2009 Third International Conference on Power Systems, Kharagpur, INDIA December 27-29

ICPS – 247

Optimum coordination of overcurrent relays in

distribution system using genetic algorithm
Prashant P. Bedekar Sudhir R. Bhide Vijay S. Kale
Research Scholar Assistant Professor Sr. Lecturer
Electrical Engineering Department, Visvesvaraya National Institute of Technology, NAGPUR (INDIA) 440 010

Abstract— This paper presents genetic algorithm (GA) method
for coordination of overcurrent (OC) relays. The OC relays are
the major protection devices in a distribution system. OC relay is
usually employed as backup protection. But in some situations it
may be the only protection provided. To reduce the power
outages, mal-operation of the backup relays should be avoided,
and therefore, OC relay coordination in power distribution
network is a major concern of protection engineer.
The OC relay coordination in ring fed distribution networks
is a highly constrained optimization problem. The purpose is to
find an optimum relay setting to minimize the time of operation
of relays and at the same time, to avoid the mal-operation of
relays. This paper presents GA technique for optimum
coordination of OC relays in a ring fed distribution system.
Constraints are incorporated in the fitness function making use
of the penalty method. Computer programs (using MATLAB)
have been developed for optimum coordination of OC relays
using GA technique.
Keywords- Overcurrent relay coordination, Constrained
optimization, Genetic algorithms.
I. INTRODUCTION
The most obvious effect of a shunt fault is a sudden built
up of current. So it is natural that the magnitude of current be
utilized as positive indication of existence of a fault. Therefore
the over-current protection is the most widely used form of
protection [1-3]. Overcurrent (OC) relay is usually employed
as backup protection. But in some situations it may be the only
protection provided.
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. A typical power system may consist of hundreds of
equipment and even more protection relays to protect the
system. Each relay in the system needs to be coordinated with
the relay protecting the adjacent equipment. If backup
protections are not well coordinated, mal-operation can occur
and, therefore, OC relay coordination is a major concern of
power system protection [4,5]. 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.
In a system where there is a source at more than one of
the line terminals, fault and load current can flow in either
direction. Relays protecting the lines are, therefore, subject to
fault currents flowing in both the directions. If non directional
OC relays were used in such system, they would have to be
coordinated with, not only the relays at the remote end of the
line, but also with relays behind them. Since directional relays
operate only when the fault current flows in the specified
tripping direction, they avoid coordination with the relays
behind them.
The directional OC relay coordination problem in
distribution system can be defined as constrained optimization
problem and can be solved using GA technique.
Genetic Algorithms (GAs) are search methods that
emulate the evolution of a population. GAs perform a global
search on the solution space of a given problem domain [6-8].
Hence GA gives the global optimum solution. The GA
optimization method has been employed for optimum
coordination of OC relays in this paper.
Thus the main contribution of this paper includes (i)
conversion of a constrained optimization problem into
unconstrained optimization problem, by incorporating the
constraints in the fitness function using penalty method, and
(ii) obtaining the optimum coordination of OC relays using
GA technique.
II. COORDINATION OF OC RELAYS IN RING FED SYSTEM
As soon as the fault takes place it is sensed by both
primary and backup protection. The primary protection is the
first to operate as its operating time being less that that of the
backup relay.
A simple radial feeder with two sections is shown in
figure 1. For fault at point F, relay RB is first to operate. Let
the operating time of RB is set to 0.1 s. The relay RA should
RA RB F
B
A
Fig 1 A radial feeder (both relays are non-directional)
wait for 0.1 s plus, a time equal to the operating time of circuit
breaker (CB) at bus B, plus the overshoot time of relay A [1].

978-1-4244-4331-4/09/$25.00 ©2009 IEEE

2009 Third International Conference on Power Systems, Kharagpur, INDIA December 27-29
ICPS – 247

This is necessary for maintaining the selectivity of relays at A where ti ,i indicates the operating time of the primary
and B.
A ring main feeder system is shown in figure 2. It allows relay at i, for near end fault. This objective is to be achieved
supply to be maintained to all the loads in spite of fault on any under the following constraints [9-12]:
section. Relays 1, and 8 are non directional whereas all other A. Coordination criteria –
relays (2, 3, 4, 5, 6, and 7) are directional OC relays. All
directional relays have their tripping direction away from the tbi ,i − ti ,i ≥ Δt (2)
concerned bus. where ti ,i is the operating time of the primary relay at i, for
near end fault
tbi ,i is the operating time for the backup relay, for the
8 1 same near end fault
Δt is the coordination time interval (CTI)
7 2
B. Bounds on the relay operating time –
ti ,i ,min ≤ ti ,i ≤ ti ,i ,max (3)
where ti ,i ,min is the minimum operating time of relay at i for
6 3 near end fault (fault at i)
ti ,i ,max is the maximum operating time of relay at i
5 4 for near end fault (fault at i)

Fig 2 A ring main feeder C. Relay characteristics –

For coordination purpose relays 2,4,6, and 8 will form one All relays are assumed to be identical and are assumed to
group and relays 1,3,5, and 7 will form other group. For have normal IDMT characteristic as [4-5,9-12] :
group one, setting is to be started from relay 2. The relay
λ (TMS )
operating times will be related as top = (4)
TR8 > TR6 > TR4 > TR2 ( PSM )γ − 1
For group two, setting is to started from relay 7. The relay where top is relay operating time,
operating times will be related as
TR1 > TR3 > TR5 > TR7 TMS is time multiplier setting, and
The actual operating time for each relay can be decided again PSM is plug setting multiplier.
considering the operating time of preceding relay, operating For normal IDMT relay γ is 0.02 and λ is 0.14. As the pickup
time CB associated with preceding relay, and the overshoot currents of the relays are pre determined from the system
time of the relay under consideration. requirements, equation (4) becomes
As the size and complexity of the system goes on
increasing it becomes more and more difficult to coordinate
top = a (TMS ) (5)
the relays. Keeping the same concept of coordination (as λ
explained above) in view, the problem can be stated as where a=
constrained optimization problem which can be solved by GA ( PSM )γ − 1
technique. Making substitution from equation (5) in equation (1), the
objective function becomes
III. PROBLEM FORMULATION m

The coordination problem of directional OC relays in min z = ∑ a i (TMS ) i (6)

interconnected power systems, can be stated as an i =1

optimization problem, where the sum of the operating times of Thus the relay characteristic constraint is incorporated in
the relays of the system, for near end fault, is to be minimized the objective function itself. The values of ai for i th relay for
[9],
m
different fault locations are predetermined. Value of TMS for
i.e., min z = ∑ ti ,i (1) each relay is to be determined using GA.
i =1
IV. GENETIC ALGORITHM

GA is an optimization technique based on the principles

of genetics and natural selection. A GA allows a population

978-1-4244-4331-4/09/$25.00 ©2009 IEEE

2009 Third International Conference on Power Systems, Kharagpur, INDIA December 27-29
ICPS – 247
composed of many individuals to evolve under specified
selection rules to a state that maximizes the “fitness”.
The classical optimization methods have limitations in
searching for global optimum point and sometimes trapped in
local optimum point. Due to the fact that GA is a multipoint
search method rather that the conventional single point search
methods, GA promises the global optimum point to be reached
[6-8]. Philosophically, GA is based on Darwin’s theory of
“survival of the fittest”.
GA begins, like any other optimization algorithm, by
defining the optimization variables, and the cost function
(objective function). It ends like other optimization algorithms
too, by testing for convergence. In between, however, this
algorithm is quite different [6,13]. The algorithm is as given
below –
01. Start.
02. Define fitness function.
03. Enter number of variables in the function.
04. Set stopping criteria.
05. Define GA parameters.
06. Create initial population.
07. Set iteration count = 1.
08. Evaluate the fitness of each chromosome in the population.
09. Sort the fitness and associated parameters.
10. If stopping criteria is satisfied
then go to step 16.
else go to step 11.
11. Select pairs for mating (i.e. perform reproduction).
12. Perform crossover.
13. Mutate the population.
14. Increment iteration count.
15. Go to step 08.
16. Display results.
17. Stop.
Representation of GA parameters and the GA operators
(reproduction, crossover and mutation) are explained in brief,
below.
A. Representation of design variables (parameters)
In GAs, the design variables are represented as string of
binary numbers, 0 and 1. If each design variable xi ,
i = 1,2,3,...., n , is coded in a string of length q , a design
vector is represented using a string of total length nq . This
string (string of total length nq ) is called a “chromosome”.
GA starts with a group of chromosome known as “population”.
A continuous design variable x can only be represented by a
set of discrete values if binary representation is used. If a
variable x , whose bounds are given by x min and x max is
represented by a string of q binary digits, its decimal value
can be computed using –
978-1-4244-4331-4/09/$25.00 ©2009 IEEE
b * ( xmax − xmin )
x = xmin + (7)
2q − 1
where b is the decimal (discrete value) equivalent of binary
number. Thus if a continuous variable is to be represented with
high accuracy, we need to use a large value of q in its binary
representation [14].
B. Genetic Operators
The basic operations of natural genetics – reproduction,
crossover, and mutation, are implemented during numerical
optimization.
“Reproduction” is a process in which the individuals are
selected based on their fitness values relative to that of the
population. Thus individuals (chromosomes) with higher
fitness values have a greater chance of being selected for
mating and subsequent genetic action. Consequently, highly fit
individuals live and reproduce, and less fit chromosomes die.
After reproduction, the “crossover” operation is implemented.
Crossover is an operator that forms a new chromosome, called
“offspring”, from two “parent” chromosomes by combining
part of the information from each. Crossover is implemented in
two steps. First, two individuals 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. The offspring obtained from
crossover are placed in 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
random location
V. APPLICATION OF GA TO OPTIMUM COORDINATION OF OC
RELAYS
As coordination of OC relays is basically an optimization
problem, GA can be applied to find the optimum solution to
this problem. The aim is to find out the optimum value of
TMS for all relays, hence the TMS of relays are taken as
variables. Out of the three sets of constraints (described by
equations 2, 3 and 4), the relay characteristic constraint is
already incorporated in the objective function (as shown in
equation 5 and 6). The constraint putting bounds on TMS is
taken care of by specifying minimum and maximum bounds as
the lower and upper limit of variable ( TMS ) values while
mapping, i.e. while converting the binary representation of
variables in to continuous decimal values. The constraints
imposed by coordination criteria are included in the objective
function using penalty method [7,8]. Thus the constrained
optimization problem is converted into unconstrained
optimization problem. After this the problem is solved using
the algorithm given in section IV.
2009 Third International Conference on Power Systems, Kharagpur, INDIA December 27-29
ICPS – 247

VI. RESULTS Application of GA –

A program is developed in MATLAB, for optimum
coordination of OC relays. The program is tested for various For applying GA technique to this problem, it is first
systems and is found to give satisfactory results in all the converted in to unconstrained optimization problem. The relay
cases. Two cases (one radial and one multi loop system) are characteristic constraint is already incorporated in the
presented here for demonstration. Detail calculations are given objective function.
in illustration I and similar calculations are performed in The constraints due to bounds on relay operating (and
illustration II for formation of objective function and bounds on TMS ); given by equation 10 and 11; are taken
constraints. Also how GA is applied for this problem is care of by defining the lower limit of the variables in the GA
explained in illustration I and similar method is adopted in program. The lower limit of x1 and is taken as 0.076 whereas
illustration II.
In each case, the GA parameters used are – lower limit of x 2 is taken as 0.1. For considering the upper
Population size = 256 bound the upper limit of variable is also set in the program.
Number of bits / parameter =8 The upper limit of all variables is taken as 1.2.
Crossover rate = 0.5 (50 %) The constraints due to coordination criteria; given by
Mutation rate = 0.15 (15 %) equation (9) is included in the objective function using penalty
method and thus the problem is converted in to unconstrained
A Illustration I – optimization problem [16].
To test the algorithm, initially a simple radial system, A population of 256 chromosomes is generated randomly.
shown in figure 1, is considered. The maximum fault current The value of variables in each chromosome is bounded by
just beyond bus A and bus B are 4000 A and 3000 A lower and upper limits. Each chromosome consists of 16 bits
respectively, the plug setting of both the relays are 1, the CT (8 bits/ parameter), selected randomly. The population is
ratio for RA is 300:1 and for RB is 100:1. Minimum operating passed through the fitness function (objective function). As the
time for each relay is considered as 0.2 s and the CTI is taken objective function is of minimization type the chromosome
as 0.57 s. Calculation of value of ai (mentioned in equation 6) giving minimum value is most fit chromosome. The
population is sorted according to fitness. The chromosomes
for relays is shown in Table I.
with higher fitness value (upper 50 % population) survive
TABLE I Calculation of ai constant for relays (reproduced in the next generation) and are called parent
chromosomes. These are used for crossover.
SN Fault Relay
Pairs of parent chromosome are made for mating. A
position RA RB simple method of pairing from top-to-bottom is used. Single
1. Just 0.14 point crossover is performed. The crossover site is selected
beyond --- randomly swapping of bits, among the parent chromosome, is
bus A (13.33) 0.02 − 1
done. Crossover generates offsprings. The crossover rate is
2. Just 0.14 0.14 kept as 50 %. The parent chromosomes along with offsprings
beyond are placed together to form the population for the next
bus B (10) 0.02 − 1 (30) 0.02 − 1
generation. The population size of 256 is maintained in all
generations.
Considering x1 and x 2 as TMS of relay RA and RB The “mutation” is applied after crossover. For mutation
respectively, the problem can be stated as – the bit is selected randomly from chromosome (the
min z = 2.63x1 + 2 x2 (8) chromosomes are also selected randomly) and is replaced by
its complement. The mutation rate is taken as 10 %. After
subject to 2.97 x1 − 2 x2 ≥ 0.57 (9) mutation, the next iteration is started (from passing the
2.63x1 ≥ 0.2 (10) population through fitness function).
The optimum solution to this problem is obtained using
and 2 x2 ≥ 0.2 (11) GA technique and is given in Table II.
The minimum operating time constraints (equation 10 and
11) give the lower limit of TMS of relays as x1 ≥ 0.076 TABLE II TMS obtained using GA technique

and x 2 ≥ 0.1 . This automatically satisfies the lower bound on Relay TMS Value obtained using GA
the value of TMS ( xi ≥ 0.025 ). The upper limit of TMS of RA x1 0.2594
for both relays is taken as 1.2 [15]. RB x2 0.1

The results for the same problem are obtained using

revised simplex method and dual simplex method also and are

978-1-4244-4331-4/09/$25.00 ©2009 IEEE

2009 Third International Conference on Power Systems, Kharagpur, INDIA December 27-29
ICPS – 247

found to be same. Thus the optimality of the result is TABLE IV Total fault current and primary-backup relationship of relays
confirmed. Fault Relay current
B Illustration II – point 1 2 3 4 5 6 7 8
A 10 10 3.3 -- -- -- -- 3.3
In this case a single end fed, multi loop distribution B 3.45 3.45 5.1 6.9 -- -- -- --
system, with eight OC relays as shown in figure 3 is C 2 2 10 4 -- -- 4 --
considered. The system is multi loop system because D 5 5 4 10 -- -- -- 4
depending on the fault point, different configurations E 20 6 2 -- 8 -- -- 2
(depending on the direction of current in various feeders) are
F 6 20 2 -- -- 8 -- 2
formed. Six different fault points are taken for illustration. The
-- indicates that the fault is not seen by the relay.
total fault current and primary-backup relationship of relay for
the six fault points are given in Table III. The optimization problem is formed in the same way as
explained in illustration I. In this case there are eight variables
( TMS of eight relays), eight constraints due to bounds on
relay operating time and nine constraints due to coordination
criteria. The optimum values of TMS obtained using GA
1 2 3 technique are as under (the subscripts indicate the relay
number) –
E F TMS1 = 0.2975 TMS 2 = 0.2975
TMS3 = 0.2270 TMS 4 = 0.1739
5 6 C TMS5 = 0.0607 TMS6 = 0.0607
TMS7 = 0.0402 TMS8 = 0.1129
A 4
The best values of TMS of relays obtained during
different generations of GA are shown in figure 4.
D

8 7

B
Fig 3 Single end fed, multi loop distribution system

All relays are assumed to have plug setting of 1 and CT

ratio of 100 : 1. The relay current for different fault points are
given in Table IV.
TABLE III Total fault current and primary-backup relationship of relays

Fault Total fault Primary Backup

point current (A) relay
A 2330
B 1200
C 1400
D 1400
E 2800
F 2800
relay
1 --
2 --
8 3
3 --
4 1, 2
3 --
7 4
4 1, 2
8 3
1 --
5 8
2 --
6 8
Fig 4 Best values of TMS against generations of GA
VII. CONCLUSION
GA method for optimum coordination of overcurrent
relays in distribution system is presented in this paper. The
optimum relay coordination problem, which is basically a
highly constrained optimization problem, is converted into
unconstrained optimization problem by defining a new
objective function (using penalty method) and by using the
bounds on TMS (because of bounds on relay operating time)
as the limits of variables while decoding (mapping) the binary
value of variables into decimal value. In both the illustrations

978-1-4244-4331-4/09/$25.00 ©2009 IEEE

2009 Third International Conference on Power Systems, Kharagpur, INDIA December 27-29
ICPS – 247
the minimum operating time for each relay is considered as
0.2 s. The CTI for illustration I is taken as 0.57 s, whereas for
illustration II it is taken as 0.6 s. It shows that depending upon
the relay and breaker specifications and system requirements
the constraints can be formed and the optimum coordination
can be obtained.
The algorithm is tested for various systems and is found
to give satisfactory results in all the cases.
REFERENCES
[1] Paithankar Y. G., and Bhide S. R., “Fundamentals of Power System
Protection,” Prentice Hall of India Private Limited, New Delhi, 2007.
[2] Badri Ram, and Vishwakarma D. N., “Power System Protection and
Switchgear,” Tata McGraw Hill Publishing Company Limited, New
Delhi, 2008.
[3] Mason C.R., “The Art and Science of Protective Relaying”, John Wiley
and Sons, New York.
[4] So C.W., and Lee K.K., “Overcurrent Relay Coordination by
Evolutionary Programming,” Electric Power System Research, vol. 53,
pp. 83-90, 2000.
[5] So C.W., Lee K.K., Lai K.T., and Fung Y.K., “Application of Genetic
Algorithm to Overcurrent Relay Grading Coordination,” Fourth
International Conference on Advances in Power System Control,
Operation and Maintenance, APSCOM-97, HongKong, pp. 283-287,
November 1997.W. D. Stevenson, “Elements of power system analysis”,
McGraw Hill Publications, Kogakusha, 1975.
[6] R. L. Haupt, and S. E. Haupt, “Practical genetic algorithms, second
edition,” John Wiley and sons, Inc., Publication Hoboken, New Jersey,
2004.
978-1-4244-4331-4/09/$25.00 ©2009 IEEE
[7] D. E. Goldberg, “Genetic algorithms in search, optimization, and
machine learning,” Dorling Kindersley (India) Pvt. Ltd., New Delhi,
2008.
[8] S. S. Rao, “Engineering optimization – theory and practice, third
edition,” New Age International (P) Limited, Publisher New Delhi,
1998.
[9] Chattopadhyay B., Sachdev M.S., and Sidhu T.S., “An Online Relay
Coordination algorithm for Adaptive Protection using Linear
Programming Technique”, IEEE Trans. on Power Delivery, Vol 11, pp.
165-173, Jan 1996.
[10] Urdaneta A.J., Nadira Ramon, and Luis G.P. Jimenez, “Optimal
Coordination of Directional Relays in Interconnected Power System”,
IEEE Trans. on Power Delivery, Vol 3, No. 3, pp. 903-911, July 1988.
[11] Urdaneta A.J., Restrepo H., Marquez S., and Sanchez J., “Coordination
of Directional Relay Timing using Linear Programming”, IEEE Trans.
on Power Delivery, Vol 11, pp. 122-129, Jan 1996.
[12] Zeienldin H., El-Saadany and Salama M.A., “A Novel Problem
Formulation for Directional Overcurrent Relay Coordination,” IEEE
International Conference, pp48-52, 2004.
[13] P. P. Bedekar, and S. R. Bhide, “Optimization of multivariable nonlinear
functions using genetic algorithms”, IEEE International Advance
Computing Conferenc (IACC), Thapar University, Patiala, India, 6-7
March 2009, pp. 1905-1909.
[14] T. J. Ross, “Fuzzy Logic with engineering applications- second edition,”
Wiley India Pvt. Ltd., New Delhi, 2008.
[15] Soman S.A., Lectures on Power System Protection
(www.cdeep.iitb.ac.in/ NPTEL online), module 5, lecture 19.
[16] K. Deb, “Optimization for Engineering Design – Algorithms and
Examples”, Prentice Hall of India Pvt. Ltd., New Delhi, 2006.