You are on page 1of 17

Received: 29 July 2016 Revised: 1 June 2017 Accepted: 5 June 2017

DOI: 10.1002/etep.2395

RESEARCH ARTICLE

Optimized settings of directional overcurrent relays in meshed


power networks using stochastic fractal search algorithm

Attia A. El‐Fergany1 | Hany M. Hasanien2

1
Electrical Power and Machines Department,
Summary
Faculty of Engineering, Zagazig University,
Zagazig, 44519, Egypt This paper presents a novel application of the stochastic fractal search algorithm to
2
Electrical Power and Machines Department, solve the optimal relay coordination problem of meshed power networks. The opti-
Faculty of Engineering, Ain Shams mization problem has 3 design variables, which are the time dial, the pickup current,
University, Cairo, 11517, Egypt
and the tripping characteristic of each relay and subjects to set of coordination con-
Correspondence straints. The objective function is adapted to minimize the total operating time of the
Attia A. El‐Fergany, Electrical Power and
primary and backup relays while maintaining the validity of coordination within
Machines Department, Faculty of
Engineering, Zagazig University, Zagazig acceptable limits. Three interconnected power systems with various topologies
44519, Zagazig, Egypt. and fault scenarios are demonstrated. Among these 3 test cases, it is worthy empha-
Email: el_fergany@ieee.org
sizing that the third power system is highly penetrated with 6 distributed generators,
which tangle the optimization problems with numerous relay pair scenarios. The
validity and the effectiveness of the proposed method are confirmed using the sim-
ulation results, in addition to necessary subsequent comparisons to other competing
approaches. The simulated numerical results and the performance measures indicate
that the stochastic fractal search algorithm is viable and able to generate competitive
optimal settings for overcurrent relays and lie in an acceptable elapsed time.

K EY WO R D S
directional overcurrent relays, meshed power systems, optimization methods, relay coordination

1 | INTRODUCTION

The reliable coordination study is essential to limit the network outage to the smallest area possible by realizing appropriate
setting of relays which facilitates a rapid restoration. In other words, the primary protection schemes should perform fast and
efficiently. However, when it fails, the backup relay should take the action after predefined coordination time margin
(CTM).1 Coordination or alternatively named selectivity is classified into (i) absolute selective and (ii) relative selective.2
Examples of absolute selective relays are differential and restricted earth fault relays. Examples of relative selective are
overcurrent and earth fault relays. In common practices, the coordination studies care with relative selectivity. The ultimate
purpose of relay coordination study is to achieve the targets by defining carefully the time dial (TD), pickup current (IP), and
tripping curve of each overcurrent relay. The coordination of directional protective devices is based on both time, current
magnitude and direction.
The high penetration of distributed generation (DG) in modern electric power systems represents a greater challenge to the
protection schemes. These issues should be solved by advanced protection technologies. The coordination of directional
overcurrent relays in loop interconnected power systems with multiple DG sources is an optimization problem with high non-
linearity. Different optimization methods have been used in literatures to solve this high constraints nonlinear optimization

Int Trans Electr Energ Syst. 2017;e2395. wileyonlinelibrary.com/journal/etep Copyright © 2017 John Wiley & Sons, Ltd. 1 of 17
https://doi.org/10.1002/etep.2395
2 of 17 EL‐FERGANY AND HASANIEN

problem. Generally, the proposed methodologies for solving the relay coordination plans can roughly be classified into 2
categories: (i) traditional methods, including graph‐theoretical, and (ii) artificial intelligence‐based approaches.
During the last decades, various optimization methods, including traditional mathematical and heuristic‐based approaches,
have been applied by researchers to provide an optimum solution for the relay settings and coordination. In the study of
Ezzeddine et al3 and Mahari and Seyedi,4 analytical approaches have been introduced to solve relay coordination problems.
Lagrange generalized with the Karush‐Kuhn‐Tucker conditions for overcurrent protection relays using nonstandard tripping
characteristics were reported.5 Graphical‐theoretically based Z‐matrix to network topology analysis to address the relay coor-
dination plans was presented.6 Examples of the traditional mathematical optimization methods are linear programming7-9 and
nonlinear programing.10 In fact, the graphical and analytical‐based methodologies provided the settings of relays. However,
not necessarily the best optimal or near‐optimal ones are realized or even being trapped in local minima.
Promising alternative solutions to tackle the difficulties and drawbacks of the conventional methods are the applications of
heuristic‐based optimization procedures, with the understanding that the relay coordination plans are formulated as an optimi-
zation problem with various complexity and nonlinearity, which can be solved using the application of heuristic‐based algo-
rithms. Among these heuristic‐based computing algorithms are genetic algorithm (GA),1,10,11 ant colony algorithm,12 various
variants of differential evolution (DE) algorithm,13-15 firefly algorithm,16 teaching learning optimization algorithm,17 artificial
bees colony,18 flower pollination algorithm (FPA),19 symbiotic organism search optimization,20 and many more,21-26 which
have been attempted to solve the relay coordination problems under various network topologies. In the study of Alam
et al,27 5 meta‐heuristic optimization methods are considered to solve the optimal relay coordination with comprehensive
comparisons of their performances. In which the meta‐heuristic optimization methods are GA, particle swarm optimization,
DE, harmony search optimizer, and seeker algorithm are analyzed.
From the aforesaid literature survey, it is clearly documented that the performance of coordination plans essentially
depends on the proper settings of overcurrent relays under different fault scenarios which still require to pay more
attentions. In contrary with the theory and principal methods based on optimization,28 the authors propose to use a recently
heuristic‐based developed algorithm, namely, stochastic fractal search algorithm (SFSA) to address the optimal relay
coordination.
The SFSA29 is a dynamic algorithm inspired by evolutionary concepts, which contain the number of search agents
uses 2 main processes: (i) the diffusing process and (ii) the updating process. Each particle of the SFSA diffuses near
its current position in the diffusing process. However, in the updating process, the SFSA procedure mimics how search
agents in the group update their position based on the position of other search agents in the group (somehow, similar to
the particle swarm optimization). The SFSA is used successfully to solve a number of engineering optimization problems
in various areas, including optimal design of planar steel frames,30 and monitoring of an aerospace structure,31 aiming at
the production of optimal solutions.
The current paper cares with relative coordination of overcurrent relays. The coordination study is formulated as nonlinear
mixed‐integer optimization problem subjected to the coordination constraints. The SFSA is applied to minimize the total oper-
ating time (TOT) of the relays in order to generate the optimal setting of network relays under study and its performance is
investigated and evaluated. Three test cases with various topologies are examined and necessary comparisons to demonstrate
the accuracy and the validity of the proposed method over other competing techniques are made.

2 | PROBLEM STATEMENT AND FORMULATION

In digital overcurrent relays, TD and I P setting values are continuous variables. However, for each relay, there is a multiplicity
selection of tripping curve characteristics. The IEC 60255 standard (Part 3)32 describes the operating time of the overcurrent
relays mathematically by Equation 1:

0 1
B β C
t ¼ @  α ATD (1)
IF
IP −1

where t is the relay operating time (s), TD is the time dial (s), IF is the fault current (A), IP is the pickup current (A), and α and β
are constants. Table 1 gives the values of α and β for various common overcurrent relay with inverse definite minimum time
(IDMT) curves.32,33
EL‐FERGANY AND HASANIEN 3 of 17

TABLE 1 Constants for various standard overcurrent relay characteristics

Curve type α β

Normal inverse (NI) 0.02 0.14


Very inverse (VI) 1.0 13.5
Extremely inverse (EI) 2.0 80
Long‐time inverse (LI) 1.0 120

Because of the aforementioned, the proposed objective function (OF) is dealt as a mixed‐integer nonlinear optimization
problem. The adapted OF is expressed to realize the relay coordination and minimize the TOT of the main and backup relays
along the network comprising all relay pairs as specified in Equation (2):
 
 
Minimize ∑ t m;k þ t b;k ∀k∈R (2)
k

where R is the number of relay pairs, and t m;k and t b;k are the operating times that are calculated using Equation (1) for main and
backup relay pairs, respectively. The OF subjects to the following inequality constraints:
 
t b;k ðback−upÞI ≥t m;k ðmainÞI þ CTMk ∀k∈R (3)
f max;ðk;2Þ f max;ðk;1Þ

CTMk; min ≤CTMk ≤CTMk; max ∀k∈R (4)

TD min;i ≤TDi ≤TD max;i ∀i∈N r (5)

I P; min;i ≤I P;i ≤I P; max;i ∀i∈N r (6)

t m;i ≥T min and t b;i ≤T max ∀i∈N r (7)

where Nr is the total number of network relays involved in the coordination study; t m;k and t b;k represent the operating time of the
primary and backup relay, respectively, for the kth relay pair; If max, (k, 1) and If max, (k, 2) are the maximum 3‐phase fault currents
of the relay pair through primary and backup relays, respectively; CTMk is the time margin for the kth relay pair; CTMmin, k
and CTMmax, k are the minimum and maximum bounds of the time margins, respectively, for the kth relay pair; TDi
is the time‐dial setting of relay ith relay; TDmin, i and TDmax, i are the lower and upper limits of time dials of the ith
relay; respectively; IP, i is the pickup current of the ith relay, IP, min, i and IP, max, i are the lower and upper pickup bounds for
the ith relay, respectively; and T min and T max are the minimum and maximum allowable operating times of the relays along
the network, respectively.
It is worth stating that the merits of using CTM between minimum and maximum values are as follows: (i) the SFSA is
encouraged to maintain the calculated CTM within this specified range (which is typically ranges from 0.2 to 0.5 s), and (ii)
if the user wishes to use it as equality constraint, it is simple by setting the values of CTMk , min = CTMk , max or even set
CTMk , max to infinity. At last, the SFSA is trying to minimize the TOT of all primary and backup relays along the network under
study. The appropriate settings of overcurrent relays require the values of the IP, TD, and the type of inverse characteristic to
achieve the optimal relay coordination. It is worthy to highlight that the inequality constraints specified in Equations 4 and 6
in addition to the type of curve are self‐constrained. However, other inequality constraints stated in Equations 3, 4, and 7 are
implemented as a penalty function, which is superimposed to the objective function.

3 | STOCHA STI C F R ACTAL SE A RCH AL G O R I T H M

The fractal search is a dynamic algorithm, where its number of agents is modified through the algorithm. The SFSA has 2
main processes, which are the diffusion process and the updating process. In the diffusion process, each agent or particle
diffuses around its current location in order to satisfy the exploitation process, which exceeds the chance of obtaining
4 of 17 EL‐FERGANY AND HASANIEN

global minima. This behavior prevents stuck in local minima. In this algorithm, the diffusion process is static, where the
best particle is taken into consideration and the rest of the particles are ignored.29 In the updating process, the particle
position is updated based on the position of other particles in the group. The updating process refers to the exploration
process of the SFSA.
To generate new particles from the diffusion process, the Gaussian distribution method is used, and it can be represented by
Equations 8 and 9:
 
GW1 ¼ GaussianðμBP ; σ Þ þ ε:BP−ε′ :Pi (8)

GW2 ¼ GaussianðμP ; σ Þ (9)

where ε and ε′ are random numbers lied in the range of 0 and 1. BP and Pi are the best particle and the ith particle positions,
respectively. In Equation 8, μBP and σ are Gaussian parameters and μBP is equal to |BP|. In Equation 9, μP is |Pi|. The standard
deviation σ is written by Equation 10:
 
 logðgÞ 

σ¼ :ðPi −BPÞ (10)
g

logðgÞ
where g is the number of iterations. The term g is implemented with the purpose of reaching the optimal solution in an
expedite way.
For a D‐dimensional design variable optimization problem, the particles are initially generated randomly within the search
space. The particle initialization can be addressed as follows:

Pi ¼ LB þ εðUB−LBÞ (11)

where LB and UB are the lower and upper bounds of the problem vector constraints, respectively.
After the initialization process, the fitness function is evaluated for each particle to get BP among all particles. Moreover, the
particles diffuse around their positions to satisfy the exploitation procedure. Then 2 statistical procedures are used in the
updating process. In the first procedure, all the particles are ranked based on their fitness values. Each particle possesses a
probability value obeying a simple uniform distribution showed in Equation 12:

rankðPi Þ
Pai ¼ (12)
N

where N is the number of particles. The higher probability means the better particle. In other words, Pai tends to 0 for the worst
particle and tends to 1 for the best particle. This equation tries to enhance the worst particles solution. For all particles, if Pai < ε,
the particle position will be updated by Equation 13; otherwise, the particle position will remain unchanged.

P′i ðjÞ ¼ Pr ðjÞ−ε:ðPt ðjÞ−Pi ðjÞÞ (13)

where P′i ðjÞ is the updated position of Pi. Pr and Pt are the position of random particles.
In the second procedure of the updating process, all the particles obtained by the first procedure are ranked again by
Equation 12. For all particles P′i, if P′ai <ε′, the particle position will be updated by Equations 14 and 15; otherwise, the particle
position will remain unchanged:
 
P′′i ¼ P′i −ε′ : P′t −BP if ε′ ≤0:5 (14)

 
P′′i ¼ P′i þ ε′ : P′t −P′r if ε′ >0:5 (15)

where P′′i is the updated position of P′i : P′r and P′t are updated positions of random Pr and Pt particles. The new particle P′′i is
replaced by P′i if it has a better fitness value. Figure 1 demonstrates a detailed flowchart of the proposed SFSA. The main
parameters that define the algorithm characteristics are N, g, and the maximum diffusion number Ndiff. In this study, the second
Gaussian distribution method GW2 is utilized, where it can deal with nonlinear constraints problems.
EL‐FERGANY AND HASANIEN 5 of 17

FIGURE 1 Flowchart of the SFSA procedures

4 | S I M U L AT I O N R E S U LTS A N D D I S C U S S I O N S

The SFSA is applied to solve the coordination problem of directional overcurrent relays for different interconnected power sys-
tems. The simulation results are implemented using MATLAB environment from MathWorks.34 It is worth mentioning that the
MATLAB code offered by Salimi at the official MATLAB website35 is used. In this paper, the 8‐bus, 9‐bus, and 15‐bus net-
work systems are demonstrated. For the purpose of fair comparisons with other challenging methods, the 3‐phase short circuit
faults are considered. These numerical results are implemented using a computer with Intel Core i7 processor, 8 MB RAM with
Windows 10 operating system. TD and I P are continuous design variables and the time‐current characteristic of the relay is
treated as a discrete design integer variable. The optimal characteristics of the proposed SFSA include N = 50, 50, and 60,
g = 1000, 1000, and 2000 for the 8‐bus, 9‐bus, and 15‐bus meshed systems, respectively, and Ndiff equals 2. The controlling
parameters of other optimizers used for comparisons are the same found in their cited corresponding papers and hence are
not repeated here. The TOT of the main and backup relays is minimized as stated in Equation 2 subject to a set of constraints
presented in Equations 3–7. The simulation results are investigated as follows:
6 of 17 EL‐FERGANY AND HASANIEN

4.1 | The 8‐bus network system


The single line diagram of the 8‐bus network system is demonstrated in Figure 2, and the system data are given in.21,25,36 IP , min, k
and IP , max, k are equal to 100% and 150% of the maximum load current, respectively. Different scenarios are implemented
for more investigation of the coordination problem. In scenario 1, the coordination problem is studied under variable time‐
current characteristics of the relay. In this scenario, TDmin, i and TDmax,i are selected as 0.01 and 1 s, respectively. The best
fitness value of the optimization problem records 8.14 s, and the TOT of the primary relays is 1.39 s. Table 2 illustrates

FIGURE 2 Single line diagram of the 8‐bus network system19,22

TABLE 2 Optimal values of the design variables for the 8‐bus system

Scenario 1 Scenario 2 Scenario 3


Relay
no. IP(A) TD(s) Curve IP(A) TD(s) IP(A) TD(s) Curve

1 417.31 0.03 VI 551.91 0.05 422.38 0.05 VI


2 673.31 0.10 EI 985.31 0.10 716.36 0.10 EI
3 507.17 0.09 EI 668.78 0.09 502.11 0.08 EI
4 757.15 0.05 VI 830.90 0.07 679.00 0.05 EI
5 493.28 0.03 VI 583.39 0.05 462.48 0.04 VI
6 527.40 0.10 EI 486.68 0.10 470.24 0.10 VI
7 578.07 0.04 EI 589.75 0.10 566.86 0.07 VI
8 496.47 0.10 EI 491.36 0.10 461.91 0.10 VI
9 451.23 0.03 VI 627.23 0.05 469.65 0.04 VI
10 463.21 0.10 EI 458.54 0.10 524.99 0.08 EI
11 602.91 0.06 EI 601.60 0.10 580.07 0.06 EI
12 592.82 0.20 EI 568.88 0.20 497.14 0.20 EI
13 611.65 0.02 VI 540.00 0.06 501.33 0.04 VI
14 686.26 0.04 VI 707.96 0.10 674.22 0.05 VI
14 1.3876 4.6455 1.7032
∑ t m;i ðsÞ
i¼1
EL‐FERGANY AND HASANIEN 7 of 17

the optimal values of design variables using the SFSA. The convergence of fitness value is depicted in Figure 3. It is worth of
noting here that the convergence is fast and smooth. The operating times of the primary and backup relays are depicted in
Table 3. The results have shown that no violation of the coordination constraints exists, where the value of CTM is between
0.2 and 0.4 s. The SFSA method is tested by making a comparison to other methods, as shown in Table 4. It can be realized
that the coordination problem has a lower time using the SFSA than that by using other optimization methods.

FIGURE 3 Objective function


convergence for the 8‐bus network system

TABLE 3 Operating times of the primary and backup relays for the 8‐bus system

Scenario 1 Scenario 2 Scenario 3


Relay pairs tm (s) tb (s) CTM (s) tm (s) tb (s) CTM (s) tm (s) tb (s) CTM (s)

1 6 0.0669 0.2736 0.2067 0.2047 0.4136 0.2089 0.1001 0.3034 0.2034


2 1 0.1225 0.3252 0.2027 0.4178 0.6200 0.2022 0.1166 0.4901 0.3735
2 7 0.1225 0.3228 0.2003 0.4178 0.6238 0.2060 0.1166 0.3973 0.2808
3 2 0.1445 0.3480 0.2035 0.3819 0.5869 0.2050 0.1294 0.3323 0.2029
4 3 0.1682 0.3747 0.2064 0.3261 0.5296 0.2034 0.1284 0.3352 0.2069
5 4 0.1019 0.3097 0.2077 0.2678 0.4680 0.2002 0.1340 0.3352 0.2012
6 5 0.0751 0.2763 0.2012 0.3075 0.5309 0.2234 0.1486 0.3536 0.2050
6 14 0.0751 0.3268 0.2517 0.3075 0.6957 0.3882 0.1486 0.3786 0.2299
7 5 0.0388 0.2763 0.2375 0.3297 0.5309 0.2012 0.1129 0.3536 0.2407
7 13 0.0388 0.3300 0.2912 0.3297 0.7049 0.3752 0.1129 0.5128 0.3999
8 7 0.0599 0.3228 0.2629 0.3280 0.6238 0.2958 0.1362 0.3973 0.2611
8 9 0.0599 0.2868 0.2269 0.3280 0.5472 0.2192 0.1362 0.3374 0.2012
9 10 0.1007 0.3026 0.2019 0.2443 0.4481 0.2038 0.1165 0.3172 0.2007
10 11 0.1213 0.3214 0.2001 0.3528 0.5534 0.2006 0.1263 0.3283 0.2020
11 12 0.1232 0.3254 0.2022 0.4120 0.6157 0.2037 0.1263 0.3267 0.2004
12 13 0.1265 0.3300 0.2035 0.4911 0.7049 0.2138 0.1276 0.5128 0.3852
12 14 0.1265 0.3268 0.2003 0.4911 0.6957 0.2046 0.1276 0.3786 0.2509
13 8 0.0521 0.2539 0.2018 0.2456 0.4605 0.2149 0.1000 0.3033 0.2032
14 1 0.0860 0.3252 0.2391 0.3362 0.6200 0.2838 0.1004 0.4901 0.3897
14 9 0.0860 0.2868 0.2007 0.3362 0.5472 0.2110 0.1004 0.3374 0.2370
TOT (s) 1.8964 6.2451 4.3483 6.8558 11.5208 4.6649 2.4456 7.5212 5.0756
8 of 17 EL‐FERGANY AND HASANIEN

TABLE 4 Comparison among the SFSA and other optimizers for the 8‐bus system

Scenario 1 Scenario 2

Method ∑14
i¼1 tm;i ðsÞ ∑20
k¼1 CTMk ðsÞ Method ∑14
i¼1 tm;i ðsÞ ∑20
k¼1 CTMk ðsÞ

FPA19a 1.73 5.32 FPA19a 5.80 4.90


21b
Proposed SFSA 1.39 4.35 SA 8.43 7.25
Proposed SFSA 4.65 4.65
a
The objective function is different and all other conditions are the same.
b
The pickup currents values are treated as discrete variables and CTMmin is equal to 0.3 s.

In addition to the above, scenario 2 cares solving the coordination problem under fixed NI time‐current characteristics of the
relay. The best fitness value of the optimization problem records 18.38 s and the TOT of the primary relays is 4.65 s. Table 2
illustrates the optimal values of design variables using the SFSA. The convergence of fitness value is depicted in Figure 3. It is
worth of noting here that the convergence is fast and smooth. The operating times of the relays are depicted in Table 3. The
results have shown that no violation of the coordination constraints exists. Moreover, to validate the proposed methodology,
a detailed comparison is made among other methods to solve the same problem, as indicated in Table 4. It can be noted from
this comparison that the proposed SFSA is superior to other methods, where the fitness value using the SFSA possesses a
shorter time with lesser sum of CTMs.
In addition, scenario 3 proposes the coordination problem solution under a minimum relay operating time condition. In this
scenario, the coordination problem is solved under variable time‐current characteristics of the relay. For obtaining realistic
responses, a minimum relay operating time condition is applied, and its value is selected to be 0.1 s. The best fitness value
of the optimization problem records 9.97 s, and the TOT of the primary relays is 1.70 s. Table 2 illustrates the optimal values
of design variables using the SFSA. The convergence of fitness value is depicted in Figure 3. The operating times of both of the
primary and backup relays are revealed in Table 3. It can be emphasized that the results have shown that no violation of the
coordination constraints exists.

4.2 | The 9‐bus network system


The single line diagram of the 9‐bus meshed network system is revealed in Figure 4, and the system data are given in several
studies.1,10,25,27 The system consists of 12 lines, 24 relays, and 32 experiments of primary‐backup relay pairs. The 3‐phase

FIGURE 4 Single line diagram of the 9‐bus network system27


EL‐FERGANY AND HASANIEN 9 of 17

short circuit faults are considered at middle of lines, and the fault currents are pointed out by Bedekar and Bhide.10 The fault at
the mid of line L9, relay no. 17, is supposed to have the relays (22 and 24) to act among its backup ones. However, these
relays canoot sense this fault current, which is lower than their normal load currents. Likewise, for faults at mid of lines
L10, L11, and L12, relays nos. 19, 21, and 23 are supposed to have relays (22 and 24), (18 and 24), and (18, 20 and 22),
respectively, to act among their backups. However, those relays cannot sense such said faults. Therefore, it can be concluded
that relays nos. 17, 19, 21, and 23 have no backup for such faults.1,10,24-27 In other words, one can say that relays nos. 18, 20,
22, and 24 cannot sense faults at the mid of lines L9, L10, L11, and L12, respectively. TDmin, i and TDmax, i are selected as
0.01 and 1.1 s, and IP , min, i and IP , max, i are equal to 120% and 200% of the maximum load current, respectively. In this case
study, 3 scenarios are tested. The design variables of the coordination problem are TD, IP, and the type of time‐current char-
acteristics/curves of the relay. These variables are considered to define scenario 1 (ie, number of control variables is equal to
72) and Tmin equals 0.10 s. On the other hand, in scenario 2, the used curve for all relays is NI while optimizing the TD, and IP
for each relay (ie, number of control variables is equal to 48). Then scenario 3 is made for further verification of the
proposed algorithm, where the 3‐phase short circuit faults are considered near to the relays (near‐end faults) and the
fault currents are described in.1 The same design variables and constraints are applied similar to scenario 2 (ie, the type
of time‐current characteristics of the relay is fixed at NI curve). In scenarios 2 and 3, Tmin equals 0.2 s similar other
competitive methods in order to maintain the same conditions for fair comparisons. The maximum operating time of
the relays along the network is selected as 2 s for all scenarios.
The SFSA is applied to solve the coordination problem, and the fitness best values are 14.86 s, 33.71 s, and 59.77 s for
scenarios 1, 2, and 3, respectively. The convergences of fitness values are shown using a semilogarithmic scale, which are
depicted in Figure 5. It is worth mentioning that the average elapsed time to implement 1 run (1000 iterations) when considering
3 control variables per each relay is 18.50 s. However, when 2 control variables per each relay (NI fixed curve) are used, the
average elapsed time is 11.20 s.
Table 5 defines the optimal values of design variables using the SFSA for these scenarios obtained over 50 independent
runs. The operating times of the primary and backup relays are arranged in Table 6. The results have shown that no violation
of the coordination constraints exists. It can be seen that in scenario 1, none of relays operates in time faster than 0.10 s, and for
other scenarios, all relays operate in time greater than 0.2 s in order to claim the reasonableness of a realistic study (refer to
Table 6).
The SFSA is compared with other methods such as hybrid gravitational search algorithm‐sequential quadratic pro-
gramming (GSA‐SQP)24 and DE,26 which are presented in Table 7. In this comparison, the same conditions and con-
straints are applied as stated in scenario 2. The reader can see obviously that lower values of the sum of tm and
sum of CTM are obtained by the SFSA as indicated in Table 7. On the whole, it can be noted from this comparison
that the proposed SFSA is superior to other competing methods, where the fitness value using the SFSA possesses a
shorter time.

FIGURE 5 Objective function


convergence for the 9‐bus network system
10 of 17 EL‐FERGANY AND HASANIEN

TABLE 5 Optimal values of the design variables for the 9‐bus system

Scenario 1 Scenario 2 Scenario 3


Relay no. IP(A) TD(s) Curve IP(A) TD(s) IP(A) TD(s)

1 158.67 0.20 VI 227.04 0.20 216.47 0.40


2 261.37 0.04 VI 380.20 0.05 295.66 0.20
3 29.35 0.10 LI 26.90 0.40 37.51 0.60
4 27.67 0.10 LI 28.70 0.30 34.64 0.50
5 96.62 0.20 VI 104.51 0.20 94.96 0.40
6 100.06 0.30 VI 94.78 0.20 102.94 0.50
7 102.65 0.30 VI 104.52 0.30 114.65 0.50
8 106.68 0.10 VI 98.09 0.20 97.25 0.30
9 27.82 1.00 VI 31.04 0.30 29.85 0.50
10 27.80 0.10 LI 31.03 0.30 28.25 0.60
11 147.96 0.10 VI 150.82 0.10 182.70 0.20
12 151.89 0.10 VI 206.40 0.10 185.64 0.40
13 45.17 0.60 VI 45.15 0.30 47.97 0.60
14 39.14 0.09 LI 37.65 0.30 41.09 0.60
15 47.33 0.70 VI 38.15 0.30 41.95 0.60
16 37.12 0.08 LI 37.47 0.30 46.39 0.60
17 561.52 0.03 VI 585.21 0.08 713.40 0.20
18 653.14 0.02 VI 666.60 0.06 610.34 0.10
19 542.33 0.03 VI 664.43 0.06 663.67 0.20
20 513.22 0.03 EI 508.94 0.08 582.05 0.20
21 589.33 0.03 VI 671.50 0.06 534.08 0.30
22 568.26 0.02 VI 536.93 0.06 725.57 0.10
23 612.00 0.03 VI 716.77 0.06 632.00 0.30
24 629.43 0.01 VI 650.75 0.05 609.43 0.09
14 2.2526 8.1168 15.7730
∑ t m;i ðsÞ
i¼1

4.3 | The 15‐bus network system


Figure 6 shows the single line diagram of the 15‐bus network system, which is highly penetrated by 6 DGs. The data of this
system are given in.19,21,27 The reader may notice that the formulation of this test case is complicated as an optimization
problem with high nonlinearity. The system consists of 42 relays and 82 experiments of primary‐backup relay pairs. TDmin, i
and TDmax, i are selected as 0.01 and 1 s, and the range of IP, min, i and IP , max, i is 100% and 150% of the maximum load current,
respectively. To simulate real situation in the field, the minimum and maximum operating times of relays along the network are
equal to [50 ms, 2 s], respectively. In this case, 2 scenarios are analyzed. Scenario 1 utilizes optimized relays characteristics, and
the type of time‐current characteristics of the relay is fixed NI curve in scenario 2.
The optimal values of design variables using the SFSA are stipulated in Table 8 for both scenarios, and the fitness values are
29.96 s and 64.51 s, respectively (cropped over 50 runs). Table 9 lists the operating time of the primary and backup relays. It can
be found that the coordination constraints are not violated and the CTM is within the allowable range. The trend of fitness func-
tion convergences is characterized in Figure 7. It is worthy to state that the elapsed times are 60.50 s and 44.45 s in order to
implement 2000 iterations for scenarios 1 and 2, respectively. It can be seen that when optimized curves are offered, the pro-
posed SFSA tends to decide on VI and EI curves as the nature of these curves possess faster responses than by NI curve.
The SFSA is checked again by making a comparison with other methods for this system. In this comparison, the same condi-
tions and constraints are applied which is presented in Table 10. It can be noted from this comparison that the proposed SFSA
EL‐FERGANY AND HASANIEN 11 of 17

TABLE 6 Operating times of the primary and backup relays for the 9‐bus system

Scenario 1 Scenario 2 Scenario 3


Relay pairs tm (s) tb (s) CTM (s) tm (s) tb (s) CTM (s) tm (s) tb (s) CTM (s)

1 15 0.1029 0.3757 0.2728 0.4319 0.6345 0.2026 0.9074 1.1223 0.2148


1 17 0.1029 0.3229 0.2200 0.4319 0.6897 0.2578 0.9074 1.1109 0.2034
2 4 0.1013 0.3185 0.2172 0.2535 0.4847 0.2311 0.6104 0.8144 0.2039
3 1 0.1291 0.4024 0.2733 0.5287 0.7485 0.2198 0.9015 1.1105 0.2091
4 6 0.1253 0.3923 0.2670 0.3826 0.5863 0.2037 0.7446 0.9453 0.2007
5 3 0.1187 0.3280 0.2093 0.4537 0.6646 0.2109 0.8119 1.0155 0.2036
6 8 0.1032 0.3192 0.2160 0.3866 0.5993 0.2127 0.8080 1.0084 0.2003
6 23 0.1032 0.3244 0.2212 0.3866 0.6771 0.2905 0.8080 1.0484 0.2404
7 5 0.1011 0.3247 0.2236 0.4530 0.6768 0.2237 0.8364 1.0454 0.2090
7 23 0.1011 0.3244 0.2233 0.4530 0.6771 0.2240 0.8364 1.0484 0.2120
8 10 0.1155 0.3195 0.2040 0.4060 0.6169 0.2109 0.7814 1.0017 0.2203
9 7 0.1498 0.3849 0.2351 0.4909 0.6962 0.2053 0.7785 0.9832 0.2047
10 12 0.1259 0.4357 0.3099 0.4869 0.7171 0.2303 0.8958 1.1088 0.2130
11 9 0.1438 0.3807 0.2369 0.4073 0.6246 0.2173 0.6663 0.8743 0.2080
12 14 0.1041 0.3677 0.2636 0.3629 0.5808 0.2179 0.9152 1.1318 0.2167
12 21 0.1041 0.3229 0.2189 0.3629 0.6754 0.3125 0.9152 1.1404 0.2252
13 11 0.1042 0.4227 0.3185 0.4655 0.6680 0.2025 0.8698 1.0720 0.2023
13 21 0.1042 0.3229 0.2187 0.4655 0.6754 0.2099 0.8698 1.1447 0.2749
14 16 0.1004 0.3753 0.2748 0.4183 0.6576 0.2393 0.9158 1.1189 0.2031
14 19 0.1004 0.3043 0.2039 0.4183 0.6596 0.2413 0.9158 1.4117 0.4959
15 13 0.1021 0.3845 0.2824 0.4565 0.6634 0.2069 0.9071 1.1106 0.2035
15 19 0.1021 0.3043 0.2022 0.4565 0.6596 0.2031 0.9071 1.1079 0.2008
16 2 0.1023 0.3544 0.2521 0.4691 0.6888 0.2197 0.8780 1.0901 0.2121
16 17 0.1023 0.3229 0.2206 0.4691 0.6897 0.2206 0.8780 1.1159 0.2379
18 2 0.1013 0.3544 0.2531 0.3124 0.6888 0.3763 0.6371 1.0946 0.4574
18 15 0.1013 0.3757 0.2743 0.3124 0.6345 0.3221 0.6371 1.1262 0.4891
20 13 0.1066 0.3845 0.2779 0.3153 0.6634 0.3481 0.6617 1.1132 0.4515
20 16 0.1066 0.3753 0.2687 0.3153 0.6576 0.3423 0.6617 1.1215 0.4598
22 11 0.1058 0.4227 0.3168 0.2918 0.6680 0.3762 0.6922 1.0749 0.3828
22 14 0.1058 0.3677 0.2619 0.2918 0.5808 0.2890 0.6922 1.1358 0.4437
24 5 0.1091 0.3247 0.2156 0.3439 0.6768 0.3329 0.5540 1.0494 0.4954
24 8 0.1091 0.3192 0.2102 0.3439 0.5993 0.2554 0.5540 1.0122 0.4583
TOT (s) 3.4956 11.3594 7.8638 12.8240 20.8809 8.0566 25.3558 34.4093 9.0536

method is tremendous of superiority to other heuristic‐based optimization methods. Therefore, the SFSA is capable of solving
the coordination problem of the directional overcurrent relays successfully.
It can be emphasized that Figures 3, 5, and 7 take into consideration the constraint's penalty function which confirm that all
stated constraints are respected by the SFSA.

4.4 | The SFSA performance measures using parametric and nonparametric tests
At last, the parametric statistical and nonparametric Wilcoxon test are used to certify the performance of the SFSA for
solving the optimal relay coordination problem. A significant level confidence of 1% is suggested. The copped results of
12 of 17 EL‐FERGANY AND HASANIEN

TABLE 7 Comparisons among the SFSA and other methods for the 9‐bus system (scenario 2)

Method ∑32
k¼1 CTMk ðsÞ ∑24
i¼1 tm;i ðsÞ

GSA24a 14.74 18.22


24a
SQP 10.20 12.46
24a
GSA‐SQP 8.89 11.47
26a
DE 14.05 13.11
Proposed SFSA 8.12 8.06
a
CTMmin equals 0.2 s.

FIGURE 6 Single line diagram of the 15‐bus network system19,22


EL‐FERGANY AND HASANIEN 13 of 17

TABLE 8 Optimal values of the design variables for the 15‐bus system

Scenario 1 Scenario 2
Relay no. IP(A) TD(s) Curve IP(A) TD(s)

1 200.37 0.07 VI 197.69 0.10


2 238.83 0.07 VI 237.37 0.10
3 349.45 0.06 VI 336.36 0.10
4 297.09 0.08 VI 316.06 0.10
5 354.92 0.06 VI 360.40 0.10
6 252.46 0.07 EI 307.25 0.09
7 251.24 0.06 EI 249.79 0.10
8 379.01 0.06 VI 413.63 0.10
9 299.08 0.05 VI 333.91 0.07
10 272.81 0.06 VI 283.21 0.10
11 303.03 0.07 VI 319.75 0.10
12 378.62 0.06 VI 350.07 0.10
13 354.34 0.04 VI 363.94 0.07
14 275.80 0.06 VI 294.91 0.10
15 246.56 0.07 VI 255.35 0.10
16 176.47 0.07 VI 173.57 0.10
17 132.57 0.07 VI 127.12 0.10
18 361.21 0.08 VI 351.18 0.10
19 286.01 0.07 VI 296.61 0.10
20 424.31 0.07 VI 438.60 0.10
21 379.55 0.08 VI 364.15 0.10
22 135.71 0.07 VI 135.63 0.10
23 271.68 0.06 VI 269.85 0.10
24 177.01 0.07 VI 177.15 0.10
25 236.52 0.06 VI 239.59 0.10
26 213.73 0.07 VI 206.30 0.10
27 275.43 0.05 EI 238.25 0.10
28 340.05 0.06 VI 341.36 0.10
29 463.70 0.07 VI 456.42 0.10
30 147.69 0.07 VI 137.68 0.10
31 229.92 0.06 VI 225.26 0.10
32 194.33 0.06 VI 191.94 0.10
33 337.84 0.05 VI 300.42 0.10
34 202.07 0.08 EI 202.51 0.10
35 248.47 0.06 EI 252.56 0.10
36 293.33 0.06 VI 296.95 0.10
37 401.85 0.06 VI 420.26 0.10
38 219.38 0.05 EI 213.14 0.10
39 200.84 0.06 EI 226.30 0.09
40 412.10 0.06 VI 411.15 0.10

(Continues)
14 of 17 EL‐FERGANY AND HASANIEN

TABLE 8 (Continued)

Scenario 1 Scenario 2
Relay no. IP(A) TD(s) Curve IP(A) TD(s)

41 221.63 0.06 VI 205.97 0.10


42 252.39 0.06 VI 287.25 0.09

∑42
i¼1 t m;i ðsÞ
3.2093 12.0309

TABLE 9 Operating times of the primary and backup relays for the 15‐bus system

Scenario 1 Scenario 2 Scenario 1 Scenario 2


Relay Relay
pairs tm (s) tb (s) CTM (s) tm (s) tb (s) CTM (s) pairs tm (s) tb (s) CTM (s) tm (s) tb (s) CTM (s)

1 6 0.054 0.261 0.207 0.248 0.452 0.204 21 17 0.051 0.256 0.206 0.234 0.467 0.233
2 4 0.052 0.258 0.206 0.241 0.452 0.211 21 19 0.051 0.256 0.206 0.234 0.459 0.226
2 16 0.052 0.285 0.233 0.241 0.498 0.257 21 30 0.051 0.254 0.203 0.234 0.448 0.215
3 1 0.082 0.284 0.202 0.282 0.500 0.218 22 23 0.072 0.329 0.258 0.271 0.590 0.319
3 16 0.082 0.285 0.202 0.282 0.498 0.216 22 34 0.072 0.278 0.207 0.271 0.480 0.209
4 7 0.074 0.279 0.204 0.262 0.476 0.214 23 11 0.050 0.257 0.207 0.258 0.464 0.207
4 12 0.074 0.277 0.202 0.262 0.473 0.211 23 13 0.050 0.283 0.233 0.258 0.480 0.222
4 20 0.074 0.277 0.202 0.262 0.487 0.225 24 21 0.078 0.430 0.352 0.265 0.577 0.313
5 2 0.100 0.330 0.230 0.321 0.534 0.214 24 34 0.078 0.278 0.200 0.265 0.480 0.215
6 8 0.055 0.281 0.227 0.290 0.500 0.211 25 15 0.090 0.323 0.234 0.318 0.523 0.205
6 10 0.055 0.261 0.207 0.290 0.514 0.225 25 18 0.090 0.421 0.331 0.318 0.527 0.209
7 5 0.053 0.285 0.232 0.306 0.530 0.224 26 28 0.093 0.301 0.209 0.298 0.531 0.234
7 10 0.053 0.261 0.209 0.306 0.514 0.208 26 36 0.093 0.299 0.206 0.298 0.556 0.258
8 3 0.076 0.278 0.202 0.269 0.488 0.219 27 25 0.073 0.277 0.204 0.341 0.547 0.206
8 12 0.076 0.277 0.201 0.269 0.473 0.204 27 36 0.073 0.299 0.226 0.341 0.556 0.215
8 20 0.076 0.277 0.200 0.269 0.487 0.218 28 29 0.117 0.319 0.202 0.330 0.542 0.213
9 5 0.072 0.285 0.213 0.216 0.530 0.314 28 32 0.117 0.323 0.205 0.330 0.541 0.212
9 8 0.072 0.281 0.210 0.216 0.500 0.285 29 17 0.055 0.256 0.201 0.255 0.467 0.212
10 14 0.066 0.268 0.203 0.272 0.499 0.227 29 19 0.055 0.256 0.201 0.255 0.459 0.204
11 3 0.075 0.278 0.204 0.269 0.488 0.218 29 22 0.055 0.257 0.202 0.255 0.470 0.215
11 7 0.075 0.279 0.204 0.269 0.476 0.207 30 27 0.085 0.288 0.203 0.280 0.497 0.216
11 20 0.075 0.277 0.202 0.269 0.487 0.218 30 32 0.085 0.323 0.237 0.280 0.541 0.261
12 13 0.079 0.283 0.204 0.269 0.480 0.210 31 27 0.066 0.288 0.222 0.276 0.497 0.220
12 24 0.079 0.287 0.208 0.269 0.474 0.205 31 29 0.066 0.319 0.253 0.276 0.542 0.266
13 9 0.065 0.268 0.203 0.225 0.429 0.204 32 33 0.087 0.303 0.216 0.290 0.557 0.266
14 11 0.056 0.257 0.201 0.248 0.464 0.217 32 42 0.087 0.293 0.206 0.290 0.550 0.260
14 24 0.056 0.287 0.231 0.248 0.474 0.227 33 21 0.127 0.430 0.303 0.367 0.577 0.211
15 1 0.052 0.284 0.232 0.236 0.500 0.265 33 23 0.127 0.329 0.203 0.367 0.590 0.223
15 4 0.052 0.258 0.205 0.236 0.452 0.216 34 31 0.086 0.301 0.215 0.350 0.557 0.207
16 18 0.079 0.421 0.342 0.281 0.527 0.246 34 42 0.086 0.293 0.206 0.350 0.550 0.201
16 26 0.079 0.279 0.201 0.281 0.490 0.209 35 25 0.064 0.277 0.213 0.330 0.547 0.216
(Continues)
EL‐FERGANY AND HASANIEN 15 of 17

TABLE 9 (Continued)

Scenario 1 Scenario 2 Scenario 1 Scenario 2


Relay Relay
pairs tm (s) tb (s) CTM (s) tm (s) tb (s) CTM (s) pairs tm (s) tb (s) CTM (s) tm (s) tb (s) CTM (s)

17 15 0.069 0.323 0.255 0.266 0.523 0.257 35 28 0.064 0.301 0.238 0.330 0.531 0.201
17 26 0.069 0.279 0.211 0.266 0.490 0.224 36 38 0.082 0.287 0.206 0.301 0.539 0.238
18 19 0.050 0.256 0.206 0.216 0.459 0.244 37 35 0.111 0.359 0.248 0.343 0.550 0.207
18 22 0.050 0.257 0.207 0.216 0.470 0.254 38 40 0.109 0.322 0.213 0.405 0.614 0.209
18 30 0.050 0.254 0.203 0.216 0.448 0.233 39 37 0.103 0.311 0.208 0.341 0.580 0.239
19 3 0.075 0.278 0.203 0.267 0.488 0.220 40 41 0.117 0.322 0.206 0.368 0.597 0.229
19 7 0.075 0.279 0.204 0.267 0.476 0.209 41 31 0.097 0.301 0.205 0.336 0.557 0.220
19 12 0.075 0.277 0.202 0.267 0.473 0.205 41 33 0.097 0.303 0.206 0.336 0.557 0.220
20 17 0.053 0.256 0.203 0.238 0.467 0.229 42 39 0.063 0.272 0.209 0.256 0.460 0.204
20 22 0.053 0.257 0.204 0.238 0.470 0.232 TOT (s) 6.065 23.892 17.837 23.00 41.512 18.521
20 30 0.053 0.254 0.201 0.238 0.448 0.211

FIGURE 7 Objective function


convergence for the 15‐bus network system

TABLE 10 Comparison among the SFSA and other methods for the 15‐bus system

Scenario 1 Scenario 2
42 82 42 82
∑ tm;i ðsÞ ∑ CTMk ðsÞ ∑ tm;i ðsÞ ∑ CTMk ðsÞ
Method i¼1 k¼1 Method i¼1 k¼1
19a 19a
FPA 2.95 20.31 FPA 13.07 20.88
Proposed SFSA 3.21 17.84 SA21b 12.23 18.21
22
EFO: electromagnetic field optimizer EFO 17.91 19.16
MEFO: modified EFO MEFO22 13.95 21.73
Proposed SFSA 12.03 18.52
a
The objective function is different, no limit for Tmin and all other conditions are the same, andbThe pickup currents values are treated as discrete variables and CTMmin
equals 0.2 s.
16 of 17 EL‐FERGANY AND HASANIEN

TABLE 11 Parametric and nonparametric performance tests of the SFSA runs

Scenarios of the 8‐bus system Scenarios of the 9‐bus system Scenarios of the 15‐bus system
Indicator 1 2 3 1 2 3 1 2

Best 8.14 18.38 9.67 14.86 33.71 59.77 29.96 64.51


Worst 8.52 19.73 10.24 16.53 36.77 64.95 32.80 68.44
Mean 8.30 18.92 10.10 15.71 35.26 62.40 31.40 66.50
Median 8.28 18.70 10.10 15.75 35.33 62.53 31.47 66.60
σ 0.14 0.58 0.12 0.49 0.90 1.52 0.84 1.16
Variance 0.19 0.34 0.14 0.24 0.81 2.31 0.70 1.34
h 0 0 0 0 0 0 0 0
P value (T‐test) 0.47 0.47 0.47 0.47 0.47 0.47 0.47 0.47

the fitness function values, which specified in Equation 2 after 50 independent runs, are implemented per each scenario.
The entries of Table 11 stipulates the various indicators used to test the SFSA.
As a result of the statistical test, the small values of standard deviation (σ) and variance of various experiments indicate the
robustness of the SFSA. In addition, the null hypotheses with 0.01 confidence (h = 0) are approved as the small values of P
value obtained by T‐test of pair samples (equal variances and medians are not assumed between samples).

5 | CONCLUSIONS

This paper has presented a novel application of the SFSA with the purpose of solving the coordination problem of directional
overcurrent relays in interconnected power systems. The optimization problem has continuous and discrete design variables.
The effectiveness of the SFSA is compared with other optimization methods for various mesh power systems such as the 8‐,
9‐, and 15‐bus networks. The simulation results have proven the proposed algorithm efficiency for solving the coordination
problem of such systems. It is found that the SFSA outperforms the other evolutionary optimizers for solving the coordination
problem. The applied algorithm shows high convergence speed, needs lower parameters to be tuned, and gives excellent results.
Finally, the SFSA is a promising algorithm, which can be applied to other optimization problems in power systems.

R E F E RENC E S
1. Adelnia F, Moravej Z, Farzinfar M. A new formulation for coordination of directional overcurrent relays in interconnected networks. Int Trans
Electr Energy Syst. 2015;25(1):120‐137. https://doi.org/10.1002/etep.1828
2. ABB: ‘Protection application handbook’, book no. 6, section 15: protection settings, pp. 301–340, Sweden, ABB Support 1999‐06.
3. Ezzeddine M, Kaczmarek R, Iftikhar MU. Coordination of directional overcurrent relays using a novel method to select their settings. IET Generat
Transm Distrib. 2011;5(7):743‐750. https://doi.org/10.1049/iet‐gtd.2010.0603
4. Mahari A, Seyedi H. An analytic approach for optimal coordination of overcurrent relays. IET Generat Transm Distrib. 2013;7(7):674‐680. https://
doi.org/10.1049/iet‐gtd.2012.0721
5. Keil T, Jager J. Advanced coordination method for overcurrent protection relays using nonstandard tripping characteristics. IEEE Trans Power
Deliv. 2008;23(1):52‐57. https://doi.org/10.1109/TPWRD.2007.905337
6. Yang XP, Duan XZ, Shi DY. Novel applications of Z‐matrix to network topology analysis in the relay coordination software. IET Generation
Transmission and Distribution. 2007;1(4):540‐547. https://doi.org/10.1049/iet‐gtd:20060486
7. Sueiro JA, Diaz‐Dorado E, Míguez E, Cidrás J. Coordination of directional overcurrent relay using evolutionary algorithm and linear
programming. International Journal of Electrical Power & Energy Systems. 2012;42(1):299‐305. https://doi.org/10.1016/j.ijepes.2012.03.036
8. Huchel L, Zeineldin HH. Planning the coordination of directional overcurrent relays for distribution systems considering DG. IEEE Trans on
Smart Grid. 2016;7(3):1642‐1649. https://doi.org/10.1109/TSG.2015.2420711
9. Bedekar PP, Bhide SR, Kale VS. Optimum time coordination of overcurrent relays using two phase simplex method. World Academy Sci Eng
Technol. 2009;52:1110‐1114.
10. Bedekar PP, Bhide SR. Optimum coordination of directional overcurrent relays using the hybrid GA‐NLP approach. IEEE Trans Power Deliv.
2011;26(1):109‐119. https://doi.org/10.1109/TPWRD.2010.2080289
EL‐FERGANY AND HASANIEN 17 of 17

11. Bottura FB, Bernardes WMS, Oleskovicz M, Asada EN. Setting directional overcurrent protection parameters using hybrid GA optimizer. Electric
Power Systems Research. 2017;143:400‐408. https://doi.org/10.1016/j.epsr.2016.09.017
12. Shih MY, Salazar CAC, Enriquez AC. Adaptive directional overcurrent relay coordination using ant colony optimisation. IET Generation
Transmission and Distribution. 2015;9(14):2040‐2049. https://doi.org/10.1049/iet‐gtd.2015.0394
13. Moirangthem J, Krishnanand KR, Dash SS, Ramaswami R. Adaptive differential evolution algorithm for solving non‐linear coordination problem
of directional overcurrent relays. IET Generat Trans Distrib. 2013;7(4):329‐336. https://doi.org/10.1049/iet‐gtd.2012.0110
14. Chelliah TR, Thangaraj R, Allamsetty S, Pant M. Coordination of directional overcurrent relays using opposition based chaotic differential
evolution algorithm. Int J Electr Power & Energ Syst. 2014;55:341‐350. https://doi.org/10.1016/j.ijepes.2013.09.032
15. Chen C‐R, Lee C‐H, Chang C‐J. Overcurrent relay coordination optimization with partial differentiation approach for the validation of
coordination violation. Electr Power Compono Syst. 2011;39(10):933‐947. https://doi.org/10.1080/15325008.2011.552092
16. Zellagui M, Benabid R, Boudour M, Chaghi A. Application of firefly algorithm for optimal coordination of directional overcurrent protection
relays in presence of series compensation. J Autom Syst Eng. 2014;8(2):92‐107.
17. Singh M, Panigrahi BK, Abhyankar AR. Optimal coordination of directional overcurrent relays using teaching learning‐based optimization
algorithm. Int J Electr Power & Energ Syst. Sept. 2013;50:33‐41. https://doi.org/10.1016/j.ijepes.2013.02.011
18. Hussain MH, Musirin I, Abidin AF, Rahim SRA. Solving directional overcurrent relay coordination problem using artificial bees colony. Int J
Electr Electron Sci Eng. 2014;8(5):705‐710. scholar.waset.org/1999.5/9998279
19. El‐Fergany AA. Optimal directional digital overcurrent relays coordination and arc‐flash hazard assessments in meshed networks. Int Trans Electr
Energy Syst. 2016;26(1):134‐154. https://doi.org/10.1002/etep.2073
20. Saha D, Datta A, Das P. Optimal coordination of directional overcurrent relays in power systems using symbiotic organism search (SOS)
optimization technique. IET Generat Trans Distrib. 2016;10(11):2681‐2688. https://doi.org/10.1049/iet‐gtd.2015.0961
21. Amraee T. Coordination of directional overcurrent relays using seeker algorithm. IEEE Trans Power Deliv. 2012;27(3):1415‐1422. https://doi.org/
10.1109/TPWRD.2012.2190107
22. Bouchekara HREH, Zellagui M, Abido MA. Optimal coordination of directional overcurrent relays using a modified electromagnetic field
optimization algorithm. Applied Soft Comput. 2017;54:267‐283. https://doi.org/10.1016/j.asoc.2017.01.037
23. Othman AM, Abdelaziz AY. Enhanced backtracking search algorithm for optimal coordination of directional over‐current relays including
distributed generation. Electr Power Compono Syst. 2016;44(3):278‐290. https://doi.org/10.1080/15325008.2015.1111468
24. Radosavljević J, Jevtić M. Hybrid GSA‐SQP algorithm for optimal coordination of directional overcurrent relays. IET Generat Trans Distrib.
2016;10(8):1928‐1937. https://doi.org/10.1049/iet‐gtd.2015.1223
25. Albasri FA, Alroomi AR, Talaq JH. Optimal coordination of directional overcurrent relays using biogeography‐based optimization algorithms.
IEEE Trans Power Deliv. 2015;30(4):1810‐1820. https://doi.org/10.1109/TPWRD.2015.2406114
26. Singh M, Panigrahi BK, Abhyankar AR. Optimal coordination of directional over‐current relays using informative differential evolution algo-
rithm. J Comput Sci. 2014;5:269‐276. https://doi.org/10.1016/j.jocs.2013.05.010
27. Alam MN, Das B, Pant V. A comparative study of metaheuristic optimization approaches for directional overcurrent relays coordination. Electric
Power Systems Research. 2015;128:39‐52. https://doi.org/10.1016/j.epsr.2015.06.018
28. Borenstei, Y., Moraglio, A.: ‘Theory and Principled Methods for the Design of Metaheuristics’, Springer‐Verlag Berlin Heidelberg, 2014,
chapter 1, pp. 1‐23. https://doi.org/10.1007/978‐3‐642‐33206‐7
29. Salimi H. Stochastic fractal search: a powerful metaheuristic algorithm. Knowledge‐Based Syst. 2015;75:1‐18. https://doi.org/10.1016/j.knosys.2014.07.025
30. Tejani GG, Bhensdadia VH, Bureerat S. Examination of three meta‐heuristic algorithms for optimal design of planar steel frames. Adv Computat
Des. 2016;1(1):79‐86. https://doi.org/10.12989/acd.2016.1.1.079
31. Rahman TAZ. Parameters optimization of an SVM‐classifier using stochastic fractal search algorithm for monitoring aerospace structure. Int J
Fluids and Heat Transfer. 2016;1:68‐78.
32. IEC 60255‐3: ‘Electrical relay. Part 3: single input energizing quantity measuring relays with dependent or independent time’, Geneva, Switzerland, June 1998.
33. Zellagui M, Benabid R, Chaghi A, Boudour M. Impact of GCSC on IDMT directional overcurrent relay in the presence of phase to earth fault.
Serbian J Electr Eng (SJEE). 2013;10(3):381‐398. https://doi.org/10.2298/SJEE130505011Z
34. Release 2016b, “MATLAB,” The MathWorks press, September 2016.
35. MathWorks/file exchange: http://www.mathworks.com/matlabcentral/fileexchange/47565‐stochastic‐fractal‐search‐‐sfs‐ (Accessed on 20 Dec., 2016)
36. Papaspiliotopoulos VA, Korres GN, Maratos NG. A novel quadratically constrained quadratic programming method for optimal coordination of
directional overcurrent relays. IEEE Trans Power Deliv. 2017;32(1):3‐10. https://doi.org/10.1109/TPWRD.2015.2455015

How to cite this article: El‐Fergany AA, Hasanien HM. Optimized settings of directional overcurrent relays in meshed
power networks using stochastic fractal search algorithm. Int Trans Electr Energ Syst. 2017;e2395. https://doi.org/
10.1002/etep.2395

You might also like