Electric Power Systems Research 237 (2024) 111017

Contents lists available at ScienceDirect

Electric Power Systems Research

journal homepage: www.elsevier.com/locate/epsr

A new and effective directional overcurrent relay coordination approach

for IIDG-based distribution networks using different setting groups for peak
and off-peak demand periods
Alisan Ayvaz
Department of Electrical and Electronics Engineering, Amasya University, 05100, Amasya, Türkiye

A R T I C L E I N F O A B S T R A C T

Keywords: Directional overcurrent relays (DOCRs) must be properly coordinated to avoid vulnerabilities in the protection of
Power system protection distribution systems. Although most researchers have focused on changes in fault currents caused by distributed
Optimal DOCR coordination generation units (DGs), changes in load currents are also important. Due to the stochastic nature of load demand
Relay setting groups
and output power of renewable-based DGs, load current magnitudes can change significantly and reach high
Inverter-interfaced distributed generation units
Enhanced white shark optimizer
values throughout the day. Therefore, to avoid maloperation, high pickup current settings for DOCRs must be
selected. However, at this time, protection coordination may become ineffective. This paper proposes a novel
DOCR coordination strategy for inverter-interfaced DG (IIDG)-based distribution systems by assigning different
setting groups to each DOCR for the peak and off-peak loading periods of the day and using a realistic approach
to determine maximum load currents. The aim is to achieve high effectiveness in DOCR coordination by lowering
selectable pickup settings. To obtain the optimal setting groups, a new variant of the white shark optimizer is
developed and used. The proposed approach reduces the total operating time of relay pairs by up to 24.5 % and
21.5 % for the IEEE 14-bus and IEEE 30-bus distribution systems, respectively, compared to the traditional
coordination approach.

1. Introduction alternative coordination approaches have been developed in recent

years. In [3], an approach is proposed to meet the requirements of
The problem of optimal coordination of directional overcurrent re­ adaptive and robust DOCR coordination and improve the power quality
lays (DOCRs) is solved to minimize the total operating time of relays of power systems, considering changes in network topology due to DG
without violating a set of constraints, including the coordination time outages. Reference [4] proposes a new method to mitigate DGs’ impact
interval (CTI) and relay setting and operating limits. With this solution, on DOCR coordination in a variable topology distribution network using
relay settings—including time dial setting (TDS), pickup (or plug) fault current limiters at optimal locations. A new transient
setting (PS), and, where applicable, the time-current characteristic curve stability-oriented DOCR coordination approach is proposed in [5] to
as specified in IEC standards—are optimally determined for all DOCRs achieve a fast and optimal protection scheme while maintaining the
[1]. Thus, an attempt is made at a fast, selective, and reliable protection stability of systems equipped with DGs. In these studies, the DGs are
coordination scheme. considered synchronous DGs because their impact on fault currents is
Recent studies on this topic show that there is a great effort to meet much greater than those of inverter-interfaced ones [6]. In fact,
the requirements for the protection of today’s complex power networks. inverter-interfaced DGs (IIDGs) (i.e. wind-based DGs) can also consid­
Distributed generation units (DGs) are important components of modern erably contribute to fault currents, but this contribution is limited to a
distribution networks that can change the magnitude and direction of few cycles after the fault occurrence [7]. Therefore, considering CTI and
load and fault currents, thereby leading to maloperation and/or mis­ interruption time of fault current by circuit breaker, IIDGs’ effects on
coordination of DOCRs [2]. These effects may arise from a DG connected fault currents are negligible for backup protection and hence DOCR
to a distribution network and also from network topology changes due to coordination studies [8]. However, load current changes due to DGs
the disconnection of a DG. At this point, traditional coordination ap­ cannot directly depend on the type of DG (i.e. synchronous or
proaches cannot overcome the issues caused by DGs. Therefore, several inverter-interfaced). It is crucial to calculate maximum load currents at

E-mail address: [email protected].

https://doi.org/10.1016/j.epsr.2024.111017
Received 23 June 2024; Received in revised form 20 August 2024; Accepted 22 August 2024
Available online 29 August 2024
0378-7796/© 2024 Elsevier B.V. All rights are reserved, including those for text and data mining, AI training, and similar technologies.
A. Ayvaz Electric Power Systems Research 237 (2024) 111017

distribution networks by reducing relay operating times and eliminating

maloperation and miscoordination problems. The proposed approach is
based on assigning different setting groups to each DOCR for the peak
and off-peak demand periods of the day and determining maximum load
currents using actual load demand data. With this strategy, it is possible
to choose lower relay pickup settings, especially for the off-peak loading
hours, and thereby achieve lower relay operating times. Thanks to
recent developments in relay technology, different setting groups for
each numerical DOCR can be specified, although in limited numbers
[13]. Furthermore, these new technological DOCRs can be easily
controlled and programmed [15]. The advantages of numerical DOCRs
have been considered in the methodology of several recent studies. In
[16], an adaptive coordination approach using setting groups is pro­
posed to address coordination problems caused by changes in network
configuration. Reference [17] proposes using the k-means clustering
algorithm to classify various network topologies into clusters consid­
ering a restricted number of available setting groups for DOCRs. Chal­
lenges in protecting networked microgrids due to changes in network
conditions and hence fault currents are addressed in [18] by assigning
Fig. 1. Time-current characteristics obtained for an example case.
four setting groups to each DOCR. In these studies, determination of
network state via communication link/s is required to select a suitable
lines accurately to avoid improper determination of pickup current setting group and assign it to relays on an online basis. Unlike these
settings and hence maloperation of DOCRs [9]. In many studies, studies, this paper proposes a communication-less strategy based on
maximum load currents are obtained under the assumption that load programming numerical DOCRs with different setting groups for the
demand is at full capacity and DGs are operated at a fixed capacity. peak and off-peak loading periods of the day.
However, due to the variable nature of system load level and generation On the other hand, another important factor in optimal coordination
output of renewable-based DGs, load currents at transmission lines are studies is the optimisation method used to determine the optimal relay
more changeable. Therefore, a comprehensive consideration of loading settings that give the minimum total relay operating time. Optimisation
and generation variations is required for a reliable protection scheme. methods according to the related literature can be classified as mathe­
According to recent literature, a few optimal coordination studies have matical programming methods, metaheuristic methods, and hybrid
given specific attention to this issue. In [10], an online protection co­ methods. The study in [17] is based on linear programming (LP). In
ordination approach that considers frequent changes in load current and [18], an interior-point optimization (IPOPT) algorithm, which is a
short-circuit current levels of transmission lines as well as network to­ nonlinear programming (NLP)-based method, is used. Although math­
pology changes due to DG outages is proposed. In [11], a protection ematical programming methods provide an advantage in terms of
framework based on active communication links between DOCRs is computation speed, using these methods carries a high risk in terms of
proposed to improve microgrid reliability and security while minimising obtaining local optimum solutions due to their high dependency on
energy not supplied. However, these methods require communication initial values of decision variables. The metaheuristics, with their sto­
infrastructure and/or an online computation process. This implies the chastic nature, are more capable of discovering promising locations in
need for expensive hardware/software solutions [12]. In addition, search space and are more likely to converge to the global optimal so­
especially for communication-assisted approaches, reliability issues lution [19]. Some of the metaheuristics used in this field are gazelle
arise related to communication links that must be secured against optimization algorithm (GOA) [2], water cycle algorithm (WCA) [11],
cyberattacks [13]. harmony search algorithm (HSA) [14], particle swarm optimization
On the other hand, in protection coordination applications, the (PSO) [20], genetic algorithm (GA) [21], improved moth-flame opti­
pickup current setting of a relay must be selected higher than the mization (IMFO) [22], harris hawks optimization (HHO) [23], imperi­
maximum load current passing through the corresponding relay. alistic competition algorithm (ICA) [24], whale optimization algorithm
Therefore, high load current magnitudes require selecting high pickup (WOA) [25], variable neighbourhood search (VNS) [26], improved
current settings for relays to prevent maloperation. Such high settings firefly algorithm (IFA) [27], and elite marine predators algorithm
reduce the sensitivity of protection. In addition, the effectiveness of (EMPA) [28]. Hybrid algorithms are another class of optimisation al­
protection coordination becomes more questionable as higher pickup gorithms, such as PSO-integer LP (PSO-ILP) [16] and GA-NLP [29].
currents are selected [14]. For an example case, Fig. 1 shows the From these studies and as seen in [30], hybrid algorithms demonstrate
time-current characteristic curves of relays R1 and R2. In this example, better performance than individual algorithms. However, a hybrid al­
R1/R2 is a primary/backup relay pair. Fig. 1 shows that the operating gorithm’s global search ability primarily depends on the metaheuristic
times of the relays increase with the decrease in the multiples of pickup algorithm that it contains. Therefore, a well-established individual al­
current settings, as indicated by the red arrows, due to the increase in gorithm with high exploration ability can yield better results than
pickup current settings assuming that fault currents remained constant. hybrid algorithms, as seen in [28] and [31]. Therefore, to obtain a more
To compensate for these long operating times, reducing the TDS of R2 by powerful global search ability in applied hybridising strategy, the
changing TDS1 to TDS2 or TDS3 could be a solution. However, this time, PSOGSA and GA-PSO-LP algorithms are developed in [4] and [32],
the problem of miscoordination arises due to the violation of CTI. In this respectively. This is a meaningful strategy to combine powerful features
case, to overcome this problem, the TDS of R1 can be decreased. This of two individual metaheuristic algorithms. However, important de­
process can be expanded for all relay pairs; however, at the final stage, ficiencies of such hybrid algorithms are the complexity of parameter
avoiding CTI violations can be impossible because a primary relay for a settings and the high computational burden of optimisation. On the
fault point can be in backup protection for other fault point/s. Conse­ other hand, although GA and PSO are pioneers and widely used algo­
quently, limiting TDS to low values can result in an infeasible optimi­ rithms in the metaheuristic area, many new metaheuristics have been
sation solution for optimal DOCR coordination. developed that perform better and faster than them in the past decade.
In this paper, a novel DOCR coordination approach is proposed to In addition, in this area, a new algorithm is generally verified by
provide an effective and reliable protection scheme for IIDG-dominated comparing it with its recent competitors. Therefore, it is important to try

2
 A. Ayvaz Electric Power Systems Research 237 (2024) 111017

Table 1
Summary of literature review in the protection of systems equipped with DOCRs and DGs.
Ref. Year Optimal Communication-less Completely offline Network Consideration of the variable nature of Optimisation
coordination approach computation procedure topology changes DG output and system load level method

[21] 2018 √ √ √ × × GA
[22] 2020 √ √ √ × × IMFO
[16] 2020 √ × √ √ × P SO-ILP
[4] 2020 √ √ √ √ × P SOGSA
[24] 2021 √ √ √ × × ICA
[26] 2021 √ √ √ × × VNS
[1] 2021 × × √ √ √ -
[3] 2021 √ × √ √ × GA
[5] 2021 √ √ √ × × GA, PSO
[14] 2021 √ × √ √ × G A,HSA,WCA
[18] 2022 √ × √ √ × I POPT
[9] 2022 × √ × √ × -
[10] 2022 √ × × √ √ FUZZY+GA
[25] 2022 √ √ √ √ × W OA
[32] 2022 √ √ √ √ × G A-PSO-LP
[2] 2023 √ √ √ × × GOA
[29] 2023 √ × × √ × G A-NLP
[11] 2023 √ × √ √ √ WCA
[28] 2023 √ √ √ × × EMPA
[13] 2023 √ × √ √ × GA
[12] 2024 √ √ × √ × G urobi/AMPL
[23] 2024 √ √ √ × × HHO
Proposed √ √ √ √ √ EWSO
method

recent optimisation techniques in the protection coordination field discussions are introduced in Section 5, and finally, Section 6 provides
because they have possibly higher potential than their predecessors. the conclusions.
This paper proposes a new variant of white shark optimizer, called
enhanced white shark optimizer (EWSO), to solve the optimisation 2. Problem formulation
problem considered in this study. The WSO is a recent metaheuristic
algorithm inspired by great white sharks’ foraging behaviour [33]. The This section presents the mathematical formulation of the optimi­
WSO algorithm, with its advantages of simplicity and low computational sation problem that is solved to find the optimal DOCR settings. DOCR
burden, has been used to overcome several complex engineering prob­ operates when the current flowing in front of it exceeds its pickup cur­
lems, such as energy management [34] and optimal power flow [35]. rent setting. It is desirable for DOCRs to operate as quickly as possible to
However, the WSO algorithm suffers from premature convergence and isolate faults to prevent high stress on power system equipment and thus
can easily fall into local optima due to its inadequate population di­ damage to them. However, the primary/backup protection mechanism
versity in the later stages of iterations. Therefore, in this paper, a new has a restricted impact on this goal. This mechanism is essential for
enhancement strategy that addresses this shortcoming by modifying the overcurrent protection in terms of selectivity to avoid unnecessary
position update rule of fish school behaviour, which is dominant in the interruption of electricity supply outside the fault zone. In this context,
later stages of evolution, is applied to the basic WSO algorithm. the optimal DOCR coordination topic is studied to meet these desires
The other contributions of this paper can be summarised as follows: and requirements. By solving this problem, it is tried to obtain a fast,
selective, and reliable protection coordination scheme.
• The k-means clustering technique is used to classify peak and off-
peak hours on a daily load profile extracted from historical load
2.1. Objective function
demand data.
• To obtain maximum load currents and thus minimum pickup current
The objective of this solution is to minimise the relay operating times
settings for DOCRs, a Monte Carlo simulation is performed by
by keeping the constraints on the coordination time interval and relay
considering variations in system load demand, uncertainty in power
characteristics within their limits. The mathematical equation of the
generation due to IIDGs, and changes in network topology due to
objective function is as follows:
single and multiple IIDG outages.
• The proposed coordination approach is validated on the distribution NRP
∑ ( )
section of the IEEE 14-bus and IEEE 30-bus systems, which are OF = Tp,k,l,h + Tb,m,l,h ∀(k, m) ∈ ΩRP (1)
h=1
widely used test systems in this research area.
• The performance of the proposed EWSO algorithm is compared with where, NRP is the total number of relay pairs. ΩRP represents the set of
the basic WSO algorithm and other two recent metaheuristics primary/backup relay pairs of the DOCRs. Tp,k,l,h and Tb,m,l,h express the
including hunter-prey optimization (HPO) [36] and golden jackal operating times of the primary relay k and backup relay m for a fault at
optimization (GJO) [37] algorithms. the location l, respectively. The relay operating time according to IEC
standards is calculated as follows:
Based on the literature review above, Table 1 presents a comparison
of the proposed approach and other methods intended for systems α
Tk,l = TDSk × ( / )β (2)
equipped with DOCRs and DGs. Ifk,l PSk − 1
The rest of the paper is organised as follows. The formulation of the
optimisation problem is given in Section 2. Section 3 describes the where, TDSk and PSk are the time dial setting and pickup setting of the kth
methodology of the proposed coordination approach. Section 4 presents DOCR, respectively. Ifk,l is the fault current observed by the DOCR k in
the mathematical model of the EWSO algorithm. The results and case of a fault occurrence at the location l. α and β are constant co

3
­
A. Ayvaz Electric Power Systems Research 237 (2024) 111017

efficients specific to the relay characteristic. to classify the peak and off-peak hours on the obtained daily load profile.
Here, the k-means clustering method helps to make better discrimina­
2.2. Problem constraints tion between the load demand hours of these two classes (i.e. peak or
off-peak). This is important because modelling load demand variation to
The objective function given by Eq. (1) is subject to a set of con­ find maximum load currents directly depends on this classification.
straints detailed below. After this step, the procedure continues in parallel for the two
separate periods of the day (peak and off-peak loading periods).
2.2.1. Coordination selectivity constraint
To prevent maloperation, the minimum CTI must be maintained for Step 3. To find the maximum load currents passing through the relays,
all primary/backup relay pairs. The expression of this criterion can be a Monte Carlo simulation is performed, considering load variation, un­
given as: certainty in power generation due to IIDGs, and IIDG outages. Here, the
upper bound of load variation is specific to each period (peak or off-
CTI ≤ Tb,m,l,h − Tp,k,l,h ∀(k, m) ∈ ΩRP (3) peak) and can be obtained from the daily load profile. On the other
hand, it is difficult to know when an IIDG will operate at its rated ca­
2.2.2. Relay characteristics constraints pacity. Therefore, IIDG output power is assumed to reach its rated value
These constraints define the limits of relays’ TDS, PS, and operating during the peak and off-peak periods.
time. The bounds on each relay’s TDS are provided by the relay manu­
facturer and can be expressed as: Step 4. The lower bound of PS to be used in the optimisation process is
min max specified for each relay.
TDS ≤ TDS ≤ TDS (4)

where, TDSmin and TDSmax are the minimum and maximum bounds of Step 5. The optimisation problem formulated in Section 2 is solved.
TDS, respectively.
The bounds on PS can be given as [38]: Step 6. The optimal relay settings to be used in the protection coor­
( ) ( ) dination for the peak and off-peak loading periods are obtained.
max PSmin , Iload
max
k
≤ PSk ≤ min PSmax , Ifault
min
k
k ∈ ΩR (5) The following subsections detail the load profile extraction step and
define relay settings.
where, ΩR is the set of DOCRs in the system. PSmin and PSmax are the
max
minimum and maximum available tap settings, respectively. Iload k
and 3.1. Load profile extraction
min
Ifaultk
represent the maximum load current and minimum fault current
max
passing in front of the relay k, respectively. Here, the value of Iload is In most DOCR coordination studies, maximum load currents are
k

identified through the proposed methodology in this paper. determined by performing a power flow solution under the system’s full
In Eq. (1), relay operating times are also constrained as follows: loading condition. After the maximum load current passing through a
relay is determined, the lower bound of the relay’s PS used in the so­
T min ≤ T ≤ Tmax (6) lution process of the optimal DOCR coordination problem is set to the
value obtained by multiplying this current by a constant threshold value
where, Tmin and T max are the minimum and maximum bounds of the (i.e. 1.25 [40]), if the obtained value is not lower than a predetermined
relay operating time, respectively. Here, Tmin depends on the relay reference limit value (i.e. 0.5 [41]). This technique is not practical
manufacturer. On the other hand, T max depends on system stability re­ because the system’s loading condition may never approach full ca­
quirements, such as critical clearing time. pacity and may often be well below it. Therefore, multiplying the ob­
tained maximum load current by the threshold value results in
3. The proposed coordination approach unnecessarily high PS values being assigned to DOCRs. On the other
hand, if the maximum loading condition is derived from the system’s
It is quite difficult to predict when a fault will occur. In case of a fault, historical load demand data, it may correspond to an extreme loading
it must be cleared as quickly as possible to prevent high stress on power condition, which also causes the same problem.
system equipment. One of the most practical ways to achieve this is to Unlike the methods presented in the literature, in this study, the
shorten relay operating times while avoiding coordination violations. maximum loading level for each hour of the day is calculated according
However, load currents in lines can reach high values during the day; to Eq. (7). Here, hourly extraction of maximum loading levels is
therefore, assigning high pickup current settings to DOCRs is required to necessary to obtain the daily load profile that is to be used in discrimi­
prevent maloperation. Due to the negativeness of high PS, long relay nating the peak and off-peak periods of the day.
operating times for DOCRs can be obtained. Therefore, as proposed in { [ ] [ ]
this study, dividing the day into two periods based on the loading con­ 1.25 × Lavg + Lstd , 1.25 × Lavg + Lstd > Lmax
LPt = t t t t t
(7)
ditions of the lines and assigning different setting groups to each DOCR 1.01 × Lt otherwise
max

for these periods can be one of the best ways to enable fast relay oper­
ations and thus reduce the stress on power system equipment, for a where, LPt expresses the obtained load profile data corresponding to the
certain part of the day. In addition, this paper suggests a realistic avg
hour t of the day, Lt , Lstd max
t , and Lt represents the average, standard
strategy for obtaining maximum load currents. This strategy helps deviation, and maximum values of the historical load data for the hour t
reduce operating times also for peak hours and provides a reliable co­ of the day, respectively.
ordination scheme by eliminating DOCR maloperation. The imple­ In Eq. (7), using historical load data provides a realistic load profile
mentation steps of the proposed coordination approach are given as extraction, and using different multiplying factors (i.e. 1.25 and 1.01)
follows: prevents unnecessarily high pickup settings due to extreme loading
[ avg ]
Step 1. First, using a set of historical load demand data, a load profile conditions (i.e. 1.25 × Lt + Lstd t ≤ Lmax
t ).
that includes the hourly maximum loading levels observed during a
season or year-round period is extracted. 3.2. Defining relay settings

Step 2. Following Step 1, the k-means clustering technique [39] is used In this study, all the DOCRs in the system are considered numerical

4
A. Ayvaz Electric Power Systems Research 237 (2024) 111017

Table 2 approach.
IEC standard relay characteristics and their coefficient values.
Characteristic ID (CID) Characteristic type α β 4. Enhanced white shark optimizer (EWSO)
1 Normal inverse 0.14 0.02
2 Very inverse 13.5 1 The mathematical background of the WSO algorithm and the
3 Extremely inverse 80 2 enhancement strategy used to develop the EWSO are given in this
section.

4.1. The basic WSO algorithm

The white shark optimizer, which was introduced by Braik et al. in

2022 [33], mimics the hunting and navigating behaviours of great white
sharks. The great white shark has extraordinary senses of smell and
hearing to detect prey in the ocean’s depths, and its powerful ability to
search for prey is mathematically formulated in the WSO algorithm. The
WSO algorithm is established on three movement activities of white
sharks in identifying prey (or the food source): 1) the movement activity
based on the hesitation of waves caused due to the movement of prey, 2)
the random movement for searching the optimal prey location, and 3)
the movement of following the best white shark.
Following the order of movement activities defined above, in the first
one, once a white shark senses waves created due to prey movement, it
starts to swim towards to prey in an undulating manner. The velocity of
this motion is expressed as follows:
[ ( ) ( )]
vik+1 = μ⋅ vik + c1 × p1 × wgbk − wik + c2 × p2 × wibestk − wik , i ∈ [1, N]
(9)

where vik+1 and vik are the velocity vectors of the white shark i in the next
and current iterations, respectively. c1 and c2 are random numbers
generated uniformly within the range of [0, 1]. wgbk represents the
current global best position. wik is the current position of the white shark
i, and wibestk is its best position obtained so far in the iteration k. N rep­
resents the population number. The algorithm control parameters, μ,
p1 and p2 , in Eq. (9) are respectively calculated as follows:
2
μ=⃒ √⃒̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅, τ = 4.125 (10)
⃒2 − τ − τ2 − 4τ⃒
( )
p1 = pmax + (pmax − pmin ) × exp − (4k/K)2 (11)

( )
p2 = pmin + (pmax − pmin ) × exp − (4k/K)2 (12)

Fig. 2. Flowchart of proposed coordination approach. where pmax and pmin are, respectively, the subordinate and initial ve­
locities of the great white shark’s movement, and their values are 1.5
and 0.5, respectively. K represents the maximum number of iterations.
relays, and the relay settings, including the time dial setting, pickup
The second movement behaviour of the great white shark is its
current setting, and time-current characteristic are optimally deter­
random navigation in search of prey. Once the white shark hears the
mined for each DOCR by solving the optimisation problem given by Eqs.
movement of prey or smells its scent, it moves toward the prey. How­
(1)–(6). Among these settings, the time-current characteristic is a
ever, the prey may leave the location where it leaves its scent. In this
discrete decision variable whereas the others are continuous. Three
case, the white shark continues to smell the prey and searches for it in
types of relay characteristics according to IEC standards are considered:
random locations. The position update rule of this behaviour is formu­
normal inverse, very inverse, and extremely inverse. Table 2 lists the
lated as:
coefficients used in Eq. (2) for each of these characteristics [42]. One of
{
them is optimally assigned to each relay setting group. wik ⋅¬ ⊕ wo + ub⋅a + lb⋅b; r1 < mv
For each DOCR in the distribution system, two setting groups (for wk+1 =
i
/ (13)
wik + vik f; otherwise
peak and off-peak periods) and a total of six setting parameters are
optimally obtained via the proposed coordination methodology. The
wo = ⊕(a, b) (14)
optimal relay setting groups can be expressed by the following matrices:
⎡ ⎤ ⎡ ⎤ ( )
a = sgn wik − ub (15)
TDS1k TDS2k
⎢ ⎥ ⎢ ⎥
SG1, peak
=⎢⎣ PSk ⎦, SGk
1 ⎥ 2, off− peak
⎣ PSk ⎦, k ∈ ΩR
=⎢ 2 ⎥
(8) ( )
k b = sgn wik − lb (16)
CID1k CID2k

Fig. 2 presents the overall flowchart of the proposed coordination

5
A. Ayvaz Electric Power Systems Research 237 (2024) 111017

fmax − fmin
f = fmin + (17)
fmax + fmin

where the symbols, ¬ and ⊕, denote the bitwise not and xor operations,
respectively. ub and lb are the upper and lower bounds of the decision
variables, respectively. fmax and fmin represent, respectively, the
maximum and minimum limits of the frequency of the updating shark
movement, and they are equal to 0.75 and 0.07, respectively. r1 is a
random number between 0 and 1. mv is an algorithm parameter that
controls the balance between the exploration and exploitation phases.
The formulation of mv is given as follows:
1
mv = (18)
(a0 + exp((K/2 − k)/a1 )

where a0 and a1 are positive constants, which equal 6.25 and 100,
respectively.
The final movement behaviour of white sharks is their tendency to
track the best-performing white shark closest to the prey. In line with
this fish school behaviour, all the white sharks, except for the two best-
Fig. 3. Historical load demand data handled in this study.
performing ones, update their positions according to Eq. (19).

wik + wk+1
i
the position vector of a randomly selected individual (i.e. rth individual)
ʹ
wik+1 = (19)
2 × rand from the swarm. β is a deciding coefficient that determines the indi­
vidual behaviour of tracking or moving away from a random individual.
→
wk+1 = wgbk + r2 × D w × sgn(r3 − 0.5) r4 < s (20)
ʹi
β is defined in the optimisation process according to Eq. (24).
{
→ ⃒ ( )⃒ − 1; ftki < ftkr
D w = ⃒r5 × wgbk − wik ⃒ (21) β= (24)
1; otherwise
s = |1 − exp (− a2 × k / K)| (22)
where ftki and ftkr represent the fitness values of the ith and rth individuals
where r2 , r3 , r4 , and r5 are random numbers between [0, 1]. a2 equals in the iteration k, respectively.
0.0005. According to the new position update rule, each individual obtains
more information about the prey location than it does based on the old
4.2. The developed EWSO algorithm rule. This increased information sharing between swarm members also
The significant deficiency of the basic WSO algorithm is insufficient increases the population diversity by providing alternative promising
population diversity in the later stages of evolution, which can cause search directions to individuals in addition to that of the best individual.
premature convergence speed and cause the algorithm to fall into local
optima. According to the mathematical model of the basic WSO, it can 5. Results and discussion
be seen that the fish school behaviour of white sharks (i.e. Eqs. (19)–
(22)) is possibly dominant in the later stages of evolution in accordance In this section, the results obtained by implementing the proposed
with the increase of s over iterations. Although, in the mathematical coordination approach on widely used test systems are presented, along
model of this behaviour, the usage of information of the best-performing with a detailed analysis of the performance and applicability of the
white shark enhances the algorithm convergence performance, still proposed strategy. The validation of the proposed method is performed
some issues remain that need to be addressed to achieve a more on the distribution sections of the IEEE 14-bus and IEEE 30-bus test
performance-optimising evolution. One of these problems concerns the systems. Three-phase-to-ground solid faults at the midpoint of lines are
use of sgn(r3 − 0.5). It equals -1 when r3 is lower than 0.5. In this case, used in short circuit analysis, and fault currents are computed using
the corresponding individual shows a behaviour of moving away from ETAP software. Each IIDG is assumed to have a rated capacity of 5 MW.
the best individual. This behaviour perhaps contributes to the diversity To model system load demand, historical load demand data taken
of solutions; however, it is not a meaningful search strategy funda­ from the website of the National Load Dispatch Centre Information
mentally and can burden the algorithm unnecessarily. The other prob­ System (YTBS) of Türkiye for 2022 is used [43]. Fig. 3 presents the data
lem is the swarm member movements’ high dependency on the best plot. It can be seen that variations in load demand exhibit different
white shark’s position. This undesired level of dependency can lead to a seasonal characteristics. For example, load demand reached high levels
decrease in population diversity and therefore cause the local optima during Jul-Sep whereas it remained at considerably lower levels during
problem. Oct-Dec. These differences significantly affect the proposed DOCR co­
To cope with these deficiencies of the basic WSO algorithm, in this ordination strategy’s effectiveness because maximum load currents in
study, an enhancement strategy is applied to the mathematical model of lines can vary depending on whether the selected time period corre­
white sharks’ fish school behaviour. Therefore, the position update rule sponds to a season or year, or what season of the year it is. Therefore, the
in Eq. (19) is modified as follows: proposed approach is implemented and investigated under two cases
( ( ) ( )) described as follows:
wik+1 = 0.5 × rand × wgbk − wik + rand × wsbk − wik + β × rand
( )
× wrk − wik r4 • Case 1: Historical load demand data for the entire year are
<s considered.
(23) • Case 2: Historical load demand data during the Oct-Dec season are
considered.
where wsbk is the position vector of the second-best white shark, and wrk is
To obtain the maximum load currents passing through the relays, a

6
A. Ayvaz Electric Power Systems Research 237 (2024) 111017

Fig. 4. Single-line diagram of the modified IEEE 14-bus distribution system.

Fig. 5. The load profiles with the peak and off-peak loading hours obtained for the IEEE 14-bus distribution system considering (a) Case 1 and (b) Case 2.

Monte Carlo-based power flow solution is performed in MATLAB for TDSmax , T min , and Tmax are set as 0.05, 1.1, 0.05, and 2, respectively. The
each day period of each case using IIDG information (i.e. capacity and CTI is selected as 0.2 s. All the simulation experiments are performed
location) and load profile data extracted based on the proposed pro­ using MATLAB on a personal computer with a 2.6 GHz processor speed
cedure in Section 3.1. At this point, it should be noted that the raw and 12 GB of random access memory. For the EWSO algorithm, the
historical data presented in Fig. 3 needs to be rearranged in accordance recommended parameter settings in [33] were found to be suitable
with the systems’ total load demands before passing the proposed pro­ based on the trial-and-error procedure. Furthermore, to avoid the
cedure steps defined in Section 3.1. For this purpose, the data are first haphazard situation caused by the high degree of randomness of meta­
scaled by assuming that the maximum load demand observed during the heuristics, the results were obtained over 20 independent runs. The
whole year corresponds to 80 % of the studied system’s total load de­ population size and maximum number of iterations are selected as 500
mand. In other words, the demand factor is selected as 0.8. and 2000, respectively.
The lower and upper bounds of the optimisation variables, TDSmin ,

Table 3
Lower bounds on the pickup settings of DOCRs obtained for the IEEE 14-bus distribution system.
DOCR Case 1 Case 2 DOCR Case 1 Case 2

Peak Off-peak Peak Off-peak Peak Off-peak Peak Off-peak

R1 0.55 0.50 0.50 0.50 R9 0.90 0.75 0.78 0.66

R2 0.83 0.67 0.70 0.57 R10 1.82 1.52 1.56 1.34
R3 0.51 0.50 0.50 0.50 R11 0.97 0.86 0.88 0.80
R4 1.02 0.88 0.91 0.81 R12 0.61 0.54 0.55 0.50
R5 0.50 0.50 0.50 0.50 R13 1.59 1.29 1.32 1.08
R6 0.50 0.50 0.50 0.50 R14 0.50 0.50 0.50 0.50
R7 1.62 1.42 1.45 1.29 R15 0.57 0.53 0.53 0.50
R8 0.54 0.50 0.50 0.50 R16 1.11 0.89 0.93 0.76

7
A. Ayvaz Electric Power Systems Research 237 (2024) 111017

Table 4
Optimal relay setting groups obtained for the IEEE 14-bus distribution system.
DOCR Case 1 Case 2
1, peak 2, off− peak
Peak demand period, SG Off-peak demand period, SG Peak demand period, SG1, peak Off-peak demand period, SG2, off− peak

TDS PS CID TDS PS CID TDS PS CID TDS PS CID

R1 0.908 0.593 3 0.464 0.708 3 0.409 0.765 3 0.061 1.813 3

R2 0.229 0.830 3 0.297 0.731 3 0.068 1.489 3 0.464 0.570 3
R3 0.222 1.548 3 0.218 1.619 3 0.296 1.319 3 0.264 1.374 3
R4 0.077 2.187 3 0.302 1.113 3 0.460 0.912 3 0.058 2.439 3
R5 0.059 1.874 3 0.672 0.554 3 0.084 1.549 3 0.137 1.201 3
R6 0.950 0.574 3 0.280 1.058 3 0.383 0.913 3 0.474 0.796 3
R7 0.050 1.982 3 0.050 1.820 3 0.080 1.451 3 0.098 1.290 3
R8 1.056 0.627 3 0.161 1.531 3 0.206 1.385 3 1.020 0.600 3
R9 0.428 0.900 3 0.062 2.200 3 0.557 0.780 3 0.050 2.374 3
R10 0.067 1.820 3 0.082 1.610 3 0.091 1.560 3 0.079 1.613 3
R11 0.411 1.185 3 0.632 0.860 3 0.705 0.883 3 0.650 0.800 3
R12 0.050 2.248 3 0.406 0.807 3 0.331 0.899 3 0.256 0.999 3
R13 0.050 1.590 1 0.051 1.290 2 0.050 1.320 1 0.050 1.080 2
R14 0.270 0.500 2 0.483 0.500 3 0.160 0.601 2 0.448 0.512 3
R15 0.167 0.570 2 0.186 0.530 2 0.185 0.530 2 0.185 0.511 2
R16 0.056 1.112 2 0.141 0.890 3 0.050 0.930 2 0.127 0.760 3
OF 9.389 s 8.459 s 8.568 s 7.692 s

Fig. 6. The time-current characteristic curves of R13 and R11 for the obtained setting groups in Case 2.

5.1. IEEE 14-bus distribution system original system are available in [44]. The system’s total load demand is
The proposed approach is first tested on a modified version of the 87.3 MW and 44.1 MVAr. In this study, it is assumed that three IIDGs are
IEEE 14-bus distribution system, as Fig. 4 shows. Detailed data of the installed at bus nos. 2, 5, and 6, as Fig. 4 shows. The utility grid is

8
A. Ayvaz Electric Power Systems Research 237 (2024) 111017

Fig. 7. Single-line diagram of the modified IEEE 30-bus distribution system.

connected to bus nos. 1 and 4 with an 800 MVA short circuit capacity in flowing through relays R13 and R11 for random system operating con­
total (each 400 MVA). The system is protected with 16 DOCRs at the ditions selected from the peak and off-peak periods, in the case of a
beginning and end of each line and contains 22 primary/backup relay three-phase-to-ground fault occurring at the midpoint of the line be­
pairs [21]. The current transformer ratios for the relays (R2, R8), (R1, tween bus nos. 4 and 6. For this fault condition, R13 is the primary relay
R3, R5, R6, R9, R14), (R12, R13), and (R4, R7, R10, R11, R15, R16) are and R11 serves as its backup. Compared to the off-peak period settings,
selected as 300:1, 200:1, 160:1, and 100:1, respectively. the peak period settings cause a 0.086 s increase in the operating time of
Following the proposed methodology, the system load profiles for R13 (from 0.091 to 0.177 s). As a result, due to the CTI constraint, the
Case 1 and Case 2, including the peak and off-peak periods, are obtained operating time of R11 also increases by about 0.089 s in backup pro­
as shown in Fig. 5. The lower bounds on the pickup settings of the tection (from 0.291 to 0.38 s). It should be noted that the CTI for the
DOCRs are determined, as presented in Table 3. primary/backup relay pair R13/R11 appears to be compatible with the
Table 4 presents the results obtained by solving the coordination calculations based on the optimisation outcomes, where it equals 0.203
problem formulated in Section 2 using the EWSO algorithm. It can be and 0.2 s for the peak and off-peak period settings, respectively. From
seen that, in each case, the total operating time of relays (OF) obtained this perspective, the simulation results verify the accuracy of the optimal
for the off-peak period is lower than that obtained for the peak period. relay settings obtained with the proposed approach.
Further, for the seasonal case (Case 2), the OF values are lower than The observed changes in relay operating times are directly related to
those in Case 1. The results confirm the loading conditions’ significant the difference in the pickup settings of R13 obtained for the peak and off-
impact on relay coordination and show that the lower the pickup set­ peak periods. The optimal PS settings for R13 in Case 2 are found to be
tings assigned to the DOCRs, the shorter the relay operating times. As 1.32 and 1.08 for the peak and off-peak periods, respectively (see
proof of this, the optimal PSs for many relays are obtained at or near the Table 4). These settings correspond to the lower PS bounds determined
minimum values given in Table 3. For example, considering the peak for R13 in Table 3 for each of these periods. The difference is due to the
loading period in Case 1, the optimal PSs for relays R2, R9, R10, R13, higher loading condition observed during the peak loading period of the
R14, and R15 are at the lower bounds, and the setting of R16 is very day, which increases the lower bound of the PS settings and prevents
close to its minimum value. A similar situation is observed in 5 relays for relays taking lower PS settings. As previously mentioned, an increase in
the off-peak period in Case 1 and in 7 relays for the peak and off-peak the pickup setting decreases the value of Ifk,l /PSk thereby increasing the
periods in Case 2. relays’ operating time according to Eq. (2). To mitigate this, a smaller
On the other hand, it is observed that the extremely inverse char­ TDS or a different relay characteristic can be selected. It is noted that the
acteristic type is dominant in the optimal relay characteristics. This TDS of R13 is already at its minimum value (i.e., 0.05) for both periods.
characteristic provides advantages in obtaining low relay operating Changing the characteristic type of R13 from normal inverse to very
times when fault currents passing through the overcurrent relays are inverse for the peak period would slightly reduce the primary operating
high or when selectable pickup currents are low [45]. Table 4 shows that time of R13 but would also lead to a larger increase in the backup
as pickup current settings decrease, the characteristic types of some operating time of R13. Moreover, the optimisation procedure already
relays change from normal inverse to very inverse or extremely inverse. provides the optimal relay settings that yield the minimum operating
Furthermore, to analyse the impact of load currents on the optimal times. Therefore, altering the optimal settings is not a viable solution.
coordination of DOCRs in more detail and to verify the optimisation Consequently, this example and the overall optimisation results indicate
outcomes in a simulation environment, the relay pair R13/R11 is that the pickup settings and, hence the load currents that define the
selected as an example. In Fig. 6, the time-current characteristic curves lower bounds of these settings, play a crucial role in the effectiveness of
of R13 and R11 obtained via ETAP for the optimal relay settings found relay coordination.
by the proposed approach for the peak and off-peak periods in Case 2 are
given. Here, the currents shown in Fig. 6 correspond to the fault currents

9
A. Ayvaz Electric Power Systems Research 237 (2024) 111017

Table 5 Case 1 result in a decrease of about 2.872 s in the OF compared to the

Selected ratios of current transformers for the relays in the IEEE 30-bus distri­ peak period settings. This decrease is 2.866 s in Case 2. Additionally,
bution system. similar to the results obtained for the 14-bus system, smaller OF values
Current 100:1 150:1 200:1 300:1 are observed in the seasonal case (Case 2) than in the annual case (Case
transformer 1) for the 30-bus system.
ratios According to the proposed protection scheme, for a significant
Relays R8, R9, R1, R3, R4, R6, R7, R10, R11, R2, R5, portion of the day—corresponding to 8 hours in Case 1 and 9 hours in
R16, R19, R14, R15, R17, R18, R12, R23, R13, R24, Case 2—, the relays will operate faster in the event of a fault during these
R20, R30 R21, R22, R26, R28, R25, R27, R34, R35
periods. This reduces the stress on power system equipment, decreases
R29, R33, R36, R37, R31, R32
R38, R39 the likelihood of damage to them, and thus promotes to the reliable and
healthy operation of the power system over a longer period.

5.2. IEEE 30-bus distribution system 5.3. Evaluation of the proposed coordination scheme
A modified version of the IEEE 30-bus distribution system is also Table 7 presents the results as 24-hour averages. Here, the results
used to test the performance of the proposed approach for a larger dis­ from the traditional coordination approach are obtained by following
tribution system. Detailed data of the original system are available in
[44]. Fig. 7 shows the single-line diagram of the system. It is assumed
Table 7
that four IIDGs are installed at bus nos. 6, 10, 16, and 17. The system is Comparison of the proposed and traditional coordination approaches based on
protected with 39 DOCRs and contains 56 primary/backup relay pairs 24 h averages.
[46]. The current transformer ratios for the relays are given in Table 5.
Method 14-bus system 30-bus system
The optimal relay settings and OF values obtained by the proposed
approach for the IEEE 30-bus distribution system are tabulated in OF (s) TOTpri (s) OF (s) TOTpri (s)

Table 6. The results indicate that dividing the day into two periods (peak Traditional coordination approach 10.909 2.194 37.224 6.688
and off-peak demand) provides significant benefits in terms of total relay The proposed method (Case 1) 9.079 1.675 30.617 5.345
The proposed method (Case 2) 8.240 1.472 29.239 4.795
operating time, as proposed in this paper. The off-peak period settings in

Table 6
Optimal relay setting groups obtained for the IEEE 30-bus distribution system.
DOCR Case 1 Case 2
1, peak 2, off− peak
Peak demand period, SG Off-peak demand period, SG Peak demand period, SG1, peak Off-peak demand period, SG2, off− peak

TDS PS CID TDS PS CID TDS PS CID TDS PS CID

R1 0.053 2.247 3 0.157 1.407 3 0.663 0.669 3 1.008 0.514 3

R2 0.224 0.815 3 0.078 1.328 3 0.050 1.663 3 0.452 0.553 3
R3 0.134 2.252 3 1.086 0.812 3 1.081 0.816 3 1.099 0.797 3
R4 0.128 0.506 2 0.175 1.083 3 0.458 0.669 3 0.811 0.500 3
R5 0.050 0.818 1 0.057 0.652 3 0.050 0.778 2 0.080 0.553 3
R6 0.430 1.445 3 0.345 1.548 3 0.131 2.500 3 0.119 2.500 3
R7 0.178 0.640 2 0.435 0.524 3 0.385 0.556 3 0.050 1.498 3
R8 0.521 1.207 3 0.456 1.281 3 0.800 0.975 3 1.100 0.826 3
R9 0.981 1.075 3 0.462 1.416 3 0.264 1.858 3 1.100 0.901 3
R10 0.202 0.691 3 0.050 1.343 3 0.064 1.201 3 0.376 0.500 3
R11 0.800 0.943 3 0.306 1.466 3 0.106 2.457 3 0.101 2.500 3
R12 0.804 0.990 3 0.866 0.941 3 1.015 0.877 3 0.116 2.500 3
R13 0.224 1.215 3 0.141 1.510 3 0.710 0.686 3 0.050 2.455 3
R14 0.222 1.076 3 0.676 0.603 3 0.632 0.625 3 0.901 0.522 3
R15 0.096 2.351 3 0.209 1.437 3 0.633 0.840 3 1.100 0.615 3
R16 0.216 1.130 3 0.501 0.732 3 0.984 0.526 3 0.050 2.188 3
R17 0.124 1.079 2 0.565 0.500 3 0.078 1.405 3 0.050 1.606 3
R18 0.118 1.185 2 0.179 0.591 2 0.229 0.549 2 0.214 0.501 2
R19 0.214 2.357 3 0.698 1.222 3 0.682 1.241 3 1.100 0.911 3
R20 1.099 0.916 3 0.442 1.319 3 0.218 1.947 3 0.118 2.500 3
R21 0.055 2.500 2 0.414 0.735 3 0.205 1.034 3 0.050 1.815 3
R22 0.125 2.485 3 1.100 0.818 3 0.511 1.225 3 1.100 0.811 3
R23 0.050 1.102 1 0.092 0.640 2 0.081 0.706 2 0.050 0.800 3
R24 0.050 0.811 1 0.056 0.663 2 0.050 0.723 2 0.055 0.571 3
R25 0.293 0.623 2 0.200 0.848 2 0.112 1.475 2 0.338 0.513 2
R26 0.050 0.956 2 0.091 0.603 2 0.086 0.626 2 0.107 0.521 2
R27 0.786 0.616 3 0.294 0.903 3 0.450 0.730 3 0.050 1.924 3
R28 0.495 1.032 3 0.324 1.197 3 0.362 1.221 3 0.069 2.500 3
R29 0.060 2.342 3 0.989 0.591 3 0.134 1.591 3 1.100 0.550 3
R30 0.147 2.466 3 0.664 1.125 3 0.536 1.250 3 1.100 0.840 3
R31 0.847 0.643 3 0.230 1.068 3 1.024 0.595 3 0.050 2.229 3
R32 0.234 0.644 2 0.340 0.700 3 0.354 0.683 3 0.050 1.728 3
R33 0.125 0.500 2 0.212 0.942 3 0.126 1.210 3 0.740 0.500 3
R34 0.050 0.863 1 0.139 0.514 3 0.064 0.580 1 0.050 0.838 3
R35 0.141 0.680 1 0.136 0.514 1 0.130 0.576 1 0.132 0.500 1
R36 0.100 1.075 3 0.339 0.604 3 0.360 0.658 3 0.476 0.500 3
R37 0.050 1.156 2 0.669 0.520 3 0.367 0.697 3 0.677 0.502 3
R38 0.050 0.500 2 0.050 0.528 2 0.050 0.503 1 0.050 0.500 2
R39 0.050 0.730 2 0.071 0.562 2 0.054 0.714 2 0.080 0.502 2
OF 31.574 s 28.702 s 30.314 s 27.448 s

10
A. Ayvaz Electric Power Systems Research 237 (2024) 111017

Fig. 8. Sensitivity analysis of EWSO algorithm’s parameters.

Steps 3–5 of the proposed methodology outlined in Section 3. However, SDGs can also make significant contributions. In such cases, network
in Step 3, the system load demand is considered to vary up to the full topology changes due to SDG outages may lead to challenges in avoiding
load demand. CTI violations. To address this issue, the proposed method can be
According to the results, the seasonal implementation of the pro­ extended by using fault current limiters to restrict the fault contribution
posed method offers greater effectiveness in relay operations. Compared of SDGs, thereby achieving robustness against topology changes, as in
to the traditional approach, it achieves a reduction of approximately [4].
24.5 % and 21.5 % in the total operating time of relay pairs (OF), and On the other hand, the proposed protection strategy relies on nu­
about 32.9 % and 28.3 % in the total operating time of primary relays merical DOCRs and their intraday setting adjustments. From this
(TOTpri) for the 14-bus and 30-bus systems, respectively. This result is perspective, the proposed method may be considered costly, particularly
notable, especially considering the significant efforts in the literature to given the prevalence of electromechanical relays in distribution systems.
reduce relay operating times by even 5-10 % [22,47]. Thus, these results However, recent advancements in relay technology demonstrate that
validate the superiority of the proposed method in terms of reliable and numerical relays have significant potential to become a fundamental
effective DOCR coordination. Additionally, the proposed coordination component of future protection systems due to their programmability
scheme is also effective for large-scale power systems. and ease of control [48]. Indeed, numerical relays are extensively
It is observed that implementing the proposed coordination scheme included in recent studies [49]. However, even if the intraday adjust­
results in smaller relay operating times compared to those achieved with ments to the relay settings are not applied, the proposed approach still
the traditional approach. Therefore, the proposed method significantly delivers effective results. For instance, using the peak period settings
contributes to improving power systems’ stability, particularly by throughout the entire day would not lead to any maloperation because
reducing primary relays’ operating times. Additionally, considering a the pickup current settings for this period of day are higher than those
seasonal period can further enhance the effectiveness of the proposed obtained for the off-peak period. Even in such a scenario, considering
approach. In this context, it should be noted that the proposed approach Case 2, the proposed strategy yields approximately 21.5 % and 18.6 %
not only improves relay operations by reducing operating times but also reductions in the total operating times of relay pairs for the IEEE 14-bus
offers a robust coordination plan by preventing relay maloperation and and IEEE 30-bus distribution systems, respectively, compared to the
miscoordination due to fluctuations in power generation from IIDGs, traditional approach.
IIDG outages, and variations in system load demand. Another advantage of the proposed approach is that it relies on an
The proposed approach is implemented on two different-scale power offline procedure for determining relay settings. Real-time calculations
systems containing IIDGs, and it is observed that the approach is effec­ of load and fault currents can incur high computational costs, which
tive for both systems. It should be noted that the proposed method is also may lead to relay maloperation due to potential delays in this compu­
applicable to distribution systems equipped with synchronous-type DGs tational process. Additionally, high communication traffic and faulty
(SDGs) or a combination of SDGs and IIDGs. However, while the external data transfer in such traffic for online procedures can result in incorrect
grid is the major contributor to fault currents in grid-connected systems, relay setting determination. Therefore, it can be concluded that offline
procedures are one step ahead of online procedures in terms of
simplicity, applicability, and reliability.
Table 8
Statistical optimisation results obtained for different algorithms under the
5.4. Performance analysis of the EWSO algorithm
traditional coordination scheme.
In this study, the EWSO algorithm was developed and utilized to
Method Min Max Avg Std Average CPU solve the coordination problem according to both the proposed and
time (s)
traditional coordination approaches. In all simulation experiments, the
14-bus HPO 11.718 13.468 12.220 0.676 76.901 control parameters of the EWSO algorithm were kept at the same values.
system GJO 11.444 12.773 12.112 0.452 48.031
The settings were established through a trial-and-error process. Fig. 8
WSO 11.040 12.527 11.500 0.532 40.865
EWSO 10.909 11.474 11.089 0.179 41.187 presents the sensitivity analysis results for the parameters used to solve
30-bus HPO 58.273 69.142 62.132 3.908 99.201 the traditional coordination problem for the IEEE-14 bus system. For
system GJO 50.471 61.220 57.089 3.702 62.772 each parameter, the settings at the center of the test points correspond to
WSO 43.387 49.484 46.655 2.223 55.108 the recommended settings given in [33], except for N and K. The
EWSO 37.224 43.054 39.898 1.738 55.323

11
A. Ayvaz Electric Power Systems Research 237 (2024) 111017

Fig. 9. The average convergence curves of the algorithms for the traditional coordination solution on (a) the IEEE 14-bus and (b) IEEE 30-bus distribution systems.

Table 9
The best optimisation results obtained using different methods for the proposed coordination approach.
Test system Method Case 1 / peak period Case 1 / off-peak period Case 2 / peak period Case 2 / off-peak period

14-bus system HPO 10.285 9.142 9.401 8.634

GJO 10.101 9.299 9.470 8.719
WSO 9.448 8.707 8.743 7.861
EWSO 9.389 8.459 8.568 7.692
30-bus system HPO 52.606 45.721 46.010 41.367
GJO 36.994 33.027 34.232 31.542
WSO 34.881 29.879 32.510 28.762
EWSO 31.574 28.702 30.314 27.448

algorithm was executed 20 times independently for each test point 6. Conclusion
shown in Fig. 8. Thus, Fig. 8 shows the best values obtained over 20 runs.
According to the sensitivity analysis, no improvement in the optimisa­ This paper proposes a new DOCR coordination approach that assigns
tion solution is observed beyond a population size of 500 and a different setting groups to each DOCR for peak and off-peak demand
maximum iteration number of 2000. These parameter settings provide a periods of the day. The aim is to ensure effective relay operations during
good balance between optimisation efficiency and speed. Additionally, fault conditions by reducing relay operating times and to provide reli­
for the other algorithm parameters, the settings recommended in [33] able protection coordination by avoiding maloperation and mis­
are found to be suitable for the problem considered in this study. coordination of the relays. This strategy considers changes in distributed
To analyse the EWSO algorithm’s performance, it is compared with generation, IIDG outages, and load demand variations to achieve a
the basic WSO and two recent metaheuristic algorithms, GJO and HPO. robust DOCR coordination scheme. The analyses and simulations are
For the WSO, GJO, and HPO algorithms, the recommended settings from conducted in two cases, representing the one-year and seasonal imple­
their original studies are used. Each algorithm is run 20 times. Table 8 mentations of the proposed approach. Additionally, to solve the coor­
presents the statistical results obtained for the traditional coordination dination problem in this study, a new variant of the white shark
scheme in terms of minimum (Min), maximum (Max), average (Avg), optimizer, called EWSO, is developed and proposed.
standard deviation (Std), and average CPU time over 20 independent The results show that the proposed approach significantly reduces
runs. It is seen that the EWSO algorithm outperforms the other methods relay operating times, considering operating times obtained via the
in terms of Min, Max, Avg, and Std. Although the results confirm that the traditional approach as a reference. Further, the proposed approach’s
WSO algorithm is the best in terms of computational efficiency, there is effectiveness is high if it is implemented seasonally for a low-load de­
almost no difference in the computational times required by the WSO mand period of the year. Based on the overall analyses and in­
and EWSO. vestigations conducted in this study, the proposed approach is a robust
Fig. 9 shows the average convergence curves corresponding to the alternative to traditional techniques for reliable and effective over­
traditional coordination solutions for the IEEE 14-bus and IEEE 30-bus current protection of IIDG-based distribution networks. The results also
distribution systems. It can be observed that, after the initial itera­ demonstrate that the EWSO algorithm is highly effective for solving the
tions, the proposed EWSO algorithm exhibits faster convergence per­ optimal coordination problem. Although the proposed coordination
formance compared to the basic WSO algorithm. Thus, this result approach relies on an offline computation procedure, the EWSO algo­
confirms the success of the proposed enhancement strategy applied to rithm—due to its ability to achieve effective results with low compu­
the WSO algorithm. tational requirements—is also more advantageous compared to other
The best fitness values obtained for the proposed coordination methods for applications requiring online optimisation processes.
scheme over 20 runs are presented in Table 9. As shown, the EWSO
algorithm produces the best results in all case/period combinations for CRediT authorship contribution statement
both the 14-bus and 30-bus test systems.
Alisan Ayvaz: Writing – original draft, Validation, Software,
Methodology.

12
A. Ayvaz Electric Power Systems Research 237 (2024) 111017

Declaration of competing interest [17] M. Ojaghi, V. Mohammadi, Use of clustering to reduce the number of different
setting groups for adaptive coordination of overcurrent relays, IEEE Trans. Power
Deliv. 33 (3) (2018) 1204–1212, https://doi.org/10.1109/tpwrd.2017.2749321.
The authors declare that they have no known competing financial [18] M.N. Alam, S. Chakrabarti, A.K. Pradhan, Protection of networked microgrids
interests or personal relationships that could have appeared to influence using relays with multiple setting groups, IEEE Trans. Industr. Inform. 18 (6)
the work reported in this paper. (2022) 3713–3723, https://doi.org/10.1109/tii.2021.3120151.
[19] A. Ayvaz, An improved chicken swarm optimization algorithm for extracting the
optimal parameters of proton exchange membrane fuel cells, Int. J. Energy Res. 46
Data availability (11) (2022) 15081–15098, https://doi.org/10.1002/er.8208.
[20] A. Yazdaninejadi, M.S. Naderi, G.B. Gharehpetian, V. Talavat, Protection
coordination of directional overcurrent relays: new time current characteristic and
Data will be made available on request. objective function, IET Gener. Transm. Distrib. 12 (1) (2017) 190–199, https://doi.
org/10.1049/iet-gtd.2017.0574.
[21] V.N. Rajput, F. Adelnia, K.S. Pandya, Optimal coordination of directional
overcurrent relays using improved mathematical formulation, IET Gener. Transm.
References Distrib. 12 (9) (2018) 2086–2094, https://doi.org/10.1049/iet-gtd.2017.0945.
[22] A. Korashy, S. Kamel, T. Alquthami, F. Jurado, Optimal coordination of standard
[1] F.B. dos Reis, J.O.C.P. Pinto, F.S. dos Reis, D. Issicaba, J.G. Rolim, Multi-agent dual and non-standard direction overcurrent relays using an improved moth-flame
strategy based adaptive protection for microgrids, Sustain. Energy Grids Netw. 27 optimization, IEEE Access 8 (2020) 87378–87392, https://doi.org/10.1109/
(2021) 100501, https://doi.org/10.1016/j.segan.2021.100501. access.2020.2992566.
[2] A. Ayvaz, An investigation of the impact of distributed generation penetration on [23] M. Rizwan, C. Gao, L. Hong, S. Ahmad, A.E. Sayed, M. Zaindin, An innovative dual-
directional overcurrent relay coordination in a distribution network, Gazi Univer. setting uni-directional hybrid characteristics for enhanced primary-backup relay
J. Sci. Part A Eng. Innov. 10 (3) (2023) 301–309, https://doi.org/10.54287/ coordination, Electr. Power Syst. Res. 235 (2024) 110545, https://doi.org/
gujsa.1332535. 10.1016/j.epsr.2024.110545.
[3] M.H. Sadeghi, A. Dastfan, Y. Damchi, Robust and adaptive coordination [24] P. Alaee, T. Amraee, Optimal coordination of directional overcurrent relays in
approaches for co-optimization of voltage dip and directional overcurrent relays meshed active distribution network using imperialistic competition algorithm,
coordination, Int. J. Electr. Power Energy Syst. 129 (2021) 106850, https://doi. J. Mod. Power Syst. Clean Energy 9 (2) (2021) 416–422, https://doi.org/
org/10.1016/j.ijepes.2021.106850. 10.35833/mpce.2019.000184.
[4] A. Elmitwally, M.S. Kandil, E. Gouda, A. Amer, Mitigation of dgs impact on [25] A.Y. Hatata, A.S. Ebeid, M.M. El-Saadawi, Optimal restoration of directional
variable-topology meshed network protection system by optimal fault current overcurrent protection coordination for meshed distribution system integrated
limiters considering overcurrent relay coordination, Electr. Power Syst. Res. 186 with DGs based on FCLs and adaptive relays, Electr. Power Syst. Res. 205 (2022)
(2020) 106417, https://doi.org/10.1016/j.epsr.2020.106417. 107738, https://doi.org/10.1016/j.epsr.2021.107738.
[5] A. Narimani, H. Hashemi-Dezaki, Optimal stability-oriented protection [26] H.R.E.H. Bouchekara, M.S. Shahriar, M.S. Javaid, Y.A. Sha’aban, M. Zellagui,
coordination of smart grid’s directional overcurrent relays based on optimized B. Bentouati, A variable neighborhood search algorithm for optimal protection
tripping characteristics in double-inverse model using high-set relay, Int. J. Electr. coordination of power systems, Soft Comput. 25 (2021) 10863–10883, https://doi.
Power Energy Syst. 133 (2021) 107249, https://doi.org/10.1016/j. org/10.1007/s00500-021-05776-4.
ijepes.2021.107249. [27] T. Khurshaid, A. Wadood, S. Gholami Farkoush, C.-H. Kim, J. Yu, S.-B. Rhee,
[6] A. Ayvaz, V.M. Istemihan Genc, Information-gap decision theory based transient Improved firefly algorithm for the optimal coordination of directional overcurrent
stability constrained optimal power flow considering the uncertainties of wind relays, IEEE Access 7 (2019) 78503–78514, https://doi.org/10.1109/
energy resources, IET Renew. Power Gener. 14 (11) (2020) 1946–1955, https:// access.2019.2922426.
doi.org/10.1049/iet-rpg.2019.1367. [28] O. Merabet, M. Bouchahdane, H. Belmadani, A. Kheldoun, A. Eltom, Optimal
[7] M.E. Baran, I. El-Markaby, Fault analysis on distribution feeders with distributed coordination of directional overcurrent relays in complex networks using the Elite
generators, IEEE Trans. Power. Syst. 20 (4) (2005) 1757–1764, https://doi.org/ marine predators algorithm, Electr. Power Syst. Res. 221 (2023) 109446, https://
10.1109/tpwrs.2005.857940. doi.org/10.1016/j.epsr.2023.109446.
[8] Fault current contributions from wind plants, in: IEEE 68th Annu. Conf. Protective [29] J.M. Tripathi, S.K. Mallik, An adaptive protection coordination strategy utilizing
Relay Engineers, IEEE PES Joint Working Group, 2015, pp. 137–227, https://doi. user-defined characteristics of DOCRs in a microgrid, Electr. Power Syst. Res. 214
org/10.1109/CPRE.2015.7102165. (2023) 108900, https://doi.org/10.1016/j.epsr.2022.108900.
[9] A.M. Joshua, K.P. Vittal, Protection schemes for a battery energy storage system [30] P. Kumar, A.S. Rana, Review of optimization techniques for relay coordination in
based microgrid, Electr. Power Syst. Res. 204 (2022) 107701, https://doi.org/ consideration with adaptive schemes of Microgrid, Electr. Power Syst. Res. 230
10.1016/j.epsr.2021.107701. (2024) 110240, https://doi.org/10.1016/j.epsr.2024.110240.
[10] L.H.P. Vasconcelos, A.R. Almeida, B.F. dos Santos Jr, N.X. Melo, J.G.S. Carvalho, [31] S. Kamel, A. Korashy, A.R. Youssef, F. Jurado, Development and application of an
D. de Oliveira Sobreira, Hybrid optimization algorithm applied to adaptive efficient optimizer for optimal coordination of directional overcurrent relays,
protection in distribution systems with distributed generation, Electr. Power Syst. Neural Comput. Appl. 32 (2020) 8561–8583, https://doi.org/10.1007/s00521-
Res. 202 (2022) 107605, https://doi.org/10.1016/j.epsr.2021.107605. 019-04361-z.
[11] F. Alasali, N. El-Naily, A.S. Saidi, A. Itradat, S.M. Saad, W. Holderbaum, An [32] S. Sadeghi, H. Hashemi-Dezaki, A.M. Entekhabi-Nooshabadi, Optimized protection
advanced dual-setting protection scheme for microgrid resilience based on coordination of smart grids considering N-1 contingency based on reliability-
nonstandard tripping characteristics of overcurrent relays, Electr. Power Syst. Res. oriented probability of various topologies, Electr. Power Syst. Res. 213 (2022)
225 (2023) 109869, https://doi.org/10.1016/j.epsr.2023.109869. 108737, https://doi.org/10.1016/j.epsr.2022.108737.
[12] B. Grisales-Soto, A. Herrera-Orozco, J. Mora-Flórez, Advanced adjustment of [33] M. Braik, A. Hammouri, J. Atwan, M.A. Al-Betar, M.A. Awadallah, White Shark
adaptive directional overcurrent relays for active distribution networks in a Optimizer: A novel bio-inspired meta-heuristic algorithm for global optimization
communication-less strategy, Electr. Power Syst. Res. 236 (2024) 110906, https:// problems, Knowl-based Syst 243 (2022) 108457, https://doi.org/10.1016/j.
doi.org/10.1016/j.epsr.2024.110906. knosys.2022.108457.
[13] A. Ataee-Kachoee, H. Hashemi-Dezaki, A. Ketabi, Optimized adaptive protection [34] H. Alhumade, H. Rezk, M. Louzazni, I.A. Moujdin, S. Al-Shahrani, Advanced energy
coordination of microgrids by dual-setting directional overcurrent relays management strategy of photovoltaic/PEMFC/lithium-ion batteries/
considering different topologies based on limited independent relays’ setting supercapacitors hybrid renewable power system using white shark optimizer,
groups, Electr. Power Syst. Res. 214 (2023) 108879, https://doi.org/10.1016/j. Sensors 23 (3) (2023) 1534, https://doi.org/10.3390/s23031534.
epsr.2022.108879. [35] M.A. Ali, S. Kamel, M.H. Hassan, E.M. Ahmed, M. Alanazi, Optimal power flow
[14] M.A. Elsadd, T.A. Kawady, A.-M.I. Taalab, N.I. Elkalashy, Adaptive optimum solution of power systems with renewable energy sources using white sharks
coordination of overcurrent relays for deregulated distribution system considering algorithm, Sustainability 14 (10) (2022) 6049, https://doi.org/10.3390/
parallel feeders, Electr. Eng. 103 (2021) 1849–1867, https://doi.org/10.1007/ su14106049.
s00202-020-01187-0. [36] I. Naruei, F. Keynia, A. Sabbagh Molahosseini, Hunter–prey optimization:
[15] Siemens, Reyrolle 7SR11 and 7SR12 Overcurrent Protection Manual. https://cach algorithm and applications, Soft Comput. 26 (3) (2022) 1279–1314, https://doi.
e.industry.siemens.com/dl/files/281/109747281/att_1133868/v1/7SR11_and_ org/10.1007/s00500-021-06401-0.
7SR12_-_Argus_Complete_Technical_Manual.pdf, 2021 (accessed 10 December [37] N. Chopra, M.M. Ansari, Golden jackal optimization: a novel nature-inspired
2023). optimizer for engineering applications, Expert Syst. Appl. 198 (2022) 116924,
[16] A. Samadi, R.Mohammadi Chabanloo, Adaptive coordination of overcurrent relays https://doi.org/10.1016/j.eswa.2022.116924.
in active distribution networks based on independent change of relays’ setting [38] A.S. Noghabi, J. Sadeh, H.R. Mashhadi, Considering different network topologies
groups, Int. J. Electr. Power Energy Syst. 120 (2020) 106026, https://doi.org/ in optimal overcurrent relay coordination using a hybrid GA, IEEE Trans. Power
10.1016/j.ijepes.2020.106026. Deliv. 24 (4) (2009) 1857–1863, https://doi.org/10.1109/TPWRD.2009.2029057.

13
A. Ayvaz Electric Power Systems Research 237 (2024) 111017

[39] T. Kanungo, D.M. Mount, N.S. Netanyahu, C.D. Piatko, R. Silverman, A.Y. Wu, An [44] Power Systems Test Case Archive, University Washington, 1993. http://www.ee.
efficient k-means clustering algorithm: analysis and implementation, IEEE Trans. washington.edu/research/ (accessed 2 December 2023).
Pattern Anal. Mach. Intell. 24 (7) (2002) 881–892, https://doi.org/10.1109/ [45] R. Hemmati, H. Mehrjerdi, Non-standard characteristic of overcurrent relay for
tpami.2002.1017616. minimum operating time and maximum protection level, Simul. Model. Pract.
[40] J. Yu, C.-H. Kim, S.-B. Rhee, Oppositional Jaya algorithm with distance-adaptive Theory. 97 (2019) 101953, https://doi.org/10.1016/j.simpat.2019.101953.
coefficient in solving directional over current relays coordination problem, IEEE [46] N. Mohammadzadeh, R.M. Chabanloo, M.G. Maleki, Optimal coordination of
Access 7 (2019) 150729–150742, https://doi.org/10.1109/access.2019.2947626. directional overcurrent relays considering two-level fault current due to the
[41] D. Solati Alkaran, M.R. Vatani, M.J. Sanjari, G.B. Gharehpetian, M.S. Naderi, operation of remote side relay, Electr. Power Syst. Res. 175 (2019) 105921,
Optimal overcurrent relay coordination in interconnected networks by using fuzzy- https://doi.org/10.1016/j.epsr.2019.105921.
based GA method, IEEE Trans. Smart Grid. 9 (4) (2018) 3091–3310, https://doi. [47] S. Gupta, K. Deep, H. Moayedi, L.K. Foong, A. Assad, Sine cosine grey wolf
org/10.1109/tsg.2016.2626393. optimizer to solve engineering design problems, Eng. Comput. 37 (2020)
[42] Z. Moravej, O. Soleimani Ooreh, Coordination of distance and directional 3123–3149, https://doi.org/10.1007/s00366-020-00996-y.
overcurrent relays using a new algorithm: grey wolf optimizer, Turk. J. Electr. Eng. [48] N. Rezaei, M.N. Uddin, An analytical review on state-of-the-art microgrid
26 (6) (2018) 3131–3145, https://doi.org/10.3906/elk-1803-123. protective relaying and coordination techniques, IEEE Trans. Ind. Appl. 57 (3)
[43] Türkiye Electricity Statistics, National Load Dispatch Center Information System (2021) 2258–2273, https://doi.org/10.1109/tia.2021.3057308.
(YTBS), 2024. https://ytbsbilgi.teias.gov.tr/ytbsbilgi/frm_istatistikler.jsf. accessed [49] A. Dagar, P. Gupta, V. Niranjan, Microgrid protection: a comprehensive review,
25 July. Renew. Sustain. Energy Rev. 149 (2021) 111401, https://doi.org/10.1016/j.
rser.2021.111401.

14