, sending and receiving end voltage profiles and angular
separations. In the above expressions, the power loss in the
UPFC is neglected.
Similarly, reactive current injection can also be calculated
from the basic power flow equations [21].
(8)
Let
(9)
and
A. GRNN- An Overview
GRNNs can model nonlinear functions and are shown to
perform well in noisy environments, given enough data [13],
and its training is faster. Though GRNN is a feed forward net-
work requiring supervised training, it is significantly different
in architecture and algorithm to the BPNN. It involves one-pass
neural network learning algorithm, which can be used for
estimating the continuous variables and does not require an it-
erative training procedure as in the BPNN. It approximates any
arbitrary function between input and output vectors, drawing
the function estimate directly from the training data. Let “ ”
be a particular measured value of the random variable “ .” If
a joint probability distribution between a set of variable “ ”
and a dependent variable “ ” exists, then the estimation of “ ,”
given a particular realization of “ ,” is expressed as

(10) (18)
SEETHALEKSHMI et al.: SYNCHROPHASOR ASSISTED ADAPTIVE REACH SETTING
Fig. 3. Architecture of GRNN [13].
When the density function is not known, it can usually
be estimated from the training set using Parzen’s nonparametric
estimator [13].
B. Architecture of the GRNN
The architecture of the GRNN is shown in Fig. 3. It includes
four layers: the input layer, the pattern layer, the summation
layer and the output layer. The number of inputs is equal to the
number of independent features. The function of the input layer
is to pass forward the activity patterns presented to the network
to all the neurons in the pattern layer. The neurons in the pattern
layer perform a nonlinear transformation of the input patterns. In
this scheme, radial basis function has been chosen in the pattern
layer and, hence, the layer units represent the centers of clus-
ters of the training data. The output from all the neurons in the
pattern layer, then, becomes input to all the neurons in the sum-
mation layer. For a single output network, as shown in Fig. 3, the
summation layer consists of a single denominator neuron and a
single numerator neuron. For each additional output unit, a nu-
merator neuron is added. The summation unit performs a dot
product between a weight vector and a vector composed of the
signals from the pattern units. For the denominator summation
neuron, the weight vector is unity. For the numerator summation
neuron, the weight connecting it to each pattern layer neuron
is equal to the value of the dependent variable for the training
case of that pattern layer neuron. The outputs from the denom-
inator and numerator summation neurons are sent to the output
layer neurons, the function of which is to divide the output of
the associated numerator summation neuron by the output of
the denominator summation neuron. In this scheme, GRNN has
4 neurons in the output layer and 5 neurons in the summation
layer.
The primary advantage of the GRNN is the speed at which the
network can be trained. There are no training parameters, such
as learning rate and momentum as in the case of the BPNN, but
there is a smoothing factor that is applied after the network is
trained. The smoothing factor allows the GRNN to interpolate
between patterns or spectra in the training set and it usually
ranges between 0.01–1. In this scheme, smoothing factor has
399
been determined experimentally, by measuring the closeness of
the predicted value to the target. For SLG, LLG, and LLLG
faults, a separate GRNN has been used. They have the same
structure and differ only in the smoothing factor.
As the GRNN does not require an iterative training, any new
training pattern corresponding to the changes in the system con-
figuration can be easily incorporated, which is vital to the prob-
lems dealing with the power system security.
C. Selection of the Input Features
Inputs for the GRNN are selected based on an eigen value
[21] based Principal Component Analysis (PCA) performed on
a set of inputs namely, complex voltages at the line terminals,
complex sending end current, UPFC control parameters and the
apparent impedance. From variances in the principal compo-
nents, it is seen that the first six principal components contribute
towards 99.6% of the variances in the data. Hence, from the
six principal component vectors, the features corresponding to
the largest coefficients have been selected. Such selected inputs
are complex voltages at line terminals ( in Fig. 1), com-
plex series injection voltage of the UPFC ( in Fig. 1), reac-
tive current injection by the UPFC ( in Fig. 1) and the resis-
tance/reactance values corresponding to each boundary. These
inputs, along with the UPFC mode of operation (Status 0-auto-
matic power flow control mode and Status 1- bypass mode) are
used by the GRNN for generating the trip characteristics. Out-
puts of the GRNN correspond to the resistance/reactance value
for each boundary. It has 12 input neurons and four output neu-
rons.
1) Adaptive Algorithm: The proposed adaptive scheme
predicts the trip boundaries of the distance relay utilizing the
real time system data. The algorithm utilizes the latest prefault
data available at the relay location along with the apparent
impedance measured by the relay and status of the UPFC. The
apparent impedance is computed using Full Cycle Discrete
Fourier Transform (FCDFT) based distance relay algorithm.
The overall adaptive protection scheme is as follows.
a) Relay receives pre-fault local and remote PMU data
through communication links.
b) UPFC control settings are computed as given in
Section III.
c) After the fault detection and classification by the corre-
sponding blocks of the distance relay (usually carried out
within one cycle), the pre-fault system state data, along
with the apparent impedance measured by the relay fault
measurement unit (normalized for each boundary) and the
UPFC status are given as inputs to the trained GRNN.
d) GRNN predicts the trip boundary impedance values cor-
responding to the inputs in (c).
e) Trip signal is generated based on the comparison between
boundary values and the measured impedance.
Post fault measurement needed for the scheme to predict
the trip boundaries is only the impedance measurement by the
relay. Digital distance relays with FCDFT measures the fault
impedance in one cycle. Hence, the scheme is able to generate
the trip decision one cycle after the fault initiation.
One may think of a differential protection scheme for a line
where the communication links are already established between
400
Fig. 4. New England system with UPFC in line 16–17.
the line terminals. The advantage of the proposed scheme, com-
pared to the differential protection scheme, is that this scheme
can very well be extended to the protection of the adjacent line
sections, which are falling in the zone 2 and the zone 3 limits
of the relay. So, with the PMUs installed at both the ends of the
line, the pre-fault information, which is utilized to obtain con-
trol settings of UPFC, can be used for identifying all the three
trip zones of a conventional distance relay.
2) WAMS Architecture and Communication Requirements:
Modern distance relays such as SEL-421 of Schweitzer Engi-
neering and D60 of General Electric are having PMU capability
to provide time stamped voltage and current phasors. Pre-fault
system data, required by the proposed scheme, can be made
available at a frame rate of 60 frames per second in 60 Hz system
and 50 frames per second in 50 Hz system [22] from the dis-
tance relays (or PMUs) installed in the line itself. This scheme
can work with a “flat architecture” [15] based WAMS, in which
relevant system data is transferred through the communication
system that ties the line terminals together. The communication
delay in the present PMU based scheme does not pose a problem
in the scheme, as it deals with only pre-fault quantities at the
relay location.
V. SIMULATION RESULTS AND DISCUSSIONS
The effect of the dynamical changes in the UPFC settings
on the distance protection scheme has been studied on two test
systems, namely 39-bus New England system and a 17-bus re-
duced equivalent of Northern Region Power Grid (NRPG) net-
work. Test systems are modeled in SIMULINK [23]. Loads are
modeled as constant impedance type for the transient studies.
Dynamic model of the UPFC [23], has been considered in this
IEEE SYSTEMS JOURNAL, VOL. 5, NO. 3, SEPTEMBER 2011
study. Constraint imposed on the operation of the UPFC in auto-
matic power flow control is on the variation in the magnitude of
its series injected voltage. The limits on series injection voltage
are taken as 0–10% of the voltage value at the sending end and
the phase angle variation between 0–360 . In the bypass mode
of operation of the UPFC, both the series and the shunt con-
trollers are bypassed; hence, the series injection voltage and the
reactive current injection by the UPFC are zero.
A. 39-Bus New England System
Placement of the UPFC in a network enhances the transmis-
sion capability of the interconnected system. From a system
point of view, though it is necessary to consider both the static
as well as dynamic aspects for deriving a placement strategy for
a FACTS controller [19], this work has considered static Avail-
able Transfer Capacity (ATC) enhancement as a criterion [24] to
decide the optimal placement of the UPFC. The UPFC is placed
in line 16–17 as shown in Fig. 4, with shunt branch at bus 16. Ini-
tial reactive current injection of the UPFC is maintained at zero
for all the cases as the system is having sufficient reactive power
support. However, the dynamic simulations have considered the
reactive current injection or absorption modes of the UPFC in
response to the faults. In the present study, a SLG fault, a LLG
fault and a LLLG fault, with fault resistance , have been sim-
ulated on the line connecting buses 16 and 17. Distance relays
are assumed to be located at stations 16 and 17.
UPFC is assumed to be initially operating in the automatic
power flow control mode. For all the cases studied, reference
power flow setting of the UPFC is MVA from bus 16
to 17, which has been decided after analyzing the system for
various loading scenarios and contingency cases such as outage
of the line and generation units. However, this is only a sample
SEETHALEKSHMI et al.: SYNCHROPHASOR ASSISTED ADAPTIVE REACH SETTING
Fig. 5. Variation in the UPFC control parameters during SLG fault. (a): Vari-
ation in series voltage injection during fault. (b): Variation in reactive current
injection during fault.
setting for the study purpose and it can be decided by the utility
depending on the actual system requirements. Base case loading
of the system corresponds to a flow of MVA in the
line, from bus 16 to bus 17 and the power flow is called as Pre
Controlled Power Flow (PCPF).
1) UPFC Transient Response in Power Flow Control Mode:
As stated in Section II, the series voltage injection undergoes
changes during fault conditions. Fig. 5(a) shows the change in
the series voltage injection, when a SLG fault takes place in
the line 16–17 at 50% distance from the bus 16 with a fault
resistance of 10 ohm. The UPFC control is applied at for
maintaining the power flow at the reference setting, followed by
a SLG fault applied at after attaining the required power
flow. In Fig. 5(a) series voltage injection has been increased to
0.06 p.u. at and following the SLG fault at
reduced to 0.055 p.u.
Similar analysis has been done for the reactive current injec-
tion from the UPFC. Fig. 5(b) shows the variation in the reactive
current injection ( ) by the UPFC following the SLG fault. Fol-
lowing the fault, increases to 1.0 p.u, due to which the mea-
sured value of the apparent impedance increases. However, the
influence of on the distance relay operation is dependent on
the control mode of the UPFC.
2) Apparent Impedance Boundaries for the 39-Bus System:
Apparent impedance, measured at the relay point, is depen-
dent on the operating states of the system. This is illustrated in
Fig. 6(a) where trip boundaries of the relay under various power
flow conditions are shown with UPFC operating in automatic
power flow control mode.
Towards realizing these trip boundaries of the relay at bus 16,
SLG fault is simulated at distances, ranging from 2% to 95%
401
Fig. 6. Trip boundaries with the UPFC in 39-bus system. (a): Boundaries with
UPFC in automatic power flow control mode. (b): Trip boundaries of the relay
with UPFC in bypass mode.
from the relay location, with the fault resistance from 0.01 to
100 ohm. Fig. 6(a) shows the four trip boundaries indicated as
“boundary a,” “b,” “c,” and “d.” Boundaries “a” and “c” corre-
spond to the fault distance of 95% and 2%, respectively, from
the relaying location, when the fault resistance is varied from
minimum to maximum value. Boundaries “b” and “d” corre-
spond to the maximum and minimum fault resistances, respec-
tively, where the distance at which fault occurs vary from 2%
to 95% of the line length. Plotted boundaries correspond to the
impedance measured using FCDFT by taking the voltage and
current signals from the relay location.
In the bypass mode, both the series and the shunt controllers
of the UPFC are bypassed within 1 cycle of the fault initiation
and steady state apparent impedance is measured for the SLG
fault at various distances and fault resistances. Fig. 6(b) shows
the corresponding trip boundaries in various power flow condi-
tions.
For SLG, LLG and LLLG faults, variation in the trip bound-
aries of the distance relay located at bus 16, with the UPFC in
automatic power flow control mode, are shown in Fig. 7 along
with the stepped zone trip characteristics of a conventional mho
and quadrilateral relays. Settings of the mho and quadrilateral
relays have been selected based on the standard practice of 85%
of line impedance for zone 1, 120% for zone 2 and zone 3 set-
ting to cover the longest adjacent line section. From Fig. 7, it is
402
Fig. 7. Comparison of the boundaries with conventional characteristics. (a):
Comparison with conventional mho characteristics. (b): Comparison with
quadrilateral characteristics.
evident that the conventional fixed settings followed in the re-
lays are not capable to cope up with the dynamic changes in the
impedance due to presence of the UPFC.
3) Training Data for GRNN for 39-Bus System: The ref-
erence power setting of the UPFC in the line 16–17 is taken
as MVA at the sending end and the training patterns
were generated for the PCPF values ranging from 10 to 80 MW
with the reactive power ranging from to 0 MVAR through
the simulation of various faults. Training cases considered are
shown in Table I. Corresponding to each case, UPFC control
parameters are evaluated. Through the data obtained from the
time domain simulations, the trip boundaries are generated for
different faults in line 16–17 with the UPFC operating in the au-
tomatic power flow control mode and also in the bypass mode.
Inputs to the GRNN are 12 in number namely, the complex
bus voltages at buses 16 and 17 , complex series
voltage injected by the UPFC , reactive current injection
by UPFC ( ) status of the UPFC (0/1) along with resistance ( )
at different points along boundary ’a’ and ’c’ and reactance ( )
at different points along boundary “b” and “d.” Output of the
GRNN is the corresponding reactance values for the boundaries
“a” and “c” and resistance values for the boundaries “b” and “d.”
The above input-output pairs are utilized in training the separate
GRNN for each type of fault.
IEEE SYSTEMS JOURNAL, VOL. 5, NO. 3, SEPTEMBER 2011
TABLE I
DETAILS OF TRAINING CASES FOR 39-BUS SYSTEM
TABLE II
DETAILS OF TEST CASES FOR 39-BUS SYSTEM
4) Test Results for 39-Bus System: Testing data for the
GRNN consisted of those patterns, which were not considered
for the training. The pre-fault system data for the testing cases
are given in Table II. The trip boundaries, identified through
the GRNN for SLG, LLG and LLLG faults, are presented in
Fig. 8(a)–(c), respectively, for UPFC in automatic power flow
control mode (status-0) as well as in bypass mode (status-1).
The generated boundaries are in close approximation with the
ideal trip boundaries generated through the simulation as can be
seen in Fig. 8.
As the BPNN has earlier been utilized for pattern recogni-
tion [12] in relays, the boundaries generated by the GRNN are
compared with that by the BPNN for the SLG, LLG and LLLG
faults in Fig. 9(a)–(c), respectively. The results reveal that the
selected architecture of the GRNN tracks the boundaries more
closely as compared to the BPNN in all the test cases. However,
the major advantage of the GRNN compared to the BPNN is
due to GRNN’s one pass training. Training time is a vital factor
in protection applications as new learning is to be incorporated
very frequently to ensure the reliability of the scheme. Hence,
GRNN is more advantageous in this application as compared to
the BPNN.
B. Reduced Equivalent of NRPG System
Adaptive scheme for generating relay boundaries has also
been tested on a 17-bus reduced equivalent [18] of practical
NRPG system in India. In the NRPG system, a TCSC exists
in line 239–233 with variable compensation of 8%–20% of the
line reactance. This series compensation helps in the increased
power transfer through the line 239–233 from the existing level
SEETHALEKSHMI et al.: SYNCHROPHASOR ASSISTED ADAPTIVE REACH SETTING 403

Fig. 8. Trip boundaries predicted by the GRNN in 39-bus system. (a): for SLG
fault – Cases C and D. (b): for LLG fault – Cases C and B. (c): for LLLG fault
– Cases A and B.

of 400–420 MW to approximately 500–550 MW. These prac-

tical operating conditions have been considered for selecting the
location of the UPFC and also to decide the reference power set-
ting of the UPFC. An UPFC is assumed to be located in the line
239–233, replacing the TCSC, with shunt branch of the UPFC
connected at bus 239. Reference power settings of the UPFC
have been taken as MVA. The base case power flow
in the line is MVA.
1) Apparent Impedance Boundaries for the NRPG System:
Apparent impedance boundaries have been generated for the
distance relay located at bus 239, through the simulation of dif-
ferent types of faults at various distances of the protected line
for both the operating modes of the UPFC. In this system, vari-
ation in the trip boundaries has also been considered with the
shunt branch of the UPFC operating in the reactive power gen-
eration and absorption modes. When the UPFC operates with
its shunt branch in the reactive power generation mode, the in-
jected reactive current ’ ’ is positive that causes the relay to
Fig. 9. Comparison of trip boundaries predicted by GRNN and BPNN. (a):
Comparison of boundaries- for SLG fault (Case A). (b) Comparison of bound-
aries- for LLG fault (Case C). (c ) Comparison of boundaries- for LLLG fault
(Case C).
under-reach. Similarly, reactive power absorption mode of the
shunt controller causes the relay to see a lower value of the
impedance and, hence, relay over-reaches. This effect has been
demonstrated in Fig. 10, wherein trip boundaries are shown for
three cases with the PCPF value equal to MVA.
Training patterns are generated by considering the possible
operating scenarios of the system. Variation of the PCPF is con-
sidered from 343.7 MW to 601 MW in line 239–233. Corre-
404
Fig. 10. Variation in trip boundaries with changes in “ ” injection of the
UPFC (NRPG system).
Fig. 11. Trip boundaries generated by GRNN for NRPG system for SLG fault.
TABLE III
DETAILS OF TEST CASES FOR NRPG SYSTEM.
sponding to each power flow condition, varying reactive power
injection from the shunt branch ( ) is considered.
2) Test Results for the NRPG System: Testing patterns consist
of all the four categories i.e., UPFC in power flow control mode
with minimal, negative and positive injections along with the
UPFC in bypass mode of operation. PCPF in the test case has
been taken as MVA, which is slightly outside the
range of the PCPF values used for training the GRNN. Details
of the test cases are given in Table III and the performance of
the GRNN is shown in Fig. 11. It is seen that the GRNN is able
to approximate the function in an acceptable way.
Test results indicate that the proposed GRNN architecture is
able to track the relay trip boundaries utilizing the pre-fault op-
erating conditions and the control settings of the UPFC.
IEEE SYSTEMS JOURNAL, VOL. 5, NO. 3, SEPTEMBER 2011
VI. CONCLUSION
In this paper, a new scheme based on Generalized Regres-
sion Neural Network (GRNN) has been proposed to predict the
trip boundaries of a distance relay in an adaptive manner, in the
presence of UPFC. Data sets have been generated for training
the GRNN through an extensive analysis of the system for var-
ious types of faults viz. Single-Line-to-Ground (SLG), Double-
Line-to-Ground (LLG) and Three- Phase-to-Ground (LLLG)
faults, with the UPFC in automatic power flow control mode
as well as in the bypass mode of operation. GRNN has been
trained with the pre-fault system states, control parameters of
the UPFC, impedance measured by the relay unit and operating
status of the UPFC as inputs. Control settings of the UPFC have
been calculated with the aid of synchro phasor measurements,
assumed to be available from the Phasor Measurement Units
(PMUs) placed at the ends of the line. The test results show
that, with a single GRNN for each type of fault, the trip region
can be accurately recognized for automatic power flow control
as well as the bypass modes of operation of the UPFC. Trip
boundaries for the relay in the presence of UPFC are compared
with that of the conventional mho and quadrilateral character-
istics which clearly indicate the need of replacing the conven-
tional settings with adaptive settings. Also, the performance of
GRNN has been found superior to that with a Back Propagation
Neural Network (BPNN) for generating the trip boundary. By
replacing the fixed type relay settings with the proposed adap-
tive settings, the capability of the conventional distance relay
can be enhanced to cope up with the dynamic variation in the
line impedance in the presence of the UPFC.
ACKNOWLEDGMENT
The authors wish to thank Department of Science and Tech-
nology, New Delhi, India for providing support to carry out this
research work.
REFERENCES
[1] N. G. Hingorani and L. Gyugyi, Understanding FACTS: Concepts &
Technology of Flexible AC Transmission Systems. New York: IEEE,
2000.
[2] F. A. Albasri, T. S. Sidhu, and R. K. Verma, “Performance comparison
of distance protection schemes for shunt-FACTS compensated trans-
mission lines,” IEEE Trans. Power Del., vol. 22, no. 4, pp. 2116–2125,
2007.
[3] P. K. Dash, A. K. Pradhan, G. Panda, and A. C. Liew, “Adaptive relay
setting for flexible AC transmission systems (FACTS),” IEEE Trans.
Power Del., vol. 15, no. 1, pp. 38–43, 2000.
[4] X. Zhou, H. Wang, R. K. Aggarwal, and P. Beaumont, “Performance
evaluation of a distance relay as applied to a transmission system with
UPFC,” IEEE Trans. Power Del., vol. 21, no. 3, pp. 1137–1147, 2006.
[5] M. Khederzadeh, “UPFC operating characteristics impact on transmis-
sion line distance protection,” in Proc. IEEE PES General Meeting,
Pittsburgh, PA, Jul. 2008, pp. 1–6.
[6] S. R. Samantaray, “Decision tree-based fault zone identification and
fault classification in flexible AC transmissions-based transmission
line,” Proc. IET Gen. Transm. Distrib., vol. 3, no. 5, pp. 425–436,
2009.
[7] S. R. Samantaray, L. N. Tripathy, and P. K. Dash, “Differential
equation-based fault locator for unified power flow controller-based
transmission line using synchronized phasor measurements,” Proc.
IET Gen. Transm. Distrib., vol. 3, no. 1, pp. 86–98, 2009.
[8] K. Seethalekshmi, S. N. Singh, and S. C. Srivastava, “Adaptive dis-
tance relaying scheme in presence of UPFC using WAMS,” in Proc.
IEEE Power Systems Conf. Exposition, Seattle, WA, Mar. 2009, pp.
1–6.
SEETHALEKSHMI et al.: SYNCHROPHASOR ASSISTED ADAPTIVE REACH SETTING 405

[9] A. G. Phadke, “Synchronized phasor measurement in power systems,” K. Seethalekshmi (S’09) received the B.Tech
IEEE Comput. Applicat. Power, vol. 6, no. 2, pp. 10–15, 1993. degree from Regional Engineering College, Calicut,
[10] L. Gyugyi et al., “Operation of the unified power flow controller Kerala, India, the M. Tech. degree from the College
(UPFC) under practical constraints,” IEEE Trans. Power Del., vol. 13, of Engineering, Trivandrum, Kerala, and is currently
no. 2, pp. 630–639, 1998. pursuing the Ph.D. in the Department of Electrical
[11] K. K. Li, L. L. Lai, and A. K. David, “Stand alone intelligent digital Engineering, Indian Institute of Technology Kanpur.
distance relay,” IEEE Trans. Power Syst., vol. 15, no. 1, pp. 137–142, Her research interests include wide area monitoring
2000. systems, synchrophasor applications in power
[12] B. R. Bhalija and R. P. Maheshwari, “High- resistance faults on two system protection and control, and optimal power
terminal parallel transmission lines: analysis, simulation studies, and flow.
an adaptive distance relaying scheme,” IEEE Trans. Power Del., vol.
22, no. 2, pp. 801–812, 2007.
[13] D. F. Specht, “A general regression neural network,” IEEE Trans.
Neural Netw., vol. 2, no. 6, pp. 568–576, 1991.
[14] M. A. Pai, Energy Function Analysis for Power System Stability. S. N. Singh (SM’02) received M.Tech and Ph.D. de-
London, U.K.: Kluwer, 1989. grees from the Indian Institute of Technology Kanpur
[15] Northern Regional Load Dispatch Center Website, [Online]. Available: in 1989 and 1995, respectively.
http://nrldc.org Presently, he is Professor in the Department of
[16] M. Arunachalam, G. Lal, C. G. Rajiv, and M. M. B. Narayanan, “Per- Electrical Engineering, Indian Institute of Tech-
formance verification of TCSC control and protection equipment using nology Kanpur. His research interests include
RTDS,” in 15th Power Systems Computation Conf., Liege, Aug. 2005. power system restructuring, FACTS, power system
[17] T. Jain, S. N. Singh, and S. C. Srivastava, “Dynamic ATC enhancement optimization and control, wind power, etc. He is a
through optimal placement of FACTS controllers,” Elect. Power Syst. Fellow of IE (India) and IETE (India).
Res., vol. 79, no. 11, pp. 1473–1482, 2009.
[18] B. Xu and A. Abur, “State estimation of systems with UPFCs using
the interior point method,” IEEE Trans. Power Del., vol. 19, no. 3, pp.
1635–1641, 2004.
[19] J. Bertsch, C. Carnal, D. Karlsson, J. Mcdaniel, and K. Vu, “Wide-area
S. C. Srivastava (SM’91) received the Ph.D. degree
protection and power system utilization,” Proc. IEEE, vol. 93, no. 5,
in electrical engineering from the Indian Institute of
pp. 997–1003, 2005.
Technology Delhi.
[20] G. Shaoyun, “Optimal Power System Operation and Control Incorpo-
Presently, he is Professor in the Department of
rating FACTS Devices,” Ph.D., Hong Kong Polytechnic Univ., Dept.
Electrical Engineering, Indian Institute of Tech-
Elect. Eng., Hong Kong, 1998.
nology Kanpur. His research interests include
[21] K. Fukunaga, Introduction to Statistical Pattern Recognitions. New
synchrophasor technology, stability and controls,
York: Academic, 1972.
and technical issues in electricity markets. He is a
[22] IEEE Standard for Synchrophasors for Power Systems, IEEE Std. C37.
Fellow of INAE (India), IE (India) and IETE (India).
118, 2005.
[23] SimPowerSystems 4 – Reference Manual. Natick, MA, The Math-
Works, Inc., 2007.
[24] S. Kim, J. Lim, and S. Moon, “Enhancement of power system security
level through the power flow control of UPFC,” in Proc. IEEE PES
Summer Meeting, Seattle, WA, Jul. 2000, pp. 38–43.

View publication stats