IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 26, NO.

1, JANUARY 2011

109

Optimum Coordination of Directional Overcurrent

Relays Using the Hybrid GA-NLP Approach
Prashant Prabhakar Bedekar and Sudhir Ramkrishna Bhide

AbstractThe time of operation of overcurrent relays (OCRs)

can be reduced, and at the same time, the coordination can be
maintained, by selecting the optimum values of time multiplier setting (TMS) and plug setting (PS) of OCRs. This paper presents hybrid genetic algorithm (GA)nonlinear programming (NLP) approach for determination of optimum values of TMS and PS of
OCRs. GA has a drawback of, sometimes, converging to the values
which may not be optimum, and NLP methods have a drawback of
converging to local optimum values, if the initial choice is nearer
to local optimum. This paper proposes a hybrid method to overcome the drawback of GA and NLP method, and determine the
optimum settings of OCRs. The main contributions of this paper
are1) systematic method for formulation of problem of determining optimum values of TMS and PS of OCRs in power distribution network as a constrained nonlinear optimization problem,
2) determining initial values of TMS and PS using GA technique
and finding final (global optimum) values using NLP method, thus
making use of the advantages of both methods (and at the same
time overcoming the drawbacks of the methods).
Index TermsConstrained optimization, genetic algorithm
(GA), nonlinear programming, overcurrent relay coordination.

I. INTRODUCTION
IRECTIONAL overcurrent relays (OCRs) have been
commonly used as an economic alternative for the
protection of subtransmission and distribution system or as a
secondary protection of transmission system [1]. To reduce the
power outages, mal-operation of the backup relays should be
avoided, and therefore, OCR coordination in power distribution
network is a major concern of protection engineer. A relay
must get sufficient chance to protect the zone under its primary
protection. Only if the primary protection does not clear the
fault, the back-up protection should initiate tripping. Each
protection relay in the power system needs to be coordinated
with the relays protecting the adjacent equipment. The overall
protection coordination is thus very complicated [2].
Several optimization techniques have been proposed for optimum coordination of OCRs [1][18]. In [2] and [3], the relay
coordination problem is formulated as mixed integer non-linear
programming (MINLP) and is solved using General Algebraic
Modeling System (GAMS) software. To avoid the complexity

Manuscript received May 13, 2010; revised July 31, 2010; accepted
September 12, 2010. Date of publication November 01, 2010; date of current
version December 27, 2010. Paper no. TPWRD-00349-2010.
The authors are with the Department of Electrical Engineering, Visvesvaraya National Institute of Technology, Nagpur 440 010, India (e-mail:
[email protected]; [email protected]).
Digital Object Identifier 10.1109/TPWRD.2010.2080289

of MINLP technique, the OCR coordination problem is commonly formulated as a linear programming problem (LPP). Various LPP techniques have been used in [4][6] for OCR coordination. In [7] and [8] optimum coordination has been obtained
considering the configuration changes of the network into account. In [9][11], GAs are proposed to find the optimum solution for relay settings. Directional OCR coordination problem
has been solved in [1] using hybrid GA, considering the effects of the different network topologies. In [12] also system
topology changes have been considered and an adaptive scheme
has been suggested for relay coordination. A new nonstandard
tripping characteristic for OCRs and its advanced method for
optimized coordination have been presented in [13]. In [14],
it has been proposed to introduce additional constraints in the
directional OCR coordination problem to tackle the sympathy
trips in which other relays in the system operate earlier than
the designated primary relay. A method of simultaneously optimizing all the settings of directional OCR, in non-linear environment by sequential quadratic programming method has
been presented in [15]. Non-linear Random Search Technique to
solve the coordination problem has been presented in [16] and
it has been shown that it is possible to achieve the acceptable
speed of primary protection while attempting to coordinate the
maximum relay pairs. A review of time coordination of OCRs
has been presented in [17].
The problem of optimum coordination of OCRs is generally
formulated as an LPP, in which pickup current settings are assumed to be known and the operating time of each relay is considered as a linear function of its TMS [1].
Instead of keeping the value of PS as fixed and determining
the optimum value of TMS, it is possible to select the optimum
values of both TMS and PS, which can reduce the time of operation and also maintain the coordination, of OCRs. Noghabi
et al. have presented hybrid GA method to determine optimum
TMS and PS of OCRs. They have used GA to determine PS of
relays and with this PS of relays the coordination problem is
converted to LPP, which is solved using standard LPP method
to determine the TMS of relays [1].
In this paper the problem of determining the optimum values
of TMS and PS of OCRs is formulated as a nonlinear programming problem (NLPP), and hybrid GA-NLP approach is
proposed to find the optimum solution. GA has a drawback
of, sometimes, converging to the values which may not be optimum, and NLP methods, being single point search methods,
have a drawback of being trapped in local optimum point, if the
initial choice is nearer to the local optimum. NLP method gives
global optimum solution, if proper initial choice is made. Due
to the fact that GA is a multipoint search method rather than the

0885-8977/$26.00 2010 IEEE

110

IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 26, NO. 1, JANUARY 2011

conventional single point search methods, GA searches a large

solution space. To make use of the advantages of GA and NLP
methods, and at the same time to overcome the drawbacks of
these methods, GA has been used to determine the initial value
of TMS and PS of OCRs. These values are then used as initial
choice in NLP method, which gives the global optimum solution.

B. Constraint Set IIBounds on Relay Operating Time

Howsoever fast we want the relay to operate; it needs a certain
minimum amount of time to operate. Also a relay should not be
allowed to take too long time to operate. Constraint imposed
because of restriction on the operating time of relays can be
mathematically stated as
(3)

II. GENERIC PROBLEM FORMULATION

where

The coordination problem of directional OCRs in a ring fed

distribution systems, can be stated as an optimization problem,
where the sum of the operating times of the relays of the system,
is to be minimized [2], [3], [5],
(1)

minimum operating time of relay at location i for

the fault at any point in the zone of operation;
maximum operating time of the relay at location i
for the fault at any point in the zone of operation.
C. Constraint Set IIIBounds on the TMS of Relays
of relays directly affect the operating time of reThe
of relays. It can be stated as
lays, which puts bounds on

where

(4)

number of relays,
operating time of the relay

, for fault at k;

weight assigned for operating time of the relay

where
.

In the distribution system, since the lines are short and are of
is assigned for
approximately equal length,
operating times of all the relays [1], [3], [4], [6].
The objective of minimizing the total operating times of relays is to be achieved under five sets of constraints, as discussed
in the following sections.

and

minimum value of
relay ;

of

maximum value of
relay ;

of

taken as 0.025 and 1.2,

respectively [18].

D. Constraint Set IVBounds on the PS of Relays

A. Constraint Set ICoordination Criteria

Fault is sensed by both primary as well as secondary relay simultaneously. To avoid mal-operation, the backup relay should
takeover the tripping action, only after primary relay fails to operate. The minimum operating time of backup relay is decided
by the operating time of primary relay, operating time of circuit
breaker (CB) associated with primary relay, plus the overshoot
time. This is necessary for maintaining the selectivity of primary
and backup relays. The sum of operating time of CB associated
with primary relay, and the overshoot time is called the coordination time interval (CTI).
is the primary relay for fault at k, and is backup relay
If
for the same fault, then the coordination constraint can be stated
as
(2)

The bounds on

of relays can be stated as

(5)

where
minimum value of

of relay

maximum value of

of relay

As a rule of thumb, the minimum pickup current setting is

equal to or greater than 1.25 times the maximum load current.
This is to ensure that the relays will not mal-operate under
normal load and small amount of overload conditions. Similarly
maximum pickup current setting is less than or equal to 2/3rd
of the minimum fault current [18]. This ensures that the relay
is sensitive to the smallest fault current.
E. Constraint Set VRelay Characteristics

where
operating time of the primary relay

, for fault at k;

operating time of the backup relay

fault (at k);

, for the same

CTI.

In this paper, a nonlinear and popular OCRs characteristic

function (which has been reported in most of the literature) as
shown in (6), has been considered. The values of and are
detailed in Table I [3], [4], [9], [18], [19]
(6)

BEDEKAR AND BHIDE: OPTIMUM COORDINATION OF DIRECTIONAL OVERCURRENT RELAYS

111

TABLE I
VALUES OF  AND FOR DIFFERENT TYPES OF OCRS

where
relay operating time;
current through the relay operating coil;
plug setting.

III. GENETIC ALGORITHM

Philosophically, GA is based on Darwins theory of survival of the fittest. GA is an optimization technique inspired by
the principles of natural evolution and natural selection. GA allows a population composed of many solutions to evolve under
specified selection rules to a state that maximizes the fitness
[20][22].
GA begins, like any other optimization algorithm, by defining
the optimization variables, and the fitness function (objective
function). It ends like other optimization algorithms too, by
testing for convergence. In between, however, this algorithm is
quite different [21]. A path through the GA is shown as a flowchart in Fig. 1. In the flow chart itrc, nitr, npair, and nmut indicate
iteration count, number of iterations to be performed, number
of pairs for mating, and number of mutations to be performed,
respectively. The specified number of iterations have been considered as the stopping criteria.
In GA, the design variables are represented as strings
,
of binary numbers, 0 and 1. If each design variable
, is coded in a string of length , a design
vector is represented using a string of total length . This is
achieved by placing the strings of all the variables side-by-side.
This string (string of total length ) is called a chromosome.
GA starts with a group of chromosomes known as population.
A. Representation of Design Variables (Parameters)
A continuous design variable can be represented by a set
of discrete values if binary representation is used. If a variable
; whose bounds are given by
and
; is represented by
a string of binary digits, its decimal value can be computed
using the following expression [20]:
(7)

Fig. 1. Flowchart of genetic algorithm.

where is the decimal (discrete value) equivalent of binary

number.
Thus, desired accuracy of the continuous variables can be
achieved, using more number of bits for representing the value
of variable in its binary representation.

112

B. Genetic Operators
The basic operations of natural geneticsreproduction,
crossover, and mutation, are implemented during numerical
optimization using GA.
Reproduction is a process in which the individuals are selected based on their fitness values relative to that of the population. The population is sorted according to the fitness of the
chromosome. The individuals (chromosomes) with higher fitness values are selected for mating and subsequent genetic action. Consequently, highly fit individuals live and reproduce,
and less fit chromosomes die. The reproduced chromosome,
which are used for mating are called parent chromosome.
After reproduction, the crossover operation is implemented. Crossover is an operator that forms two new chromosomes, called offsprings, from two parent chromosomes
by combining part of the information from each. Crossover
is implemented in two steps. First, two individual strings are
selected from the mating pool generated by the reproduction
operator. Next, a crossover site is selected at random along
the string length, and the binary digits are swapped between
the two strings following the crossover site. For example if the
chromosomes P and Q are selected for mating and a single
point crossover (with crossover site 4) takes place, two new
chromosomes (represented here as R and S) will be obtained as
shown

The offsprings obtained from crossover, along with the parent

chromosome, forms the new population.
The mutation is applied after crossover. A mutation is the
occasional, random alteration of a binary digit in a string. Thus
in mutation, a 0 is changed to 1, and vice versa, at a randomly selected location within the chromosome. For example, if chromois selected for mutation and mutation
some
takes place at bit three, a new chromosome
will be obtained which will then replace chromosome S.
C. Selecting GA Parameters
There is no deterministic method to select the optimum GA
parameters for a particular application. However some hints
given in the GA literature do help in deciding the proper choice
of GA parameters. The first intensive study of GA parameters
was done by De Jong in 1975 [20], [21]. He suggested that a
good GA performance requires the choice of a high crossover
probability, a low mutation probability, and a moderate population size [20]. Selecting a small population size takes a very
small amount of computer time. However, if the population
size is too small it may not sample the entire solution space
adequately.
In the binary GA the size of gene (number of bits used to
represent a variable) is important from the point of view of the
accuracy of the solution and the length of time needed for the
GA to converge. For every decimal point of accuracy needed in
the solution, the variable needs 3.3 bits in the gene [21]. A long
gene slows down the convergence of the GA. Hence the variables should be represented to the minimum accuracy needed.

IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 26, NO. 1, JANUARY 2011

The crossover rate determines the number of chromosomes

that enter the mating pool. If crossover rate is too high, good
building blocks dont have the opportunity to accumulate within
a single chromosome. A low crossover rate, on the other hand,
doesnt produce a sufficient number of new offspring [21]. A
crossover rate of 40 to 60% is a moderate choice.
When used sparingly with the reproduction and crossover operators, mutation serves as a safeguard against a premature loss
of important genetic material at a particular position [22]. The
constant mutation rate of 0.05 to 0.2 (5 to 20%) performs better
as compared to very small or high mutation rate [21]. As suggested by De Jong the mutation probability should be inversely
proportional to the population size.
D. Incorporating Constraints in GA
The GA basically find the optimum of an unconstrained
problem [20], [22]. To solve a constrained optimization
problem, we need to transform the original constrained problem
into an unconstrained problem. Transformation methods are the
simplest and most popular optimization methods of handling
constraints. The constraints can be included in the objective
function with the help of penalty method [23].
IV. NONLINEAR PROGRAMMING PROBLEM
In an optimization problem, if the objective function and/or
constraint/s are nonlinear, the problem is called NLPP. In case
of OCR coordination problem the relay characteristic, described
by (6), is nonlinear in nature because of which the objective
function, the operating time constraints, and the coordination
constraints become nonlinear.
Various methods are available to solve the constrained NLPP.
In this paper, the function available in MATLAB optimization
toolbox has been used to find the global optimum solution of
the relay coordination problem. The function uses a sequential
quadratic programming (SQP) method.
A. Sequential Quadratic Programming
Various techniques available for solution of a constrained
nonlinear programming problem can be broadly classified into
two categories as direct methods, in which the constraints
are handled in an explicit manner, and indirect methods, in
which the constrained optimization problem is solved as a
sequence of unconstrained optimization problems [22]. SQP is
a direct method and is one of the best methods of optimization.
SQP makes a quadratic model of the objective function using
Newtons method, and models the constraints as simultaneous
equations using Karush-Kuhn-Tucker (KKT) conditions to
the Lagrangian of the constrained optimization problem [22],
[24]. The problem approximated in this way is called quadratic
programming (QP) subproblem. For example, if the nonlinear
inoptimization problem, with equality constraints and
equality constraints, is stated as

subject to
and

(8)

BEDEKAR AND BHIDE: OPTIMUM COORDINATION OF DIRECTIONAL OVERCURRENT RELAYS

Then, the optimization problem is approximated as

Find

which minimizes

subject to
and

(9)

with the Lagrangian function given by

(10)
where
indicates the gradient (vector of first order partial
indicates the search direction [15], [22],
derivatives), and
[24].
Iterations are performed till the convergence is obtained. Each
iteration in the SQP algorithm involves three major steps [22],
[24]:
1) solve the QP subproblem, and obtain the search direction;
2) line search and merit function calculation;
3) updating of the Hessian matrix of the Lagrangian function, using BroydonFletcherGoldfrabShanno (BFGS)
formula.
V. APPLICATION OF HYBRID GA-NLP APPROACH TO RELAY
COORDINATION OPTIMIZATION
The problem of optimum coordination of OCRs is formulated
as NLPP. After formulation of the problem, few generations of
GA are used to determine the TMS and PS of OCRs. These
values are then used as initial choice in NLP method, which
gives the global optimum solution.
A. Finding Initial Solution Using GA
For applying GA technique to find the initial solution, the
problem is first converted into unconstrained optimization
problem. As the objective function, coordination constraints
and operating time constraints are written using the relay
characteristics (given in (6)), the relay characteristic constraint
gets automatically incorporated in the objective function, coordination constraints and operating time constraints.
and
, are taken
The constraints due to bounds on
care of by defining the lower and upper limit of the variables
representing TMS and PS of relays, respectively, in the GA
program.
The constraints due to operating time of relays, and the constraint due to coordination criteria, are included in the objective
function using penalty method, as illustrated in Section VI-A
and thus the problem gets converted into an unconstrained optimization problem.
Number of bits to represent each parameter is decided and
then a population of suitable number of chromosomes is generated randomly. The value of variables in each chromosome is
bounded by lower and upper limits, as described above.
After this, the iterations (generations) of GA are started. The
population is passed through the fitness function (objective
function). Then the population is sorted according to fitness. As
the objective function is of minimization type the chromosome

113

giving minimum value is most fit chromosome. This chromosome has been treated as elite chromosome in this paper. The
chromosomes with higher fitness value survive (reproduced in
the next generation) and are called parent chromosomes. These
are used for mating.
Pairs of parent chromosomes are made for mating. Using the
pairs of parent chromosome, crossover is performed. For each
pair the crossover site is selected randomly. One pair (two parent
chromosomes) generates two offsprings after crossover. All the
parent chromosomes and all offsprings are placed together to
form the population for the next generation. The same population size is maintained in all generations.
The mutation is applied after crossover. The number of mutations to be performed is decided by mutation rate, which is one
of the GA parameter to be supplied at the beginning of the program. For each mutation the chromosome is selected randomly
(if selected chromosome is elite then mutation is not performed
on this chromosome and another chromosome is selected). The
bit to be mutated is selected randomly from the chromosome
and is replaced by its complement.
At this point, an iteration of GA is complete. As, in this paper,
the stopping criteria has been considered as number of iteration of GA, the process is stopped after performing pre-specified number of iterations. The elite chromosome, at this stage,
gives the result.
B. Finding Final (Optimum) Solution Using NLP
The function available in MATLAB optimization toolbox, to
solve a constrained nonlinear optimization problem, can be used
to find the global optimum solution of the relay coordination
problem. The results obtained, after performing pre-specified
number of iterations of GA, are taken as the initial values of variables while applying NLP method. The lower and upper bounds
of variables (TMS and PS of OCRs), are decided in the same
way as mentioned in Section II-C, and Section II-D. The objective function and the constraints (relay operating time constraints, and coordination constraints), are defined in two different function files. These functions are called in the main program where the MATLAB optimization toolbox function is used
to give the optimum solution.
VI. IMPLEMENTATION OF PROPOSED METHOD
The proposed method was tested successfully for various systems, out of which two are presented in this paper. In first illustration all 24 relays have been considered as numerical relays
with standard IDMT characteristics. In the second illustration
seven relays with different characteristics have been considered.
Out of these seven OCRs, six have been considered as numerical and one was considered as electromechanical.
A. Illustration I
to
A single end fed, ring main system with nine buses (
), and 24 relays, as shown in Fig. 2, was considered. All of the
relays were considered as digital (numerical), directional OCR
with standard IDMT characteristics, and have their tripping direction away from the bus.
1) System Information: Bus 1 is receiving the power, which
has been represented by a source of 100 MVA, 33 kV with a

114

IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 26, NO. 1, JANUARY 2011

TABLE III
PRIMARY BACKUP RELATIONSHIP OF RELAYS

Fig. 2. Single-line diagram of a nine-bus system.

TABLE II
LOAD CURRENTS

TABLE IV
MAXIMUM LOAD CURRENT, AND MINIMUM AND MAXIMUM FAULT CURRENT

source impedance of
p.u. Base MVA is 100 and base
p.u. The
kV is 33. Each line has an impedance of
load currents are shown in Table II. The OCRs are numbered
as 1 to 24. The CT ratio for each relay is 500:1. Twelve fault
points (one on each line), marked as A to L, were considered.
The primary-backup relationship of relays for these fault points
is shown in Table III.
The maximum load current, and minimum and maximum
fault current through the relays is shown in Table IV.
2) Problem Formulation: The minimum operating time of
each relay was taken as 0.2 s and the CTI was also taken as 0.2 s
for this problem. The current seen by the relays in case of fault at
different points was found by performing short circuit analysis.
Variables
to
were taken to represent TMS of OCRs 1
to
were taken to represent the PS of relays
to 24, and
1 to 24, respectively. The objective function was formed using
(1) and (6) as
(11)
is the current seen by the relay under consideration,
where
for nearest fault (fault in its primary protection zone).
Coordination constraints were formed using (2) and (6), and
relay operating time constraints were formed using (3) and (6).
The constraints due to bounds on TMS and PS of relays were
formed using (4) and (5), respectively.

00 indicates that the load current is not seen by the relay

3) Finding Initial Solution Using GA: The constraints due to
bounds on
, were taken care of by defining the lower and
to
, and the constraints due
upper limit of the variables
, were taken care of by defining the lower and
to bounds on
to
, in the GA program.
upper limit of the variables
The constraints due to operating time of relays, and the constraint due to coordination criteria, were included in the objective function using penalty method. As GA was used to find the
initial solution only, a simple penalty method is proposed in this
paper, in which while finding the value of objective function for
a chromosome a penalty is added to the value of objective function, if any constraint is violated. As the objective function is

BEDEKAR AND BHIDE: OPTIMUM COORDINATION OF DIRECTIONAL OVERCURRENT RELAYS

115

04. Calculate the objective function vale for kth chromosome

using (11)

05.

06. If nth constraint is violated for kth chromosome

07.

08. If
then go to step 09
else go to step 06.
09.

.
(i.e., number of chromosomes)

10. If
then go to step 11
else go to step 03.

11. Return objective function value for each chromosome.

With this, when the population is sorted according to objective function values of the chromosomes in the population, the
chromosomes with higher value of objective function (for which
one or more constraints are violated) go to the bottom and are
automatically discarded in the next generation (iteration).
The GA parameters used were as below
Population size
Number of bits per parameter
Crossover rate
Mutation rate

Fig. 3. Best value of TMS of relays against generations of GA. (a) Relay 1 to
8, (b) relay 9 to 16, and (c) relay 17 to 24.

of minimization type, a large number is taken as penalty. It is

described by the following algorithm
01. Set
02. Set

(a large number).
(start with first chromosome of the population).

03. Take kth chromosome.

GA technique was used to find the initial solution, as described in Section V-A.
4) Finding Final (Optimum) Solution Using NLP: The result
obtained at the end of 40th iteration of GA was taken as the initial value of variables while applying NLP method. The function
available in MATLAB optimization toolbox, for optimization of
constrained NLPP, was used to find the global optimum solution
of the relay coordination problem as described in Section V-B.
5) Results and Discussion: Best values of TMS and PS of
relays plotted against generations of GA (for 40 generations) is
shown in Figs. 3 and 4, respectively.
The values obtained after performing 40 iterations (generations) of GA were taken as initial values of variables and final
values were found using NLP method. The results obtained
using GA and NLP method have been shown in Table V.
The objective function value has been calculated as the sum of
the operating time of each relay for the fault in its primary zone
of protection. The objective function values in Table V show the
and
of relays
value obtained by putting the values of
in (11). As the objective is to minimize the function, the lowest
objective function value is the best.

116

IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 26, NO. 1, JANUARY 2011

TABLE V
TMS AND PS OF RELAYS

TABLE VI
OPERATING TIME OF PRIMARY AND BACKUP RELAYS

Fig. 4. Best value of PS of relays against generations of GA. (a) Relay 1 to 8,

(b) relay 9 to 16, and (c) relay 17 to 24.

With these final values, it can be found that, relay 1 will take
0.3300 s to operate for fault at point A whereas it will operate in
0.4927 s and 0.8322 s for faults at points B and C, respectively.
This is desirable, because for fault at point A, relay 1 is primary
relay and hence it is first to operate, whereas for fault at point
B, relay 3 should get first chance to operate. If it fails to operate then relay 1 should takeover tripping action (relay 1 is first
backup). Similarly for fault at point C relay 5 should get first

chance to operate, if it fails to operate relay 3 should takeover

tripping action and if relay 3 also fails to operate, then only relay

BEDEKAR AND BHIDE: OPTIMUM COORDINATION OF DIRECTIONAL OVERCURRENT RELAYS

117

TABLE VII
RESULTS OBTAINED USING ONLY GA AND ONLY NLP METHOD

Fig. 5. Single-end-fed distribution system.

TABLE VIII
RELAY TYPE, OPERATING TIME AND CT RATIO (ILLUSTRATION II)

1 should take over the tripping action (relay 1 is second backup).

Similar description can be given for operating time of each relay
for different fault points.
Operating time of primary and backup relays for different
fault points is shown in Table VI. It can be seen that the coordination (with minimum CTI of 0.2 s) is maintained in all cases.
The time taken by relay 2 (backup for relay 16) to operate for
fault at point H is 2.165 seconds. It is because small amount of
fault current flows through relay 2 and major amount of fault
current flows through relay 17 (other backup for relay 16). Similar is the description for operating time of backup relays 2 and
11 for fault at point I and K, respectively.
For the purpose of comparison the same problem was solved
using only GA method and only NLP method. In case of GA
method, the problem was solved using different GA parameters
out of which results obtained in one case where 100 iterations
of GA were performed have been presented in Table VII. It is
found that GA does not converge to the global optimum in many
cases.
In case of NLP method the problem was solved with different
initial choices. In some cases it gives the same optimum solution as obtained using hybrid GA-NLP method. But, it was
found that the results obtained in many cases were not the global
optimum. Moreover in some cases the NLP method terminates
without finding even the feasible solution. Results for one such
case, where initial choice of variables was taken as the upper
limit of the corresponding variables, are presented in Table VII.

Thus it can be said that the NLP method highly depends on the
initial choice. If proper initial choice is not made it gives local
optimal solution or terminates with infeasible solution.
B. Illustration II
In this case a single end fed, multi loop distribution system
with 3-buses, and 8 relays, as shown in Fig. 5, was considered.
Bus 1 is receiving the power, which has been represented
by a source of 25 MVA, 11 kV with a source impedance of
. Base MVA is 25 and base kV is 11. Lines between bus-1 and bus-2, and line between bus-2 and bus-3 have
p.u. Line between bus-1 and bus-3
an impedance of
p.u. The relay type, operating
has an impedance of
time, and CT ratio is given in Table VIII. The maximum load

118

IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 26, NO. 1, JANUARY 2011

TABLE IX
MAXIMUM LOAD CURRENT, AND MINIMUM AND MAXIMUM. FAULT CURRENT

TABLE X
TMS AND PS OF RELAYS

TABLE XI
CURRENT SEEN BY THE RELAYS AND THEIR OPERATING TIME

from 0.5 to 2 in discrete steps of 0.25. Four fault points (one on

each line), marked as A to D, were considered.
As relays 2 and 4 are instantaneous OCRs and relay 6 is definite time OCR their operating time for any fault location is fixed
and is as shown in Table VIII. Hence, the problem is to find
the optimum values of TMS and PS of relays 1, 3, 5, 7, and
8. After formulation of the problem the initial solution was obtained using 40 iterations of GA and final solution was found
using NLP method, in the same way as explained for illustration I. The same GA parameters (as in illustration I) were used.
The results obtained are presented in Table X.
The current seen by the relays 1, 3, 5, 7, and 8, and their
operating time for different fault points is presented in Table XI
(operating time of relays 2, 4, and 6 are independent of fault
location, as shown in Table VIII).
Noghabi et al. have presented hybrid GA method to determine optimum TMS and PS of OCRs. They have decomposed
the coordination problem in two subproblems. GA has been
used to solve the first subproblem, i.e., to determine PS of relays. Thus each chromosome in the genetic population was used
to represent PS of relays. With this chromosome information the
OCR coordination problem gets converted to LPP, which was
solved using standard LPP method to determine the TMS of relays [1]. Thus the method presented in [1] is hybrid GA-LPP
method.
To compare the proposed hybrid GA-NLP method with hybrid GA-LPP method, the same problem was also solved using
hybrid GA-LPP method. The results obtained are presented in
Table XII. The GA parameters selected were
Population size
Number of bits per parameter
Crossover rate
Mutation rate

Note I
is in ampere, t is in second.
indicates that the fault is not seen by relay

00

TABLE XII
TMS AND PS OF RELAYS USING TWO HYBRID METHODS

As mentioned in illustration I the objective function value has

been calculated as the sum of the operating time of each relay
for the fault in its primary zone of protection. It was found that,
with the TMS and PS obtained using hybrid GA-LPP method,
the operating time of relays 1, 3, 5, 7, and 8 for different fault
points is same as that mentioned in Table XI, except operating
time of relay 5 for fault at point D (with GA-LPP method it
comes out to be 0.7491 s).
It can be seen from Table XII that the proposed hybrid
GA-NLP method reduces the time required to achieve the
global optimum solution to a large extent, as compared to that
using hybrid GA-LPP method.
VII. CONCLUSION

current, and minimum and maximum fault currents for each

relay have been presented in Table IX.
Relays 1 to 7 have been considered as numerical OCRs, with
different characteristics (as given in Table VIII). Relay 8 has
been considered as electromechanical relay for which TMS
range is from 0 to 1 in discrete steps of 0.01, and PS range is

A hybrid GA-NLP method for determination of optimum

values of TMS and PS of OCRs is presented in this paper.
A systematic procedure for formulation of the problem as an
optimization problem has been developed. The problem is
formulated as a constrained NLPP in this paper. GA has been
applied to find the initial solution of this problem, which was
then used by NLP method to find the global optimum solution.
Thus the proposed method makes use of the advantages of
both GA and NLP method, and at the same time overcomes

BEDEKAR AND BHIDE: OPTIMUM COORDINATION OF DIRECTIONAL OVERCURRENT RELAYS

the drawbacks of these methods. It has been shown that the

proposed method outperforms the other methods.
The algorithm was tested for various system configurations,
including multi loop systems, and was found to give satisfactory results in all the cases. It has been shown that the proposed method can be easily applied to a system with combination of numerical and electromechanical relays and combination
of OCRs with different characteristics. Same procedure can be
adopted for optimum setting of earth fault relays.
REFERENCES
[1] A. S. Noghabi, J. Sadeh, and H. R. Mashhadi, Considering different
network topologies in optimal overcurrent relay coordination using a
hybrid GA, IEEE Trans. Power Del., vol. 24, no. 4, pp. 18571863,
Oct. 2009.
[2] A. J. Urdaneta, N. Ramon, and L. G. P. Jimenez, Optimal coordination of directional relays in interconnected power system, IEEE Trans.
Power Del., vol. 3, no. 3, pp. 903911, Jul. 1988.
[3] H. Zeienldin, E. F. El-Saadany, and M. A. Salama, A novel problem
formulation for directional overcurrent relay coordination, in Proc.
Large Engineering Systems Conf. Power Engineering, Halifax, NS,
Canada, 2004, pp. 4852.
[4] B. Chattopadhyay, M. S. Sachdev, and T. S. Sidhu, An online relay
coordination algorithm for adaptive protection using linear programming technique, IEEE Trans. Power Del., vol. 11, no. 1, pp. 165173,
Jan. 1996.
[5] A. J. Urdaneta, H. Restrepo, S. Marquez, and J. Sanchez, Coordination of directional relay timing using linear programming, IEEE Trans.
Power Del, vol. 11, no. 1, pp. 122129, Jan. 1996.
[6] H. K. Karegar, H. A. Abyaneh, V. Ohis, and M. Meshkin, Pre-processing of the optimal coordination of overcurrent relays, Elect.Power
Syst. Res., vol. 75, pp. 134141, 2005.
[7] H. A. Abhyaneh, M. Al-Dabbagh, H. K. Karegar, S. H. H. Sadeghi,
and R. A. J. Khan, A new optimal approach for coordination of directional overcurrent relays in interconnected power system, IEEE Trans.
Power Del., vol. 18, no. 2, pp. 430435, Apr. 2003.
[8] S. K. Bhattacharya and S. K. Goswami, Distribution network reconfiguration considering protection coordination constraints, Elect. Power
Compon. Syst., vol. 36, pp. 11501165, 2008.
[9] C. W. So and K. K. Li, Overcurrent relay coordination by evolutionary
programming, Elect. Power Syst. Res., vol. 53, pp. 8390, 2000.
[10] F. Razavi, H. A. Abyaneh, M. Al-Dabbagh, R. Mohammadi, and H.
Torkaman, A new comprehensive genetic algorithm method for optimal overcurrent relays coordination, Elect. Power Syst. Res., vol. 78,
pp. 71372, 2008.
[11] C. W. So, K. K. Li, K. T. Lai, and K. Y. Fung, Application of genetic
algorithm to overcurrent relay grading coordination, in Proc. 4th Int.
Conf. Advances in Power System Control, Operation and Management,
Hong Kong, China, 1997, pp. 283287.

119

[12] A. Y. Abdelaziz, H. E. A. Talaat, A. I. Nosseir, and A. A. Hajjar, An

adaptive protection scheme for optimal coordination of overcurrent relays, Elect. Power Syst. Res., vol. 61, pp. 19, 2002.
[13] T. Keil and J. Jager, advanced coordination method for overcurrent
protection relays using nonstandard tripping characteristics, IEEE
Trans. Power Del., vol. 23, no. 1, pp. 5257, Jan. 2008.
[14] D. Birla, R. P. Maheshwari, and H. O. Gupta, An approach to tackle
the threat of sympathy trips in directional overcurrent relay coordination, IEEE Trans. Power Del., vol. 22, no. 1, pp. 851858, Jan. 2007.
[15] D. Birla, R. P. Maheshwari, and H. O. Gupta, A new nonlinear directional overcurrent relay coordination technique, and banes and boons
of near-end faults based approach, IEEE Trans. Power Del., vol. 21,
no. 3, pp. 11761182, Jul. 2006.
[16] D. Birla, R. P. Maheshwari, H. O. Gupta, K. Deep, and M. Thakur,
Random search technique in directional overcurrent relay coordination, Int. J. Emerging Elect. Power Syst., vol. 7, no. 1, 2006, Article 1.
[17] D. Birla, R. P. Maheshwari, and H. O. Gupta, Time overcurrent relay
coordination: A review, Int. J. Emerging Elect. Power Syst., vol. 2, no.
2, 2005, Article 1039.
[18] S. A. Soman, Lectures on Power System Protection module 4 & 5.
[Online]. Available: www.cdeep.iitb.ac.in/NPTEL
[19] Y. G. Paithankar and S. R. Bhide, Fundamentals of Power System Protection. New Delhi, India: Prentice-Hall of India Pvt. Ltd., 2003.
[20] D. E. Goldberg, Genetic Algorithms in Search, Optimization, and Machine Learning. New Delhi, India: Dorling Kindersley (India) Pvt.
Ltd., 2008.
[21] R. L. Haupt and S. E. Haupt, Practical Genetic Algorithms, 2nd ed.
Hoboken, NJ: Wiley, 2004.
[22] S. S. Rao, Engineering OptimizationTheory and Practice, 3rd ed.
New Delhi, India: New Age International Pvt. Ltd., 1998.
[23] K. Deb, Optimization for Engineering DesignAlgorithms and Examples. New Delhi, India: Prentice Hall of India Pvt. Ltd., 2006.
[24] MATLAB Optimization Toolbox. Natick, MA.

Prashant Prabhakar Bedekar is currently pursuing the Ph.D. degree in electrical engineering at Visvesvaraya National Institute of Technology, Nagpur,
India.
Currently, he is an Assistant Professor in the Department of Electrical Engineering at Government College of Engineering, Amravati, India. His research
interests include power system protection, optimization, and AI techniques.

Sudhir Ramkrishna Bhide received the B.E., M.Tech., and Ph.D. degrees from
Nagpur University, Nagpur, India.
Currently, he is an Associate Professor in the Department of Electrical Engineering, Visvesvaraya National Institute of Technology, Nagpur. He has authored a book Fundamentals of Power System Protection (Prentice Hall of India,
2003). His special areas of interest include the application of AI techniques to
power system protection.