Renewable and Sustainable Energy Reviews 145 (2021) 111072

Contents lists available at ScienceDirect

Renewable and Sustainable Energy Reviews

journal homepage: http://www.elsevier.com/locate/rser

Metaheuristic search in smart grid: A review with emphasis on planning,

scheduling and power flow optimization applications
M. Papadimitrakis a, N. Giamarelos a, M. Stogiannos a, E.N. Zois a, N.A.-I. Livanos b,
A. Alexandridis a, *
a
Department of Electrical and Electronic Engineering, University of West Attica, Thivon 250, Aigaleo, 12241, Greece
b
EMTECH SPACE P.C., Korinthou 32 & S. Davaki, Metamorfosi 144 51, Athens, Greece

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

Keywords: A multitude of optimization tasks ensue in the context of the smart grid, often exhibiting undesirable charac­
Evolutionary computation teristics like non-convexity, mixed types of design variables and multiple - and often conflicting - objectives.
Metaheuristic search These tasks can be broadly categorized into three classes of problems, namely optimal power flow (OPF),
Optimal power flow
scheduling and planning. Metaheuristic search methods form a generic class of optimization techniques, that
Optimization
have been shown to work successfully for complex problems. Not surprisingly, they have been widely applied in
Planning
Scheduling the smart grid, their use spanning almost every smart grid-related optimization task. In this work, we review the
Smart grid use of metaheuristic search for OPF, scheduling and planning through a unified approach, keeping in mind that
Swarm intelligence these problems share many common challenges and objectives. The use of different metaheuristic methods is
discussed extensively with regard to problem handling, multi-objective optimization performance and method
accuracy in relation to computational complexity. An attempt to arrive at quantitative conclusions is also being
made, by compiling tables which present collective results on common test grids. Lastly, the paper identifies
promising directions for future research, concerning metaheuristic search application practices, method devel­
opment and new challenges that we believe will shape the future of smart grid optimization.

perspective, this new paradigm offers significant potential for the

application of new methods, capable of handling the aforementioned
1. Introduction challenges. These methods must be able to cope with a larger amount of
design variables (DVs) of diverse nature, while taking into account a
Electricity generation and distribution has undergone a major significant number of incoming state measurements from the grid. In
development in the last decades, moving from a conventional central­ addition, they must provide satisfactory solution accuracy combined
ized generation towards a distributed, small-scale, producer-consumer with reasonable computational complexity, even when applied to mul­
(prosumer) model, connected to the distribution network [1]. This tiple optimization objectives. These objectives vary according to the
evolution has been made possible by the development of a reliable in­ network and time horizon specifications; Table 1 summarizes the most
formation and communication infrastructure, but it also gave rise to common ones.
certain challenges, which were met through the emergence of the smart Optimization tasks formed in the context of power grids can be
grid [2]. The technological framework defined by the smart grid enables categorized into three major classes, namely the optimal power flow
a more reliable, more efficient and more economical operation, capable (OPF) [3], scheduling [4] and planning problems [5,6]. OPF describes a
of accommodating increased utilization of renewable energy sources broader set of optimization objectives that apply to the efficient oper­
(RES) and energy storage systems (ESS). On the one hand, system op­ ation of the smart grid bounded by a number of operational constraints.
erators now have an abundance of incoming information and available This amounts to the grid meeting the short-term active and reactive
control decisions at their disposal in order to control critical network power needs of the load buses in an efficient manner. In tasks pertaining
state variables. On the other hand, they must deal with modern chal­ to generation scheduling and planning, the network operators set up
lenges in grid operation that arise from distributed generation and power source scheduling plans, based on the total expected active and
storage, as well as RES stochasticity [1]. From an optimization

* Corresponding author.
E-mail address: [email protected] (A. Alexandridis).

https://doi.org/10.1016/j.rser.2021.111072
Received 11 May 2020; Received in revised form 17 March 2021; Accepted 1 April 2021
Available online 19 April 2021
1364-0321/© 2021 Elsevier Ltd. All rights reserved.
M. Papadimitrakis et al. Renewable and Sustainable Energy Reviews 145 (2021) 111072

List of abbreviations GSA gravitational search algorithm

GWO grey wolf optimization
ABC artificial bee colony HSA harmony search algorithm
ACO ant colony optimization ICA imperialist competitive algorithm
AIS artificial immune systems LB load balance
ALO ant-lion optimization LP linear programming
BFA bacteria foraging algorithm MG micro grid
CHP combined heat and power MINLP mixed integer nonlinear programming
COA coyote optimization algorithm MOO multi-objective optimization
CP contingency planning NRP network reconfiguration problem
CS cuckoo search NSGA-II non-dominated sorting genetic algorithm
CSA crow search algorithm OF objective function
DE differential evolution OPF optimal power flow
DERMS distributed energy resources management systems PSO particle swarm optimization
DERs distributed energy resources PV photovoltaic
DG distributed generation RES renewable energy sources
DSM demand side management RPL real power losses
DVs design variables SA simulated annealing
EC evolutionary computation SI swarm intelligence
ED economic dispatch TLBO teaching-learning based method
EM emissions TST transient stability
EMA exchange market algorithm TS tabu search
ESS energy storage systems TSA tree seed algorithm
FA firefly algorithm UC unit commitment
GA genetic algorithm VNS variable neighborhood search
GC generation cost VPI voltage profile improvement
GR grid resiliency WOA whale optimization algorithm

Table 1
Most common optimization objectives.
Optimization objective Description

Generation cost (GC) Minimization of the total generation cost

Real power losses (RPL) Minimization of power losses
Transient stability (TST) Maximization of grid stability against transient effects
Voltage profile improvement (VPI) Minimization of voltage-related indices
Emissions (EM) Minimization of emissions
Grid resiliency (GR) Minimization of grid-contingency-related indices

reactive power demand. While these two problems share similar ob­ horizon boundaries, as will be discussed later.
jectives, they refer to different horizons; scheduling is oriented towards The extensive arsenal of mathematical programming tools has been
addressing the load demand using the currently available resources, put to use against some of the most challenging grid optimization
while planning forms strategic solutions in order to cope with expected problems before the smart-grid era has emerged. Indeed, standard
demands in the far future by investing in expandability, capacity and mathematical optimization methods have been implemented success­
emissions control. The three problems typically span different time ho­ fully, including Newton, generalized reduced gradient, simplex and
rizons, as shown in Fig. 1; however, latest developments on distributed interior point methods [8–10]. However, most of these methods come
energy resources management systems (DERMS) technology [7] chal­ with three inherent disadvantages, namely (a) poor performance on
lenge this paradigm. In particular, DERMS functionalities allow for the non-convex optimization problems, (b) inability to handle discrete
automation of some of the smart grid tasks, thus obscuring their discrete design variables and (c) unsuitability for multi-objective problems, as

Fig. 1. The current paradigm in smart grid optimization; the three optimization problems span discrete horizon boundaries.

2
M. Papadimitrakis et al. Renewable and Sustainable Energy Reviews 145 (2021) 111072

they cannot easily deal with discontinuous or concave Pareto fronts • Furthermore, we provide a summary of the reviewed techniques in
[11]. Unfortunately, a significant portion of the smart grid-related terms of computational complexity. The commonly reported metric
optimization problems are non-convex, multi-objective in nature and of execution time is supplanted by the total objective function (OF)
contain mixed integer and continuous design variables. Thus, it is easily evaluations until convergence. This metric, which, to the authors’
understood that nowadays, the aforementioned methods may not be the best knowledge has not been used by other reviews in the smart grid
best choice at hand. domain up to now, enables a more objective comparison. At the same
Metaheuristic methods on the other hand constitute a class of opti­ time, it allows conclusions to be drawn independently from factors
mization algorithms that are in principle better equipped to overcome such as type of computing platform, software and special conditions
the difficulties commonly encountered in smart grid optimization and under which each method was run.
provide better quality solutions than traditional approaches. By relying • We identify promising directions for future research on the devel­
on stochastic search, they reduce the risk of getting trapped in local opment and use of metaheuristic search methods for the optimiza­
minima, while the generation and evolution of multiple solutions tion problems encountered in the context of the smart grid. These
employed by most of metaheuristic methods bestows increased explo­ directions include improving the practices of metaheuristic search
ration capabilities. In addition, they can handle non-continuous and application, intensifying method development towards specific di­
concave Pareto fronts and the population-based ones can generate rections and identifying emerging smart grid challenges that befit the
several elements of the Pareto optimal set in a single evaluation. use of metaheuristics.
Metaheuristic methods do come with disadvantages as well. The first
is their overall higher computational burden, which is very important in The structure of this work is as follows: Section 2 introduces some of
the OPF problem solution since they usually require many more function the most popular metaheuristic search methodologies that have been
evaluations than standard mathematical optimization approaches. The applied in smart grid optimization. The literature review for the OPF,
second is the absence of convergence proof for most problems. The third scheduling and planning problems is presented in sections 3-5, respec­
and most persistent, is the need of complex parameter tuning; speed of tively. Section 6 contains a discussion on the literature review given in
convergence is sensitive to tuning and may require tedious trial-and- the previous sections and presents qualitative and - wherever possible -
error runs in order to evaluate the optimal parameter set. quantitative conclusions regarding the performance of metaheuristic
Notwithstanding their disadvantages (which, as will be discussed in methods. The paper concludes by identifying promising avenues for
section 6 can be often mitigated in practice), metaheuristic optimization future research.
methods have been widely employed in the smart grid. Indeed, a
plethora of successful applications have been reported in all three 2. Metaheuristic search methods
optimization problems depicted in Fig. 1. As a response to the growing
research on the application of metaheuristics, a number of survey and Metaheuristic search algorithms form a broad category of optimi­
review papers appeared, focusing on the use of these methods on the zation techniques, which usually receive inspiration from natural, or
individual tasks of OPF [12–15], scheduling [16,17], or planning man-made processes. Though some of the early methods for meta­
[18–20]. These reviews thoroughly cover the literature up to their heuristic search appeared near the middle of the 20th century, the last
respective publication dates and do produce interesting discussions. decades have been an era of overproduction of related works; this can be
Surprisingly though, little attempt is being made to compare the attributed to the ever-growing usage and capabilities of computing
reviewed techniques using the reported results on common benchmark systems. On the other hand, the proliferation of related papers is not
test grids. Moreover, most reviews rarely touch upon the subject of without its critics [23], who mainly focus on the absence of novelty for
computational complexity, which is of great practical importance, some of the works that rely too much on the biological metaphor, while
especially for real-time OPF implementations. neglecting the true mechanics underlying the search process.
This paper reviews the latest applications of metaheuristic optimi­ This section provides a brief overview of the most influential meta­
zation in the optimal power flow, scheduling and planning problems. heuristic search methods, outlines their strengths and weaknesses and
The contributions of this work to the existing reviewing literature are as paves the way for the literature review on their applications in smart
follows: grid optimization tasks.

• We extend the literature review with respect to the application of 2.1. Evolutionary computation
metaheuristic search methods on the individual problems of OPF,
scheduling and planning during the last few years. More importantly, Evolutionary computation (EC) is a widely used computer science
we approach smart grid optimization as a whole, recognizing that discipline comprising methods that simulate the evolution of members
these three problems share a number of common challenges and of a population which are regarded as possible solutions to the optimi­
objectives. Our approach is supported by the recent developments on zation problem. Genetic algorithms (GAs) [24] and differential evolu­
DERMS and microgrid technology [21,22], which lead grid operators tion (DE) [25] comprise the two most distinctive representatives of EC,
to consider these three problems under more flexible horizon albeit a number of other EC techniques have also been proposed [26].
boundaries and in intertwined formulations. Thus, by providing a GAs are able to perform reliably and can easily collaborate with
total perspective of smart grid optimization problems, we draw existing models and systems [27], as well as integrate into hybrid ap­
conclusions of universal applicability regarding the effectiveness of proaches [28]. Additionally, they are easily scalable with parallel
different metaheuristic optimization methods. implementation abilities [29,30] and they impose no restrictions on the
• We provide a comparison of the single-objective and multi-objective functions they process. Their disadvantages include increased difficulty
optimization accuracy of the reviewed metaheuristic methods based in encoding the optimization problem, as well as sensitivity to tuning
on the reported results on common testbeds such as the IEEE-30, or parameters.
the 10–100 generator test grids, wherever possible. Even though Differential evolution [31] was developed in an attempt to improve
limited comparisons are common in individual research papers, to the slow convergence of GAs. The key difference is the use of the
the authors’ best knowledge, very few papers in the reviewing self-referential mutation technique which essentially improves explo­
literature have reported them collectively, much less used them to ration over GAs, based on the differences between random solution
draw qualitative conclusions of universal nature for the problems at pairs. Over the past twenty years, DE has been thoroughly studied and
hand. applied to problems of constrained optimization, parallel computation
[32] and multi-objective optimization (MOO) [25,33]. DE is robust, easy

3
M. Papadimitrakis et al. Renewable and Sustainable Energy Reviews 145 (2021) 111072

to use and requires relatively few control variables. One major disad­ suitable for large-scale optimization problems as it combines the sig­
vantage is the fact that the algorithm heavily relies on the tuning pa­ nificant advantage of reduced computational complexity with reason­
rameters in order to converge efficiently. able performance in terms of accuracy, albeit it cannot compete with
population-based methods in that respect.
2.2. Swarm intelligence
3. Optimal power flow
Swarm intelligence (SI) methods exhibit characteristics found in
decentralized, self-organized groups of biological organisms. SI systems 3.1. Problem formulation
typically consist of a population of simple agents that interact locally
with each other, as well as with their environment; these interactions, The OPF problem, in its generic form, is a large-scale, non-convex,
albeit local, aim to lead to the emergence of a “smart” global behavior. mixed integer nonlinear programming (MINLP) optimization problem,
Examples of SI methods inspired by nature include ant colonies [34], belonging to the NP-hard class. This is partly owed to the nature of the
bird flocking [35], herd farming [36], beehives [37], bacterial growth design variables, which may be discrete, and to the non-continuity, non-
[38], whale hunting [39], dragonfly swarming [40] and the relation differentiability and non-convexity of the objective function. Since the
between trees and their seeds [41], while many other metaphors are mathematical formulation of the OPF is extensively presented in the
used, like the harmony search algorithm (HSA) which models the pro­ literature [3], this work will be confided to a brief descriptive
cess of musical improvisation [42]. introduction.
Particle swarm optimization (PSO) [43] comprises one of the most The OPF problem takes into account operational variables of the grid
important SI methodologies. PSO requires no special encoding, thus (such as voltage magnitudes and angles at every node). Τhe design
enjoying an advantage over methods like GAs in terms of easiness to use. variables generally include active and reactive power generation levels
Its simplicity, combined with effectiveness and speed, makes PSO ideal of the node, transformer tap positions, capacitor bank switch positions,
for use in applications where computational cost is a critical parameter. etc. The problem includes equality constraints, which describe the
Due to these merits, PSO has been widely used, while various modifi­ power flow equations and inequality constraints, corresponding to the
cations were proposed [35,44], aiming to cure its defects, which are operational and safety constraints of the system. Several techniques can
mainly associated with premature convergence. be applied to ease the optimization problem, such as conic relaxations,
A different popular SI method is the ant colony optimization (ACO). or assumptions on the differentiability and continuity of the objective
ACO has proven to be effective in solving combinatorial optimization function, e.g. ignoring valve-point effects of thermal generators [56], or
(CO) problems and has found application in a number of fields in in­ rounding values of reactive compensation devices to the nearest integer
dustry [45], but presents some limitations in dealing with continuous [57].
design variables. Various modifications of the original algorithm have
been reported in the literature [45–47]. 3.2. Metaheuristic methods for OPF

2.3. Artificial immune systems 3.2.1. Swarm intelligence

The first application of a PSO method in OPF takes place in [35],
Artificial immune systems (AIS) are inspired by theoretical immu­ reporting result insensitivity to different numbers of particles. The
nology, simulating the processes used by the biological immune system robustness of PSO-based methods to suboptimal hyperparameter tuning
to respond to external threats. AIS follow a distributed model with an is also demonstrated in [58]. In [59], a PSO method equipped with a
absence of any point of total control, using exclusively local information. multi-stage penalty function for the incorporation of state variable
Due to their inherent decentralized nature, AIS require minimal CPU and constraints is presented. In [60] a faster-converging “coordinate aggre­
memory resources, in contrast to population-based techniques. On the gation” PSO is presented, where the new speed vector of each particle is
other hand, AIS may require customization in order to solve optimiza­ computed using the best-performing particle of the previous iteration.
tion problems, unlike the rest of the methods presented in this section. Convergence acceleration ideas have also been developed for the ACO
AIS-based techniques have been extensively studied and applied to method, such as the decomposition of the constraints to active and
many engineering fields [48], including power grid applications [49, non-active sets [47], or the replacement of ill-performing solutions with
50]. They provide inspiration for hybrid methodologies with remarkable a recombination of the best-performing ones using a GA scheme [46].
characteristics [51]. In [61,62] PSO-based methods are presented, capable of honoring
transient stability constraints described as sets of differential and alge­
2.4. Non population-based metaheuristics braic equations. Their novelty resides in dynamic inertia weights and in
the addition of a stochastic term for toggling between exploration and
Unlike the previously mentioned categories, which evolve a popu­ exploitation. The same strategy is used in [63] for a stability-oriented
lation or swarm of solutions, some of the early metaheuristic methods PSO where the weights are updated according to a stability condition.
were based on modifying a single solution, the most prominent being In [39], the weights of a whale optimization algorithm (WOA) are
simulated annealing (SA) [52] and tabu search (TS) [53]. SA draws updated according to chaotic maps. Dynamic weights are also used in
inspiration from the search of a minimum energy state which occurs [64] through an uncertainty-oriented fuzzy PSO approach. Another
during the process of annealing in metallurgy. Its distinct characteristic enhanced PSO algorithm capable of handling RES uncertainty is pre­
is that it allows for temporarily accepting a worse solution with a sented in [65], which utilizes inverter-based reactive power injections as
probability, which becomes smaller as the iterations progress. SA tuning a design variable; this is oriented to real time applications.
involves only a few parameters while an obvious advantage involves the Uncertainty-oriented methods have also been developed on bacteria
significantly reduced computational cost as a result of operating on a foraging algorithm (BFA) [66], in order to deal with time-varying loads
single solution. Οn the other hand, accuracy is usually inferior compared and wind power uncertainties. In addition, a dragonfly algorithm (DFA)
to population-based methods. Recent advances and SA modifications are is used for multi-objective OPF to a RES-heavy grid [67]. In [68]
reported in [52]. PSO-based method is applied to congested transmission grids taking into
Tabu search owes its name to the so-called tabu lists, that record the account energy pricing schemes while honoring transmission system
search history in order to avoid cycling, i.e. revisiting previously found constraints. Lastly, a more realistic modelling of a microgrids’ slack bus
solutions. The basic idea behind TS has been subjected to a number of characteristics occurs in [69] and the OPF problem is solved using a
modifications, which improve the algorithm’s efficiency [54,55]. TS is PSO-based algorithm.

4
M. Papadimitrakis et al. Renewable and Sustainable Energy Reviews 145 (2021) 111072

Swarm-based techniques that accommodate for reactive power the ED problem, involves allocating power demand to the committed
compensation variables include a variation of the original BFA that units. For this to be accomplished in an efficient way, ED is carried out
directly handles taps [70], or a WOA variant that optimizes shunt for day ahead horizon with hourly intervals. The objective function aims
compensator values in large-scale systems [71]. A variant of the water to minimize the total fuel cost as described in [16]. ED shares constraints
cycle algorithm is employed for reactive power dispatch [72], that (a) - (c) with the UC problem, using additionally the ramp rate limits.
employs a Gaussian solution mutation mechanism in order to dynami­ The aforementioned problems are formulated using as design vari­
cally manage the trade-off between exploitation and exploration. In [73] ables the generated power, the commitment state and the on/off-service
an imperialist competitive algorithm’s (ICA) solution initializes a classic time of units. The objective is to minimize the power system operating
sequential quadratic programming method, resulting in enhanced cost, as well as the environmental cost. Other common scheduling tar­
domain exploration; a tree seed algorithm (TSA) tailored towards the gets include voltage balance, storage operations scheduling and DSM.
same end, is presented in [41]. Lastly, a moth swarm algorithm (MSA) This last procedure is achieved by allocation of shifting loads, peak
and a gravitational search algorithm (GSA) are presented in [74,75] demand management and more recently, improvement of intermittent
respectively, oriented towards the ED of combined heat-power (CHP) generation balancing [85].
plants. The two methods prove to be especially adept at handling the Due to multiple requirements regarding power systems operation,
mathematical adversity of the valve-point effect, entailed in the the objectives are either extended to minimization of power losses,
modelling of CHP plants. congestion cost and ESS cost [86], or multi-objective formulation is
applied [37,87]. Recently, the increasing integration of distributed en­
3.2.2. Evolutionary computation ergy resources (DERs) in power grids has led to the search of new
In [76] a large-scale-system variant of the GA method is presented. techniques for centralized management, i.e. DERMS and virtual power
The design variables are 4-bit coded and the algorithm employs an extra plants (VPPs) [88,89]. The introduction of additional variables such as
search level whenever finer resolution is required. The convergence the status of switches, amount of branch currents, reactive power of
speed of the standard GA can be increased either by initializing a feasible generators, or reliability indices is frequently necessary, leading to a
point [77], or by adaptively changing crossover and mutation proba­ combinatorial optimization problem [90,91]. The set of constraints may
bilities according to the population diversity [78]. Due to the inherent also be related to the specific elements of a distribution grid such as ESS,
parallelization of solutions in the DE, different transient responses of the RES, electric vehicles (EVs), etc. [44].
system can be evaluated at the same time, thus confirming the suitability
for TS-OPF [32]. 4.2. Metaheuristic methods for scheduling
In addition, DE algorithms have been found to be applicable to a
number of multi-objective problems [25,33]. A notable multi-objective 4.2.1. Swarm intelligence
DE algorithm is proposed in [79], where each ε-level of the Over the last decade, SI algorithms and especially PSO, have gained
ε-constraint method is forcedly initialized using the best previous solu­ great appeal in distributed networks scheduling. The problems of ED
tions, resulting in less OF evaluations. In [80] a success-history-based [44,92] and unit commitment [93] for micro grids (MGs) in presence of
adaptive DE (SHADE) adaptively changes its hyper-parameters based uncertainties from wind and solar energy, were dealt using standard
on a historical memory of successful past tunings. The method is applied PSO and recently, hybrid DE-PSO algorithms [94]. In this context,
for multi-objective OPF in the face of grid uncertainties. congestion management [86] and ED [90] are addressed by conven­
tional PSO, thus achieving significant gains in ESS cost reduction and
3.2.3. Artificial immune systems improvements in load profile. Premature convergence is avoided either
AIS methods are useful candidates for OPF applications, mainly due by randomization of particle velocity [95], or by incorporation of the GA
to their low computational complexity. An AIS method with adaptive mutation process [96]. The economic dispatch of an MG in the presence
crossover rates that toggle between exploration and exploitation is of demand response programs is dealt with PSO in [97], considering
provided in [81], while in [51] an AIS algorithm uses the Jacobian of the CHP generation units as a way to increase the efficiency of the operating
power flow equations in order to improve candidate solution mutation. system. A multi-objective PSO approach is applied in [98] for a joint
In [82] a variant of the aforementioned approach, which is oriented for ED-emissions optimization, underlying the beneficial contribution of
large systems is presented. The method rejects clusters of individuals dual-mode CHP units from economical, as well as environmental aspect.
that converge on similar local optima from participating in mutation. Other PSO enhancements include: (a) altering the cost function to better
model the battery charging/discharging operations [99], (b) updating
3.2.4. Non-population based metaheuristics particle speed through quantum bit logic, thus achieving quicker
A standard TS algorithm application takes place in [54], with both convergence [100] and (c) using Gaussian mutation for the strategic
real and discrete design variables, while in [83] a version with an parameters and self-parameterization of the maximum and minimum
adaptive TS length, capable of handling polar constraints is presented. particle velocities [101]. These works aim to minimize the total costs
subject to several constraints regarding the operation of the system,
4. Optimal network scheduling units and battery.
Bio-inspired metaheuristics modified BFA [38,102] and cat swarm
4.1. Problem formulation optimization (CSO) [103], have been proposed for solving the sched­
uling problem of distributed energy resources. According to Refs. [34,
The main operational goal during power generation scheduling is the 45], ACO manages to deal with DSM and optimal storage operations
continuous satisfaction of load demand without interruptions. To this scheduling based on load and renewable production forecast, respec­
end, several scheduling tasks have been established, namely unit tively. The problem of ED in microgrid scale is formulated as linear
commitment (UC), economic dispatch (ED) and demand side manage­ programming (LP) in Refs. [104,105]. A great reduction of the total
ment (DSM). In order to achieve a reliable, as well as economic power generation cost and computation time is obtained by the so-called
supply, power system scheduling is based on the following decision- ant-lion optimization (ALO) and HSA methods, respectively. DSM [91]
making processes. Initially, the unit commitment process takes place, and ED [106] are addressed by artificial bee colony (ABC) and
which is responsible for the minimization of both fuel and start-up cost multi-dimension ICA optimizers, respectively, aiming to market clearing
[84]. The operational constraints comprise of: (a) system power balance, price minimization. Unit commitment as well as demand participation
(b) spinning reserve requirements, (c) unit minimum up and down time are investigated through shuffled frog leaping algorithm (SFLA) [107]
and (d) unit generation limits. Another decision to be made, known as and modified SFLA [108], allowing for the direct accommodation of

5
M. Papadimitrakis et al. Renewable and Sustainable Energy Reviews 145 (2021) 111072

minimum up/downtime constraints during generating feasible solu­ 5. Optimal network planning
tions. In [109], the ED optimization problem considering contingency
issues is decomposed and the binary/continuous variables are handled 5.1. Problem formulation
by different versions of the firefly algorithm (FA). The same problem is
examined using grey wolf optimization (GWO) subject to emissions Planning of power systems is a complex process due to existing al­
minimization [110], exhibiting notable performance and easiness of ternatives, goals, constraints and uncertainties. Typical planning issues
ramp-rate constraint handling. In a recent work [111] GWO is utilized encountered are:
for voltage control through minimization of transformer tap movements,
photovoltaic (PV) power curtailment and voltage rise/drop violations. • Power generation mix selection and sizing: the aim of this task is to
Other multi-objective swarm-based techniques that treat the afore­ achieve high cost-effectiveness, low environmental impact and high
mentioned objectives while accommodating for generation stochasticity reliability.
include a PSO-based method [112] and binary/real coded ABC [37]. • Siting problem: covers power sources allocation and power lines
Integrated approaches based on a modified dragonfly method [40] and a layout in order to keep quality constraints; the objective of low
natural aggregation algorithm (NAA) [113] are used to minimize the power losses is also introduced.
total system cost and the generation cost respectively. The optimal en­ • Reconfiguration: this procedure is primarily performed to minimize
ergy management is attempted through an enhanced version of the most losses, as well as load imbalances, ensuring at the same time a
valuable player algorithm (MVPA) [114], considering CHP units into a satisfactory voltage profile across the system.
microgrid scale.
It should be stressed that simultaneous network reconfiguration and
4.2.2. Evolutionary computation sizing of the distributed generation (DG) produces better results
A standard EC algorithm implementation for UC problem can be compared to the single approach in terms of real power losses (RPL)
found in [115] and an evolutionary lightning flash algorithm is pre­ [130,131] and voltage profile improvement (VPI) [132]. As it can be
sented in [87], involving non-convex and nonlinear characteristics of seen, network scheduling and planning share common aspects, both in
the objective function. DE algorithms have been applied extensively terms of optimization tasks (coordination of generating units, demand
featuring single-objective [116] and multi-objective [117] DSM opti­ participation) and objectives (cost, losses, emission minimization).
mization, as well as ESSM under uncertainties of RES supply and load Nevertheless, due to its intrinsic characteristics, planning is sometimes
demand [118]. Variant approaches addressing ED were also developed, oriented to long-term optimization, as it involves procedures related to
enhanced either by sequence-based mathematical-programming-based network topology and structure, like network expansion [133] and
initialization [88], or by introducing a series of operations into DE al­ equipment maintenance.
gorithm, such as individual ranking, population dividing and population Network planning can be formulated as a single or multi-objective
restructuring [119]. More recently, variants of EC inspired from quan­ problem. However, active power loss minimization is traditionally
tum bits (q-bits) have emerged, allowing the coding of all binary states considered, under the following constraints: (a) distributed generation
in a single solution vector [120,121]. The large-scale energy resource limits, (b) power injection limits, (c) power balance, (d) node voltage
scheduling problem is addressed by hybrid-adaptive differential evolu­ limits, (e) line current limits and (f) radiality. This problem is mathe­
tion with decay function (HyDE-DF) [122], featuring the ability of matically formulated as in [18]. Numerous decision variables at the
self-tuning of parameters and fast convergence capabilities. Other EC disposal of the network operator participate in the problem formulation,
techniques incorporate local search optimizers to enhance exploitation such as: (a) real and reactive power injected by DGs, (b) voltage at
capabilities, forming memetic evolutionary algorithms [123,124]. generation buses, (c) power imported/exported from/to the main grid,
GA is intrinsically suited to solve specific classes of scheduling (d) load shedding values, (e) DG size and location, (f) investment costs
problems like UC, as it allows for simple solution coding for binary and (g) technology assets (storage, dispatchable loads, capacitors etc.)
variables [125]. A memory-based genetic algorithm (MGA) featuring
improved search capability is utilized in [24], whereas in Refs. [126, 5.2. Metaheuristic methods for planning
127], two dynamic pricing schemes are introduced for minimization of
the energy production cost. Advanced GAs have been reported to be 5.2.1. Swarm intelligence
effective in dealing with the UC problem, such as the hybridized GA-TS The mainstream of research efforts on smart grid planning deals with
employed in [28]. A GA is applied in [57] considering interruptive loads issues like optimal sizing, location, number and replacement of units, as
with uncertainties. The minimization of both operating cost and voltage well as their combinations. Similar problems are formulated as LP, or
stability index is achieved by employing a multi-objective GA [128]. MINLP and addressed using conventional PSO, or its variants, in Refs.
[134–136]; satisfactory results are reported regarding convergence
4.2.3. Artificial immune systems speed, robustness and computational effort. A two-stage method is
An efficient and complex combinatorial wind–thermal scheduling described in [137], which initially combines binary PSO together with
problem formulation is presented in [56]. This formulation features a an optimal power flow algorithm to find a feasible reconfiguration set.
reduced number of binary string variables for modelling the Then, the method dispatches reactive power by switching
start-up/shut-down status and ramp rate limits of thermal units. The already-installed capacitors to minimize their losses. Traditional PSO
maximization of profit is achieved by an AIS algorithm. and several modified approaches are introduced for network reconfi­
guration along DG installation [132,138,139], or plain reconfiguration
4.2.4. Non-population based metaheuristics [140,141]. According to [142], the hybrid crow search algorithm
A TS based algorithm is proposed in [53] for day-ahead energy (CSA)-PSO is more efficient in dealing with the above issues.
consumption profile scheduling. This methodology achieves a consid­ Although PSO still remains the most popular among them, the list of
erable reduction in the computational complexity as compared to energy SI algorithms employed for planning is constantly growing. Coyote
consumption scheduling (ECS) algorithms. In [52], an SA method that optimization algorithm (COA) [143] and GWO [144] are applied in DER
generates only feasible neighboring solutions, thus greatly accelerating planning, presenting greater power losses reduction compared to other
UC optimization, is presented. Finally, the improved variable neigh­ methods such as GA and PSO. Stud krill herd algorithm (SKH) [36], as
borhood search (IVNS) algorithm, proposed in [129] for DER scheduling well as the exchange market algorithm (EMA) [145], are addressing the
including DR, appears to perform better than standard VNS and PSO problem of multiple DG allocation and sizing in radial distribution
variants. networks. Two bioinspired metaheuristics, FA and cuckoo search (CS)

6
M. Papadimitrakis et al. Renewable and Sustainable Energy Reviews 145 (2021) 111072

Table 2
Summary of papers on optimal power flow.
# Method Objective Discrete DVs Complexity (CPU/ CPU time/ OF Systems tested Methods compared Population
DVs (max) time, comments) iteration (s) evaluationsa on (system, to size
buses)

[35] PSO GC, VPI N 25 Pentium I – 5000 IEEE 30 GA, Gradient 50

166MHz/- projection
[59] PSO GC, VPI N 25 Pentium IV 2GHz/ 0.417 2000 IEEE 30 GA, Simplex 50
16.72s
[60] CA-PSO RPL, VPI N 75 Pentium V 1.4 0.25 1600 IEEE 30, IEEE GA, PSO, IPM 20
GHz/20s, IEEE 118 118
[61] TS-PSO GC, TS Y 25 Pentium V 1.8 28.8 600 IEEE 30, 39 bus GA 30
GHz/576s, IEEE 30
[62] EPSO GC, TS Y 60 10x Core2 3.2 1.4 2000 IEEE 145, 39 bus Multiple PSO 20
GHz/140s, 39 bus
[64] MF-APSO RPL, VPI N 8 i7 3.6 GHz/136s 2.26 480 24 Bus w/8 PV PSO 8
gen.
[63] PSO RPL N – i7 3.4GHz/- – 5000 IEEE 30, 183 bus – 50
[58] PSO–SSO GC, EM, Y 75 − /− , IEEE 30 – 4500 IEEE 30, IEEE WOA, PSO, etc. 30
RPL, VPI 57, IEEE 118
[65] PPSOGSA GC, RPL, Y 25 − /− , IEEE 30 – 5000 IEEE 30 PSO, GA, DE, etc. 50
VPI
[47] ACO GC, VPI Y 6 Pentium IV 1.4 – – IEEE 30 GA, evolutionary 20
GHz/20s programming
[46] ACO GC, EM N 6 Dual core (2014)/- – – IEEE 30 NSGA, GA 100
[70] BFA RPL, VPI Y 14 − /− – 1000 39 bus SQP-IPM –
[76] IGA GC Y 25 50MHz/5.25min – – IEEE 30 Gradient-based 50
method
[77] GA GC Y 150 Athlon 1.8 MHz/ 0.424 20,000 IEEE 30, IEEE GA 80
106s, IEEE RTS96 118, IEEE RTS96
[78] EGA- GC, RPL, Y 24 − /− – – IEEE 30 PSO-fuzzy 200
DQLF VPI
[32] DE TS, GC N 33 61x Xeon 2.66 1.95 12,000 9, 39, 162 bus – 120
GHz/195s, 162 bus
[33] MODE GC, RPL, Y 34 − /− – – IEEE 30 KHA, ICA 50
VPI, EM
[25] ESDE GC, RPL, Y 34 i3 2.2 GHz/130s, – 18,000 IEEE 30, IEEE DE, Tent map, Sine –
VPI, EM IEEE 57 57, 59 bus map etc.
[80] DE VPI, RPL Y 27 i5 2.7 GHz/148s, 1.783 5000 IEEE 30, IEEE 57 – 60
IEEE 30
[79] MODEA VPI, GC, Y 27 − /− – 10,000 IEEE 30, IEEE 57 PSO, DE 50
RPL
[81] MOAIA VPI Y 12 Celeron 633MHz/- – 1000 IEEE 30 Immune GA 100
[82] CGbAIS GC, VPI Y 65 Core2 2.4 GHz/ – – IEEE 118, 325 DE, evolutionary 50
105s, 325 buses bus PSO
[54] TS GC, VPI Y 16 /- – – IEEE 30 Evolutionary –
programming
[83] RTS GC, EM N 6 -/200s – – IEEE 30 TS, Simplex –
[73] HMICA- GC, RPL, Y 130 -/28.74s, IEEE 30 – – IEEE 118, IEEE ICA, GA, EA –
SQP RESP 30, IEEE 57
[41] TSA GC, RPL Y 55 i7 2.4 GHz/309s, 0.618 6000 IEEE 57, IEEE PSO, ABC, Moth 12
IEEE 300 300 swarm
[39] CWOA TS, GC Y 21 Pentium IV 2.3 2.8 2000 39 bus, 162 bus WOA, DE 50
GHz/112s, 162 bus
[71] WOA VPI, RPL Y 39 i7 2.9GHz/-, 114 – 900 IEEE 14, IEEE PSO 30
bus 30, 113 bus
[72] NGBWCA VPI, RPL Y 75 Pentium IV/-, 118 – 50,000 IEEE 30, IEEE GA, PSO, DE, etc. 50
bus 57, IEEE 118
a
Values refer to the same tested system that is mentioned in the “Complexity” column.

algorithm, are combined in [146] for multi-objective planning, pro­ distributed generation. The problem of optimal DG [153] along with ESS
ducing a technique that has proven useful in assessing the dynamic [154] placement and sizing is dealt with dominated groups search
stability of the power system. Multi-objective allocation and sizing of DG optimization (DGSO) and inherited competitive swarm optimization
is investigated by (a) a CS algorithm that demonstrates superior (ICSO), respectively.
convergence speed [147], (b) an ALO with loss sensitivity factors
capable of handling large-scale distribution systems [148] and (c) an 5.2.2. Evolutionary computation
HSA capable of generating a more converged Pareto front approxima­ Evolutionary optimization approaches have been used to explore
tion than a non-dominated sorting genetic algorithm (NSGA-II) [149]. RPL minimization through optimal placement and sizing of DG [155].
The minimization of both overloads and voltage deviations is combined RPL problem is addressed in a single [26] or multi-objective optimiza­
in a multi-objective framework presented in [150], where ICA is tion framework [131], using different evolutionary algorithms, namely
employed for solving the flexible alternating current transmission sys­ GOMEA and SPEA2, respectively. Fuzzy set theory is used to select the
tems (FACTS) allocation problem. Satisfactory results are obtained by best compromise solution in case of multiple objectives.
NAA [151] and SEUMRE [152] when applied to energy storage system A probabilistic formulation of shunt capacitor placement problem in
management (ESSM) and ESSM-DSM, respectively, considering distribution networks is followed in [156] and dealt with a GA. The

7
M. Papadimitrakis et al. Renewable and Sustainable Energy Reviews 145 (2021) 111072

Table 3
Summary of papers on scheduling optimization.
# Method Objectives/ Complexity (CPU/time, CPU time/ OF Systems tested on Methods compared to Population
Tasks comments) iteration (s) evaluationsa size

[86] PSO ESSM − /− – 15,000 IEEE 24 – 300

[101] Mut-PSO DSM -/68s 0.618 2200 937 bus PSO, EPSO 20
[95] PSO ED, EM Core 3.4 GHz/7862s, 100 7.862 200,000 10–100 units NSGA, GA, BFA, EA, 200
units etc.
[97] PSO ED, DSM − /− – – 33 bus MG – –
[98] PSO ED, EM Pentium IV 1.66 GHz/-, 40 – 18,000 10–40 units PSO, LR, ESA 30
units
[99] PSO ED − /− – 5000 11 units – 50
[100] PSO ED Pentium IV 2 GHz/158s, 40 0.527 9000 10–100 units – 30
units
[104] ALO ED -/1.69s 0.042 400 5 units GA, ABC, BFA 10
[107] SFLA ED Pentium IV 2 GHz/1430s, 28.6 10,000 10–100 units LR, BFA, GA, PSO 200
100 units
[108] MSFLA ED − /− – – 32 bus MG GA, PSO, DE, ICA, –
SFLA
[109] FFA ED, CP Pentium IV 3.2 GHz/945s, 1.89 50,000 IEEE 24, IEEE 118, 38, EP, PSO, GA, etc. 100
100 units 100 units
[114] MVPA ED i7 2.50 GHz/- – 1000 3 MGs GA, PSO, BH, ABC, 100
EM
[75] GSA ED Pentium 2 GHz/43.2s, 48 8.64 500 5, 7, 24, 48 units PSO, EP, DE, RCGA, 100
units BCO, etc
[116] EC DSM − /− – 60,000 smart grid – 200
[110] GWO ED, EM, CP i3 2.2GHz/-, 10 units – 5000 10–100 units, 38 units BFA, GA, EP, FLA, SA, 50
etc.
[112] FS- UC i7-4790 3.6GHz/2490.3s 3.11 48,000 26 units MOPSO, TV-MOPSO 60
MOPSO
[37] BRABC UC Pentium IV 3.2GHz/14.02s, 0.028 100,000 6, 10, 40 units, IEEE PSO, EP, GA, etc. 200
40 units 30
[40] ANFMDA ED Core i5/34.8s 0.87 8000 3 units DA, ABC, WOA, 200
HOMER
[115] EP ED -/1 h 42 m, 100 units 1.224 250,000 10–100 units LR, GA, EP 50
[88] SDE ED Core i7 3.6 GHz/2.37s 0.024 331,100 6 units PSO, PSOw, CBA, 3311
MFA
[119] CDE ED − /− – 90,000 7 bus DE, PSO 400
[87] LFA ED, EM -/2.90s, IEEE 10 0.58 50 IEEE 10, 14, 30, 40, EP, PSO, HS 10
160
[92] PSO ED − /− – 125,000 3 MGs – 250
[120] IQEA ED Core2 Duo 2.66 GHz/292s, 0.73 7200 10–100 units LR, GA, EP, PSO, etc. 18
100 units
[121] QEA ED Core 2.39 GHz/80s, 100 0.4 3600 10–100 units GA, SA, PSO, LR etc. 18
units
[122] HyDE-DF Xeon E5-2620 v2/326.79s 0.817 2000 201 bus DE, PSO, Vortex, 5
HyDE
[123] DE ED Core i5-2410 M 2.3GHz/ 0.118 3000 10–100 units PSO, GA, ACO, ICA 30
11.82s, 100 units etc
[124] NSGA-II ED, EM Core i5 3.4 GHz/4254s, 100 17.579 12,100 10, 100 units NSGA, SA, ICA 50
units
[125] GA ED Pentium IV 1.6GHz/242.5s, 0.808 15,000 10–100 units LR, GA 50
100 units
[24] MGA ED Core i7 2.6GHz/2.06s 0.016 13,000 IEEE 37 GA, PSOcf, PSOw 100
[28] GA ED − /− – 35,000 10 units FLA, GA 50
[57] GA DSM − /− – 12,600 – RDO, RS, SA 210
[127] GA ESSM i7-3630QM 2.4 GHz/- – 2400 – – 200
[129] IVNS ED, DSM − /− – 70,000 33, 180 bus CBBO, PSO, VNS –
[56] AIS UC Core i3 2.67 GHz/100s 5 1200 12 units LR, Fuzzy, PSO 60
[94] EVDEPSO UC Core i7 2.6GHz/53.56s 0.188 2850 25 bus VNS, PSO, IDE, FFA 10
[106] MICA DSM − /− – 432,000 – C-PLEX 100
[126] GA DSM Core i7 1.7GHz/- – 328,000 – Analytical 400
approaches
[102] BFA ED -/6s 1.5 40 MG GA, ABC 10
a
Values refer to the same tested system that is mentioned in the “Complexity” column.

authors in Refs. [157,158] apply a biased random-key genetic algorithm a GPU-based version utilizing minimum spanning trees to sort candidate
(BRKGA) and NSGA-II, in a multi-stage, or a multi-objective manner, in topologies [29], are applied to the DG allocation problem and achieve
order to address the network reconfiguration problem (NRP). The significant speedup. The authors of [160] propose GA for optimal DG
maximization of priority loads restoration during blackout with multiple placement and sizing, while in Ref. [161] NRP is also taken into
line faults is addressed through a self-healing scheme [159], where consideration. The crucial task of contingency planning (CP) in distri­
reconfiguration is carried out using a GA. In [130] NRP, along with bution networks was studied using GAs [162], while maintaining load
selection of optimal number and size of DGs is considered. The paral­ balance (LB). A multi-objective GA is employed for RPL along with VPI
lelization of GAs has attracted attention lately; a multi-processor GA in distribution networks, using TOPSIS to pinpoint the compromise so­
with fuzzy-adaptive rules for efficient processor data handling [30] and lution [163]. Lastly, the joint problem of DG and storage allocation is

8
M. Papadimitrakis et al. Renewable and Sustainable Energy Reviews 145 (2021) 111072

Table 4
Summary of papers on planning optimization.
# Method Objectives/ Complexity (CPU/time, CPU time/iteration OF Systems tested on Methods compared Population
Tasks comments) (s) evaluationsa to size

[151] NAA DSM − /− – 120,000 IEEE 15 - 48

[137] BPSO LB − /− – 1000 16 buses - 50
[132] PSO RPL, VPI -/16.2s 0.184 4400 IEEE 33 GA 50
[135] PSO DSM, LB − /− – 8 – BA, SMO 50
[140] MPSO RPL i7 2.4GHz/- – 250 IEEE 33, IEEE 69 PSO 50
[36] KHA RPL i3 3.3GHz/- – 960 IEEE 33, IEEE 69 GA PSO, etc. 10
[146] FA/CS RPL, VPI i5/11s, IEEE 30 1.1 500 ІЕЕЕ 14, IEEE 30 ABC, GSA, etc. 50
[147] CS RPL − /− – 1200 IEEE 69 GA, PSO 20
[148] ALO RPL − /− – 15,000 69 buses GA, CSA, etc. 30
[153] DGSO RPL, VPI -/44s, 69 buses 74.8 30 IEEE 30, 69 buses GA, PSO, etc. 51
[154] ICSO RPL, VPI i3 3.19 GHz/- – 2 IEEE 37 – 30
[165] GA DSM, GC − /− – 20 Smart grid – 100
[159] GA RPL, LB, VPI -/14s 7 100 33 buses – 50
[26] GOMEA RPL − /− – 100,000 31 buses GA 1024
[158] NSGA-IІ RPL, VPI, RESP − /− – 144 16 nodes GA 16
[168] LAHC CP − /− – 5000 570 feeders ILS HC –
[50] AIS RPL, VPI, LB 2.8GHz/32.3s, IEEE 69 0.646 2500 IEEE 33, 69 GA 50
[152] SEUMRE DSM -/315.64s 14.819 426 IEEE 33 PSO, GA 20
[49] AIS RPL i7 3.1GHz/0.365s, IEEE 33 0.183 100 IEEE 33, 84, 136 ACO, AIS 50
[166] AIS RPL Q8200 2.33GHz/16.9s 1.69 80 IEEE 33 Heuristics 8
[167] AIS RPL C-60/- – 96 IEEE 33, 69 Heuristics 8
[143] COA RPL, VPI -/3047s, IEEE 123 6.094 12,500 IEEE 123, 137 PSO, GA etc. 25
buses
[150] ICA LB, VPI − /− – 40,000 IEEE 14, 39 ABC, EP, etc. 800
a
values refer to the same tested system that is mentioned in the “Complexity” column.

considered through the prism of DSM of industrial [164] or residential the smart grid. Nevertheless, it is impossible to single out one method, or
[165] loads and is solved by a GA. even one class of methods, that performs better across all tasks, a remark
well in accordance with the “no free lunch” theorem [113]. Keeping this
5.2.3. Artificial immune systems in mind, this section provides a discussion on cross-comparisons be­
Power loss reduction has been addressed with the use of AIS [50], tween metaheuristic approaches and points of superiority versus con­
through optimal location and sizing of distribution static compensators ventional mathematical optimization methods, within the following
(DSTATCOM). In Refs. [49,166], AIS algorithms are applied to the lines: (a) Problem handling (i.e. how the methods cope with discrete
problem of radial distribution systems reconfiguration with respect to variables, uncertainty, constraints, etc.), (b) computational complexity
operational constraints. The same problem is solved in [167], while also with regard to accuracy and (c) multi-objective optimization perfor­
taking into account generation and load uncertainties. mance. A unified approach is being followed to draw conclusions about
the three different power grid optimization problems, discussing com­
5.2.4. Non-population based metaheuristics mon elements that connect them and facilitating the identification of
A number of stochastic local search metaheuristics for integrated directions for future research in the section to follow.
reconfiguration, reinforcement and extension plans of large-scale net­
works, is evaluated in [168].
6.1. Problem handling

6. Discussion and conclusions

The aforementioned power grid problems are naturally, in their
exact formulation, non-convex, non-differentiable and sometimes non-
To form a basis for discussion, in Tables 2–4 we present a quick
continuous. Some of these shortcomings can be compensated by
overview of the reviewed papers on OPF, scheduling and planning,
applying assumptions on the differentiability of OFs [56] and the con­
respectively, including only the papers that can be compared in terms of
tinuity of discrete design variables [57,169]. The non-convexity prop­
quantitative results and/or computational complexity. The tables
erty is usually alleviated by employing convex relaxation techniques
contain information such as the adopted method, the objective, hyper­
such as semi-definite programming [152]. While these assumptions can
parameter settings, indicators of computational complexity and the
assist any class of search methods, in practice traditional optimization
methods used for comparison. Furthermore, Tables 5–7 present results
methods require them in order to converge to an acceptable solution in
for the fuel cost, emissions, power losses and voltage control objectives,
reasonable time [12]. Unfortunately, this may lead to subpar results,
wherever these are reported on common benchmarks such as IEEE-30
since rounding discrete variables leads to large errors [169], while
[35] or 10–100 generator [115] grids. To be more specific, Table 5 is
relaxation techniques may fail to produce a globally optimal solution.
concerned with OPF results, while Tables 6 and 7 show unit commitment
Furthermore, they have trouble incorporating discrete variables, due to
and grid reconfiguration results, which pertain to scheduling and
incompatibility with the resulting high-order semi-definite program­
planning optimization, respectively. It should be pointed out that some
ming constraints [170].
works report results from a single run, while in other works, multiple
Since most metaheuristic methods are non-gradient-based, they can
simulation runs are performed. To establish a fair basis of comparison,
readily handle the non-differentiability of generator cost curves which
Tables 5–7 depict the best result reported in each work, as far as the
contain sine components [54,88], or are piecewise quadratic [77,151].
objective function is concerned. Finally, function evaluations are shown
Moreover, there exist metaheuristic methods, such as GAs, that inher­
in Tables 5 and 6, which, as will be discussed later, can be used to assess
ently encode discrete variables in chromosomes [78,156], while
the methods’ computational complexity.
population-based methods like PSO and ACO are traditionally oriented
It is obvious that the main metaheuristic method categories have
towards continuous variables. A number of authors choose to employ
been applied invariably to all the major optimization tasks pertaining to
both classes of methods for the solution of the unit commitment task,

9
M. Papadimitrakis et al. Renewable and Sustainable Energy Reviews 145 (2021) 111072

Table 5
Results on ІЕЕЕ-30 for single-objective and multi-objective OPF.
Single-objective OPF

# Method Fuel cost ($/h) Emissions (ton/h) Voltage metric Losses (MW) Function
Evaluations

[83] RTS 605.39d 0.1941 – – –

[25] ESDE 799.03 0.2047 0.1241f 2.84 10,000
[35] PSO 800.41 – 0.1246f – 5000
647.69d
[58] PSO–SSO 798.98 0.205 0.124f 2.86 4500
[77] GA 617.11d – – – 30,000
801.05
[59] PSO – – – 4.38 2000
[65] PPSOGSA 800.52 – 0.0898h 3.10 5000
[80] DE – – 0.0888a,h 4.41a 12,000
[79] MODEA 799.08 – 0.0939h 2.853 10,000
[71] WOA – – – 4.59 900
[76] IGA 800.80 – – – –
[60] CA-PSO – – 0.1225h 5.09 600
[33] MODE 607.19d 0.1942 – – –
[61] TS-PSO 585.17d,e – – – 600
[54] TS 802.29b – – – –
[73] HMICA-SQP 863.72c – – – 400
[78] EGA-DQLF 799.56 – 0.1040f,a 3.20 –
[47] ACO 803.12 – – – –
[72] NGBWCA – – 0.0458h 4.48 –

Multi-objective OPF

# Method Fuel cost ($/h) Emissions (ton/h) Voltage metric Losses (MW)

[83] RTS 623.11d 0.1970 – –

[25] ESDE 799.60a – 0.1246f,a 5.22b
827.15b
[35] PSO 806.38 – 0.0891h –
[65] PPSOGSA 829.59 – 0.1103h 6.11
[33] MODE 615.50d 0.2012 – –
[79] MODEA 820.88b – 0.1249f,b 5.59a
799.69a
[78] EGA-DQLF 822.87a – 0.1056 g,b
5.61a
802.06b
[58] PSO–SSO 834.80a 0.243a 0.1246f,b 4.09c
830.35b 0.224c
865.18c
a
Incorporates stochastic RES generation.
b
Solutions violate the slack bus lower q-limit.
c
Includes valve point loading effects.
d
Piecewise generator cost curve.
e
Transient stability constraints.
f
Voltage stability index (L-index).
g
Sum of squared voltage stability indices.
h
Total voltage deviation.

Table 6
Results for single-objective day-ahead unit commitment.
# Method Fuel cost ($) Function Evaluations

10 20 40 60 80 100

[52] SA 565,828 1,126,521 2,250,063 – 4,498,076 5,617,876 –

[115] EC 564,551 1,125,494 2,249,093 3,371,611 4,498,479 5,623,885 25,000/50,000/100,000/150,000/200,000/250,000
[120] EC 563,938 1,123,297 2,243,039 3,363,059 4,485,875 5,605,795 1440/7200/7200/7200/7200/1440
563,938 1,123,297 2,242,982 3,362,507 4,484,088 5,603,355
[100] PSO 563,977 1,123,297 2,242,957 3,361,980 4,482,085 5,602,486 9000 (40-unit)
[107] SFLA 564,769 1,123,261 2,246,005 3,368,257 4,503,928 5,624,526 2000 (10-unit)
[125] GA 566,404 1,127,244 2,254,123 3,378,108 4,498,943 5,630,838 12,300 (20-unit)
[28] GA 562,979 – – – – – 20,000 (40-unit)
[121] EA 563,938 1,123,607 2,245,557 3,366,676 4,488,470 5,609,550 3600 (all systems)
[95] PSO 563,938 1,123,297 2,242,953 3,361,738 4,482,417 5,601,457 400,000 (100-unit)
[124] GA 563,938 – – – – 5,605,918 10,000 (all systems)
[110] GWO 563,936 1,123,892 2,239,051 3,355,735 4,471,858 5,591,197 –
[109] FFA – – – – – 5,601,298 60,000 (100-unit)
[98] PSO 565,149 2,255,384 – – – –

10
M. Papadimitrakis et al. Renewable and Sustainable Energy Reviews 145 (2021) 111072

Table 7 6.2. Computational complexity & accuracy

Results of grid reconfiguration optimization.
# Method IEEE-33 IEEE-69 When discussing the performance of an optimization algorithm, the
balance between computational complexity and accuracy for the task at
Loss Minimum Loss Minimum
Reduction voltage (p. Reduction voltage (p. hand must be properly assessed. In the power grid case, it used to be that
(kW) u.) (kW) u.) the OPF problem was primarily concerned with achieving such a bal­
[140]a MPSO 56.70 0.940 100.30 0.942
ance, due to its short time horizon. However, with the increased pene­
[138]a MCPSO 63.75 0.942 123.62 0.949 tration of RES, the scheduling problem horizon is also shifting from the
[132]a PSO 76.30 0.893 – – typical day-ahead formulation to an hours-ahead one – therefore, its
GA 67.40 0.893 computational complexity must be considered too. Lastly, the planning
[141]a SPSO 69.53 0.942 126.38 0.949
problem usually employs a longer optimization horizon and therefore its
BPSO 66.81 0.936 125.84 0.923
[155] b
DE 116.77c – 152.41c – execution time may be of less practical importance. Accuracy of the
174.19d 203.72d results, on the other hand, determined by the search method’s ability to
[36]b KHA 123.82d 0.968 150.12d 0.979 efficiently explore the search domain and to converge to satisfactory
[145]b EMA 116.74d 0.963 150.10d 0.979 solutions, remains important for all three power grid problems. In this
[130]b GA 148.49c 0.972 – –
[163]b GA 123.81d 0.968d 151.12d 0.972
context, metaheuristic methods have a significant advantage over
[166]a AIS 139.55 0.93 – – standard mathematical optimization approaches as they are better
[167]a AIS 127.2 0.95 77.32 0.9522 equipped to handle the non-convexity present in most power grid
[160]a GA 63.13 0.938 125.36 0.943 problem formulations. Gradient-based techniques on other hand, usu­
a
reconfiguration. ally get easily trapped in local minima, thus compromising accuracy
b
reconfiguration & 2 DGs placement and sizing. [10]. This point is especially exacerbated in the case of the CHP eco­
c
DG type I: Active generation model. nomic dispatch problem, where accurate modelling calls for the incor­
d
DG type II: Active and reactive generation model. poration of valve-point loading effects [88]. This is accomplished
through the addition of a sinusoidal term to the cost function of the CHP
dedicating the discrete-coded method to handling the binary generator units, which generates a high number of local minima [75]. Lastly,
status variables and the real-coded method to handling the continuous population-based metaheuristics, though being more computationally
generator output variables [98,109]. intensive, are suitable for parallelization [13], thus fully exploiting the
Unfortunately, the OPF problem often cannot be decomposed in this advantages of multi-core processors, or computing clusters [32].
manner. This necessitates the application of a continuous algorithm, A common criticism against metaheuristic methods involves the lack
employing rounding operations to enforce discreteness wherever of rigorous mathematical analysis and more specifically, convergence
needed. A number of authors claim that this practice does not sacrifice analysis. While this remains true for the majority of the methods, there
solution quality in the case of transformer tap variables [62], since the exist theoretical studies proving convergence for significant meta­
discrete interval is so small that it can be ignored. However, VPI per­ heuristics like PSO [173] and more research is underway [174].
formance seems to suffer when treating reactive shunt compensator Furthermore, the plethora of successful applications even on NP-hard
variables as continuous valued [169]. Considering the relevant results of problems, as well as the large number of reported results on
PSO-based methods of [35,60], versus their DE-based counterparts of well-known power grid benchmark studies (Tables 2–4), provide some
[25,80], it is apparent that the latter achieve significantly better VPI evidence on the superiority of metaheuristic search methods over
performance, as shown in Table 5. This is further reflected on the results traditional mathematical optimization approaches from a practical point
of RPL minimization, which is also dependent on effective reactive of view.
power control. Indeed, GA [78] as well as DE-based [25] methods seem As far as cross-comparison between metaheuristic methods is con­
to achieve superior performance through direct handling of integer cerned, it must be stressed that an attempt to objectively rank different
values of reactive control devices, when compared to their swarm-based metaheuristic methods in terms of effectiveness on the smart grid opti­
counterparts. mization problems is a rather difficult task. Firstly, method accuracy is
Smart grid optimization problems often involve variables exhibiting dependent on problem formulation, a multitude of which exist. Sec­
a considerable amount of uncertainty, e.g. the output power of RES, or ondly, the computational burden cannot be directly inferred by the
the load demand [171]. A probabilistic approach is a common way to method runtime; even though this is the most reported metric in the
deal with uncertainty, aiming to estimate the statistical parameters of literature, it can vary in practice even when the method is run multiple
the relevant variables [86,130]. Stochastic scenario-based optimization times on the same CPU. Regarding accuracy, objective conclusions can
methods can then be formulated, where the objective function is opti­ only be drawn when cross-comparing results on the same testbeds and
mized for a set of possible scenarios [93,134,152,156]. Fuzzy set theory cases. Similarly, a more objective metric regarding computational
can also be applied to represent uncertainties, especially when there is complexity can be set as the total number of function evaluations:
insufficient information for the probabilistic parameters of the uncertain
Neval = niter × npop (1)
variables [112]. Even though such techniques result to more complex
objective functions, metaheuristic search methods are well equipped to where niter is the number of algorithm iterations until convergence and
handle them efficiently. npop is the population, or swarm size. This metric has been computed for
Finally, it must be noted that the vast majority of optimization
all the works where relevant information was available and results are
methodologies can inherently handle only unconstrained problems. It is
reported in Tables 5 and 6 Therefore, the results of each method can be
a fact that the OPF, scheduling and planning problems include hard
compared on this basis and conclusions can be drawn regarding suit­
constraints based on the system’s operational characteristics, such as
ability towards a specific application.
voltage and current constraints [35,50], ramp-up/down limits and
As shown in Table 5, evolutionary computation methods (DE/GA)
min/max operating times [91] and power constraints of DG units [30].
seem to establish improved solutions for all single-objective OPF for­
To overcome this problem, several techniques have been devised, such
mulations on an average basis, but their main downside is the high
as the penalty functions method (PFM), the feasible solutions method
number of function evaluations required for convergence. Indeed, EC
(FSM), specialized operators, repair algorithms, etc [88,172].
methods that surpass swarm-based ones on the respective objectives, do
so at a burdensome computational price. For example, when comparing

11
M. Papadimitrakis et al. Renewable and Sustainable Energy Reviews 145 (2021) 111072

OPF can be formulated as MOO problems, since they comprise multiple

and often conflicting, objectives [78]. An effective MOO method must
approximate the Pareto front as closely as possible. Conventional
mathematical optimization algorithms may fail in that regard, even
though they are simple to implement. Firstly, they produce a single
solution per iteration; therefore, they need to employ approaches like
the weighted sum [37], or the ε-constraint [79] techniques, which
require multiple runs in order to generate a Pareto front approximation.
In the case of weighted sum, the approximation may fail to be
well-converged or well-diversified when the underlying front is
non-convex or non-continuous. The importance of these two charac­
teristics is visually depicted in Fig. 2, where an underlying Pareto front
and its approximations are shown. Approximation 1 is well-diversified
Fig. 2. Examples of approximations of the Pareto front of a two- but not well-converged, approximation 2 is well-converged but not
objective function.
well-diversified and approximation 3 is both well-converged and
well-diversified. Another possible shortcoming of conventional mathe­
PSO-based [35] and GA-based [77] methods, a 4.6% decrease in the fuel matical optimization algorithms is that the approximation they provide
cost objective came at a 500% increase in function evaluations. Ulti­ may be unacceptable, even when the underlying Pareto front is convex.
mately, the practitioner must decide if additional accuracy is warranted, An example is shown in [33], where a sequential quadratic program­
since it may compromise the feasibility of a real-time OPF imple­ ming algorithm is coupled with the weighted sum method.
mentation. These observations are confirmed for larger OPF problems Metaheuristic MOO methods can deal with challenging Pareto fronts
too: Test systems such as the IEEE-57 and IEEE-118 translate to signif­ and, especially population-based metaheuristics can potentially
icantly higher function evaluations for EP methods [77,80], compared generate an approximation of the Pareto front with a single evaluation
to swarm-based methods. The latter appear generally more insensitive to [11,158]. The quality of the Pareto front approximation can be assessed
problem dimensionality [60,73]. on the basis of solution set diversity and distance from the ideal solution.
The above observations do not seem to apply for the single-objective A number of metrics exist which measure the diversity, such as hyper­
scheduling and planning problems. Swarm-based techniques achieve, on volume or inter-generational distance [158]. In the OPF literature, a few
average, superior accuracy throughout the different scheduling test authors evaluate the proposed MOO methods on this basis; most choose
systems [110], as shown in Table 6. Also, we can confirm the presence of to cross-compare the reported results using only a compromise solution,
EC algorithms that require less OF evaluations than PSO both on me­ as discussed in the next paragraph. In contrast, the usage of the afore­
dium (40 units) and large scale (100 units) systems. Regarding the mentioned metrics in the scheduling and planning literature is wide­
planning problem, we can only comment on the quantitative results, as spread, while compromise solutions are rarely reported.
OF evaluations are not available. As shown on Table 7, PSO variants Regardless of the means used for evaluating the results of a MOO
achieve larger losses reduction compared to GA, when reconfiguration is method, once the non-dominated solution set is obtained, a final
examined. Nevertheless, when DG sizing and placement are examined, compromise solution must be implemented. This is a selection made by
EC methods do exhibit superior performance compared to swarm-based the decision maker (DM), who can be either a physical entity (such as
ones [155]. However, the number of instances is so small that it pro­ the grid operator), or an automated process; the latter is preferred in the
hibits universal conclusions. case of the OPF problem, due to its short time horizon. The most
A general observation to be noted is that the population parameter of straightforward approach and the most common one in the reviewed
most swarm-based techniques (notwithstanding a few special exceptions OPF literature, is to select the knee-point solution of the set (that is, the
[110,120]) lies in the relatively small range of {20, 50}, regardless of solution with the minimum Euclidean distance from the ideal solution)
test system size, problem type, or problem dimensionality. Generally, [64]. More sophisticated approaches (prevalent in all three power grid
this is not a novel observation for SI-based methods. Sensitivity analyses problems) use fuzzy DM techniques [25,78,134,179]. Heuristic DM
for various non-power grid related engineering problems have arrived at techniques include the ranking of solutions based on a secondary
the same conclusion [175], noting that too small a population would fail “objective” that is not present in the MOO problem; for example the
to explore all of the search domain, while too large would require more minimization of RPL [64], or the TOPSIS method [46,163].
computational resources, without effectively aiding towards a higher In summary, the literature review provides evidence that there is a
solution exploitation. Moreover, although swarm-based methods seem growing interest towards robust MOO algorithms in the face of different
to dominate applications for the power grid problems, they demonstrate system states [58], generation scenarios [80], or generation cost fluc­
some variability between the converged solutions [66,67]. DE-based tuations [46]. Other studies [25,73,121] perform sensitivity analysis on
methods seem to perform better in this respect, as illustrated in [79], the hyper-parameters of the proposed MOO methods, confirming their
where the proposed method repeatedly converges on the same robustness to a suboptimal tuning.
compromise solution.
Another feature adding to the complexity of metaheuristic methods 7. Recommendations and suggestions for future research
is the large number of hyperparameters. While a suboptimal hyper­
parameter tuning will seldom incur failure of convergence, one must As far as the practices of metaheuristic search application for smart
specifically tune their metaheuristic method in order to achieve the best grid optimization are concerned, the following recommendations are
results possible for the problem at hand. Research has progressed in derived:
order to overcome this disadvantage as metaheuristic algorithms such as
GWA [110] and TLBO [176,177] aim to use fewer number of parame­ • There is a need for a consistent benchmarking framework regarding
ters. Another interesting approach involves using meta-optimization the application of optimization tasks in smart grid, so that different
algorithms based on EC for tuning hyperparameters [178]. metaheuristic approaches can be compared objectively. As far as the
planning tasks are concerned, with the exception of network recon­
6.3. Multi-objective optimization performance figuration, there are no common benchmark problems to test the
proposed optimization approaches. But even for OPF and scheduling,
Optimization tasks encountered within planning, scheduling and where benchmarks such as the IEEE-30 and 10–100 generator test

12
M. Papadimitrakis et al. Renewable and Sustainable Energy Reviews 145 (2021) 111072

grids exist, a direct comparison between different methods is not an

easy task, as it has become apparent from the previous discussion.
The reason is that each study usually applies different optimization
objectives and reports different metrics. Thus, a compact set of key
performance indicators should be established for each optimization
task, in order to assess the methods’ accuracy and complexity. With
respect to the latter, we propose that the number of function evalu­
ations until convergence should be reported along with the usual
CPU run time, so that computational complexity can be objectively
inferred. Regarding MOO, the use of standard metrics such as
hypervolume and inter-generational distance [124] is encouraged
for evaluating the resulting Pareto front approximations in OPF
problems.
• The need of hyperparameter tuning for optimal performance has
been also identified as an important drawback of existing - and quite
probably of upcoming - metaheuristic optimization methods. A va­
riety of different optimization tasks occurs within the smart grid and
the use of meta-optimization algorithms could help practitioners to
adjust the hyperparameters of metaheuristics to the task at hand.
Nevertheless, researchers in the field of metaheuristics are encour­
aged to report hyperparameter sensitivity analyses for their proposed Fig. 3. The future paradigm in smart grid optimization: The emerging DERMS
methods. technology enables some tasks to transcend their usual time horizon boundaries
• Research on load and renewable generation prediction using ma­ and be incorporated in a common time frame.
chine learning models, including neural networks and other similar
approaches that can cope with the complexity and nonlinearity of the exploiting the search space. Taking [58, 73, 142] as starting points,
problem, has boomed in the last decades [180,181]. The vast ma­ more hybrid approaches should be investigated.
jority of the reviewed literature though, does not incorporate such
models within the formulation of the optimization problem; most of Finally, new smart grid challenges are emerging that could be
the papers treat future load and renewable generation as known, or addressed using metaheuristic methods:
at best use estimations for their statistical parameters. Machine
learning models could provide far more accurate estimations for • RES inverter-based reactive power control has recently been made
these quantities, albeit at the cost of increasing the computational possible by the actualization of DERMS. While this has attracted
complexity, due to the nonlinearity of such approaches. Meta­ significant research attention for real-time grid voltage control
heuristic search methods though, are very well equipped to handle [184], very few works were implemented in OPF [65]. Its inclusion
efficiently the computational burden imposed by employing data entails challenges on an optimization level, since the bound con­
driven learning-based models and a few implementations in the ED straints of the problem are transformed to nonlinear inequalities (an
[91,102,104] and UC problems [124] have only appeared recently. inverters’ reactive power capacity is bounded by the injected active
Intensifying research on this subject and expanding the application power generation, which is a design variable).
to different smart grid optimization tasks is a fruitful research • The development of decentralized units based on renewables has
direction. given rise to energy markets. The clearing of the all-European elec­
tricity market is already an active field [185] and the EUPHEMIA
Regarding the development of metaheuristic search techniques, the algorithm [186] deals with the unification of power exchanges
authors believe that research should be more focused on improving the through the solution of a rather complicated multi-zonal market
search mechanisms with regard to the challenges of the smart grid coupling and the harmonized allocation of potentially large numbers
optimization tasks. To this end, the following directions are suggested: of energy orders. However, the procurement of ancillary services still
presents serious difficulties [187]. Thus, it is essential for smart grid
• Most metaheuristic search methods are inherently capable of paral­ scheduling optimization to integrate market price schemes [188].
lelization, which can lead to significant performance gains [32]. The non-convexity, nonlinearity and use of mixed-type variables
Intensification of research on more efficient ways of parallelization [185] make the use of metaheuristics a promising approach in this
[30], including distributed and cellular algorithms for direction, but only limited research has taken place during the last
population-based metaheuristics and multi-start algorithms for three years, featuring implementations to the ED [91] and DSM [106,
non-population based ones [182], could further boost the perfor­ 126] tasks. Furthermore, the integration of energy market clearing
mance, allowing applicability to real-time OPF. intensifies even more the need for predictive models to obtain
• Cooperative algorithms have been developed that partition the so­ day-ahead energy price forecasts [127].
lution candidate pool into clusters that each seek to optimize • Recent developments on DERMS, together with the increasing
different parts of the solution vector through cooperation [183]. practical application of microgrids [2], have transcended the stan­
Though such schemes have already proven to be very effective for dard horizon boundaries of the classic planning, scheduling and
high-dimensional problems, the development of similar methods for optimal power flow problems, allowing for some of their respective
modern smart grid optimization, which employs a much higher optimization tasks to converge, such as in Refs. [164,189]. A new
number of design variables than before, is up to now very scarce paradigm emerges, as shown in Fig. 3. DERMS, together with the new
[138]. communication infrastructure, has enabled the automatic evaluation
• Hybrid metaheuristic algorithms form an emerging trend in smart of several smart grid tasks, such as active & reactive power dispatch
grid optimization. The combination of diverse metaheuristics can (OPF), unit commitment (scheduling) and network reconfiguration
lead to new exciting approaches since hybridization can be used to (planning), allowing them to be carried out concurrently. This new
combine the advantages of different metaheuristics in exploring and paradigm accommodates for the efficient incorporation of RES gen­
eration and microgrid operation. It is the authors’ intuition that the

13
M. Papadimitrakis et al. Renewable and Sustainable Energy Reviews 145 (2021) 111072

future of smart grid optimization involves tightly coupled tasks [20] RezaeeJordehi A. DG allocation and reconfiguration in distribution systems by
metaheuristic optimisation algorithms: a comparative analysis. In: Proc. - 2018
under a holistic optimization framework, entailing a number of
IEEE PES innov. Smart grid technol. Conf. Eur. ISGT-europe 2018; 2018. https://
serious challenges that can be addressed successfully by meta­ doi.org/10.1109/ISGTEurope.2018.8571802.
heuristic optimization methods. [21] Thakar S, Vijay AS, Doolla S. System reconfiguration in microgrids. Sustain
Energy, Grids Networks 2019;17. https://doi.org/10.1016/j.segan.2019.100191.
[22] Khazraj H, Khanghah BY, Ghimire P, Martin F, Ghomi M, Da Silva FF, et al.
Declaration of competing interest Optimal operational scheduling and reconfiguration coordination in smart grids
for extreme weather condition. IET Gener, Transm Distrib 2019;13:3455–63.
https://doi.org/10.1049/iet-gtd.2019.0507.
The authors declare that they have no known competing financial [23] Sörensen K. Metaheuristics - the metaphor exposed. Int Trans Oper Res 2015;22:
interests or personal relationships that could have appeared to influence 3–18. https://doi.org/10.1111/itor.12001.
the work reported in this paper. [24] Askarzadeh A. A memory-based genetic algorithm for optimization of power
generation in a microgrid. IEEE Trans Sustain Energy 2018;9:1081–9. https://doi.
org/10.1109/TSTE.2017.2765483.
Acknowledgements [25] Pulluri H, Naresh R, Sharma V. An enhanced self-adaptive differential evolution
based solution methodology for multiobjective optimal power flow. Appl Soft
Comput J 2017;54:229–45. https://doi.org/10.1016/j.asoc.2017.01.030.
This research has been co-financed by the European Union and Greek [26] Grond MOW, Luong HN, Morren J, Pan Bosman, Slootweg HJG, La Poutré H.
national funds through the Operational Program Competitiveness, Practice-oriented optimization of distribution network planning using
Entrepreneurship and Innovation, under the call RESEARCH – CREATE – metaheuristic algorithms. In: Proc. - 2014 power syst. Comput. Conf. PSCC 2014.
Institute of Electrical and Electronics Engineers Inc.; 2014. https://doi.org/
INNOVATE (project code: T1EDK-00244). 10.1109/PSCC.2014.7038349.
[27] Leonori S, Paschero M, Massimo F, Mascioli F, Rizzi A. Optimization strategies for
References Microgrid energy management systems by Genetic Algorithms. Appl Soft Comput
J 2019;105903. https://doi.org/10.1016/j.asoc.2019.105903.
[28] Sarjiya, Mulyawan AB, Setiawan A, Sudiarso A. Thermal unit commitment
[1] Hatziargyriou N. Microgrids: architectures and control. 2013. https://doi.org/
solution using genetic algorithm combined with the principle of tabu search and
10.1002/9781118720677.
priority list method. In: Proc - 2013 int conf inf technol electr eng "intelligent
[2] Strasser T, Andrén F, Kathan J, Cecati C, Buccella C, Siano P, et al. A review of
green technol sustain dev ICITEE 2013; 2013. p. 414–9. https://doi.org/10.1109/
architectures and concepts for intelligence in future electric energy systems. IEEE
ICITEED.2013.6676278.
Trans Ind Electron 2015;62:2424–38. https://doi.org/10.1109/
[29] Roberge V, Tarbouchi M, Okou FA. Distribution system optimization on graphics
TIE.2014.2361486.
processing unit. IEEE Trans Smart Grid 2017;8:1689–99. https://doi.org/
[3] Frank S, Rebennack S. Division of economics and business working paper series A
10.1109/TSG.2015.2502066.
primer on optimal power Flow : theory, formulation, and practical examples,
[30] Asrari A, Lotfifard S, Ansari M. Reconfiguration of smart distribution systems
vols. 1–42; 2012.
with time varying loads using parallel computing. IEEE Trans Smart Grid 2016;7:
[4] Shahidehpour M, Yamin H, Li Z. Market operations in electric power systems.
2713–23. https://doi.org/10.1109/TSG.2016.2530713.
2002. https://doi.org/10.1002/047122412x.
[31] Jebaraj L, Venkatesan C, Soubache I, Rajan CCA. Application of differential
[5] Ehsan A, Yang Q. Optimal integration and planning of renewable distributed
evolution algorithm in static and dynamic economic or emission dispatch
generation in the power distribution networks: a review of analytical techniques.
problem: a review. Renew Sustain Energy Rev 2017. https://doi.org/10.1016/j.
Appl Energy 2018;210:44–59. https://doi.org/10.1016/j.apenergy.2017.10.106.
rser.2017.03.097.
[6] Jordehi AR. Optimisation of electric distribution systems: a review. Renew
[32] Cai HR, Chung CY, Wong KP. Application of differential evolution algorithm for
Sustain Energy Rev 2015. https://doi.org/10.1016/j.rser.2015.07.004.
transient stability constrained optimal power flow. IEEE Trans Power Syst 2008;
[7] Petrovic N, Strezoski L, Dumnic B. Overview of software tools for integration and
23:719–28. https://doi.org/10.1109/TPWRS.2008.919241.
active management of high penetration of DERs in emerging distribution
[33] Varadarajan M, Swarup KS. Solving multi-objective optimal power flow using
networks. EUROCON 2019 - 18th Int. Conf. Smart Technol. 2019. https://doi.
differential evolution. IET Gener, Transm Distrib 2007;2:896–903. https://doi.
org/10.1109/EUROCON.2019.8861765.
org/10.1109/ICPST.2006.321538.
[8] Momoh JA, El-Hawary ME, Adapa R. A review of selected optimal power flow
[34] Tisseur R, De Bosio F, Chicco G, Fantino M, Pastorelli M. Optimal scheduling of
literature to 1993 Part I: nonlinear and Quadratic Programming Approaches.
distributed energy storage systems by means of ACO algorithm. In: Proc. - 2016
IEEE Trans Power Syst 1999;14:96–103. https://doi.org/10.1109/59.744492.
51st int. Univ. Power eng. Conf. UPEC 2016, vol. 2017. Institute of Electrical and
[9] Momoh JA, El-Hawary ME, Adapa R. A review of selected optimal power flow
Electronics Engineers Inc.; 2016. p. 1–6. https://doi.org/10.1109/
literature to 1993 Part II: Newton, linear programming and interior point
UPEC.2016.8114101.
methods. IEEE Trans Power Syst 1999;14:96–104. https://doi.org/10.1109/
[35] Abido MA. Optimal power flow using particle swarm optimization. Int J Electr
TPWRS.1987.4335072.
Power Energy Syst 2002;24:563–71. https://doi.org/10.26634/jee.4.1.1252.
[10] Frank S, Steponavice I, Rebennack S. Optimal power flow: a bibliographic survey
[36] ChithraDevi SA, Lakshminarasimman L, Balamurugan R. Stud Krill herd
I Formulations and deterministic methods. Energy Syst 2012;3:221–58. https://
Algorithm for multiple DG placement and sizing in a radial distribution system.
doi.org/10.1007/s12667-012-0056-y.
Eng Sci Technol an Int J 2017;20:748–59. https://doi.org/10.1016/j.
[11] Coello Coello CA, Lamont GB, Van Veldhuizen DA. Evolutionary algorithms for
jestch.2016.11.009.
solving multi-objective problems. second ed. New York: Springer; 2007. https://
[37] Chandrasekaran K, Simon SP. Multi-objective unit commitment problem with
doi.org/10.1007/978-0-387-36797-2.
reliability function using fuzzified binary real coded artificial bee colony
[12] Frank S, Steponavice I, Rebennack S. Optimal power flow: a bibliographic survey
algorithm. IET Gener, Transm Distrib 2011;6:1060–73. https://doi.org/10.1049/
II Non-deterministic and hybrid methods. Energy Syst 2012;3:259–89. https://
iet-gtd.2012.0193.
doi.org/10.1007/s12667-012-0057-x.
[38] Panda A, Tripathy M, Barisal AK, Prakash T. A modified bacteria foraging based
[13] AlRashidi MR, El-Hawary ME. Applications of computational intelligence
optimal power flow framework for Hydro-Thermal-Wind generation system in the
techniques for solving the revived optimal power flow problem. Elec Power Syst
presence of STATCOM. Energy 2017;124:720–40. https://doi.org/10.1016/j.
Res 2009;79:694–702. https://doi.org/10.1016/j.epsr.2008.10.004.
energy.2017.02.090.
[14] Abdi H, Beigvand SD, Scala M La. A review of optimal power flow studies applied
[39] Prasad D, Mukherjee A, Shankar G, Mukherjee V. Application of chaotic whale
to smart grids and microgrids. Renew Sustain Energy Rev 2017;71:742–66.
optimisation algorithm for transient stability constrained optimal power flow. IET
https://doi.org/10.1016/j.rser.2016.12.102.
Sci Meas Technol 2017;11:1002–13. https://doi.org/10.1049/iet-smt.2017.0015.
[15] Niu M, Wan C, Xu Z. A review on applications of heuristic optimization
[40] Murugaperumal K, Ajay P. Energy storage based MG connected system for
algorithms for optimal power flow in modern power systems. J Mod Power Syst
optimal management of energy: an ANFMDA technique. Int J Hydrogen Energy
Clean Energy 2014;2:289–97.
2019;44:7996–8010. https://doi.org/10.1016/j.ijhydene.2019.02.002.
[16] Rahman I, Mohamad-Saleh J. Hybrid bio-Inspired computational intelligence
[41] El-Fergany AA, Hasanien HM. Tree-seed algorithm for solving optimal power flow
techniques for solving power system optimization problems: a comprehensive
problem in large-scale power systems incorporating validations and comparisons.
survey. Appl Soft Comput J 2018;69:72–130. https://doi.org/10.1016/j.
Appl Soft Comput J 2018;64:307–16. https://doi.org/10.1016/j.
asoc.2018.04.051.
asoc.2017.12.026.
[17] Mahdi FP, Vasant P, Kallimani V, Watada J, Fai PYS, Abdullah-Al-Wadud M.
[42] Huang CM, Huang YC. Hybrid optimisation method for optimal power flow using
A holistic review on optimization strategies for combined economic emission
flexible AC transmission system devices. IET Gener, Transm Distrib 2014;8:
dispatch problem. Renew Sustain Energy Rev 2018. https://doi.org/10.1016/j.
2036–45. https://doi.org/10.1049/iet-gtd.2014.0096.
rser.2017.06.111.
[43] Jordehi AR. Particle swarm optimisation (PSO) for allocation of FACTS devices in
[18] Badran O, Mekhilef S, Mokhlis H, Dahalan W. Optimal reconfiguration of
electric transmission systems: a review. Renew Sustain Energy Rev 2015. https://
distribution system connected with distributed generations: a review of different
doi.org/10.1016/j.rser.2015.08.007.
methodologies. Renew Sustain Energy Rev 2017;73:854–67. https://doi.org/
[44] Borges N, Soares J, Vale Z. A robust optimization for day-ahead microgrid
10.1016/j.rser.2017.02.010.
dispatch considering uncertainties. IFAC-PapersOnLine 2017;50:3350–5. https://
[19] Zahraee SM, Khalaji Assadi M, Saidur R. Application of artificial intelligence
doi.org/10.1016/j.ifacol.2017.08.521.
methods for hybrid energy system optimization. Renew Sustain Energy Rev 2016.
https://doi.org/10.1016/j.rser.2016.08.028.

14
M. Papadimitrakis et al. Renewable and Sustainable Energy Reviews 145 (2021) 111072

[45] Silva A, Marinheiro J, Cardoso HL, Oliveira E. Demand-side management in [71] oualid Medani K ben, Sayah S, Bekrar A. Whale optimization algorithm based
power grids: an ant colony optimization approach. In: Proc. - IEEE 18th int. Conf. optimal reactive power dispatch: a case study of the Algerian power system. Elec
Comput. Sci. Eng. CSE 2015. Institute of Electrical and Electronics Engineers Inc.; Power Syst Res 2018;163:696–705. https://doi.org/10.1016/j.epsr.2017.09.001.
2015. p. 300–6. https://doi.org/10.1109/CSE.2015.31. [72] Heidari AA, Ali Abbaspour R, Rezaee Jordehi A. Gaussian bare-bones water cycle
[46] Mousa AAA. Hybrid ant optimization system for multiobjective economic algorithm for optimal reactive power dispatch in electrical power systems. Appl
emission load dispatch problem under fuzziness. Swarm Evol Comput 2014;18: Soft Comput J 2017. https://doi.org/10.1016/j.asoc.2017.04.048.
11–21. https://doi.org/10.1016/j.swevo.2014.06.002. [73] Ben Hmida J, Chambers T, Lee J. Solving constrained optimal power flow with
[47] Allaoua B, Laoufi A. Optimal power flow solution using ant manners for electrical renewables using hybrid modified imperialist competitive algorithm and
network. Adv Electr Comput Eng 2009;9:34–40. https://doi.org/10.4316/ sequential quadratic programming. Elec Power Syst Res 2019;177:105989.
aece.2009.01006. https://doi.org/10.1016/j.epsr.2019.105989.
[48] Stogiannos M, Alexandridis A, Sarimveis H. An enhanced decentralized artificial [74] Elattar EE. Optimal power flow of a power system incorporating stochastic wind
immune-based strategy formulation algorithm for swarms of autonomous power based on modified moth swarm algorithm. IEEE Access 2019;7:89581–93.
vehicles. Appl Soft Comput J 2020;89:106135. https://doi.org/10.1016/j. https://doi.org/10.1109/ACCESS.2019.2927193.
asoc.2020.106135. [75] Beigvand SD, Abdi H, La Scala M. Combined heat and power economic dispatch
[49] Souza SSF, Romero R, Pereira J, Saraiva JT. Artificial immune algorithm applied problem using gravitational search algorithm. Elec Power Syst Res 2016;133:
to distribution system reconfiguration with variable demand. Int J Electr Power 160–72. https://doi.org/10.1016/j.epsr.2015.10.007.
Energy Syst 2016;82:561–8. https://doi.org/10.1016/j.ijepes.2016.04.038. [76] Lai LL, Ma JT, Yokoyama R, Zhao M. Improved genetic algorithms for optimal
[50] Taher SA, Afsari SA. Optimal location and sizing of DSTATCOM in distribution power flow under both normal and contingent operation states. Int J Electr Power
systems by immune algorithm. Int J Electr Power Energy Syst 2014;60:34–44. Energy Syst 1997;19:287–92. https://doi.org/10.1016/s0142-0615(96)00051-8.
https://doi.org/10.1016/j.ijepes.2014.02.020. [77] Todorovski M, Rajičić D. An initialization procedure in solving optimal power
[51] De Mello Honório L, Da Silva AML, Barbosa DA. A gradient-based artificial flow by genetic algorithm. IEEE Trans Power Syst 2006;21:480–7. https://doi.
immune system applied to optimal power flow problems. Lect Notes Comput Sci org/10.1109/TPWRS.2006.873120.
2007;4628. https://doi.org/10.1007/978-3-540-73922-7_1. LNCS:1–12. [78] Kumari MS, Maheswarapu S. Enhanced genetic algorithm based computation
[52] Simopoulos DN, Kavatza SD, Vournas CD. Unit commitment by an enhanced technique for multi-objective optimal power flow solution. Int J Electr Power
simulated annealing algorithm. IEEE Trans Power Syst 2006;21:68–76. https:// Energy Syst 2010;32:736–42. https://doi.org/10.1016/j.ijepes.2010.01.010.
doi.org/10.1109/TPWRS.2005.860922. [79] Shaheen AM, El-Sehiemy RA, Farrag SM. Solving multi-objective optimal power
[53] Assaf T, Osman AH, Hassan MS, Mir H. Fair and efficient energy consumption flow problem via forced initialised differential evolution algorithm. IET Gener,
scheduling algorithm using tabu search for future smart grids. IET Gener, Transm Transm Distrib 2016;10:1634–47. https://doi.org/10.1049/iet-gtd.2015.0892.
Distrib 2018;12:643–9. https://doi.org/10.1049/iet-gtd.2017.0247. [80] Biswas PP, Suganthan PN, Mallipeddi R, Amaratunga GAJ. Optimal reactive
[54] Abido MA. Optimal power flow using tabu search algorithm. Elec Power Compon power dispatch with uncertainties in load demand and renewable energy sources
Syst 2002;30:469–83. https://doi.org/10.1109/MPER.2002.1005652. adopting scenario-based approach. Appl Soft Comput J 2019;75:616–32. https://
[55] Lenin K, Ravindhranath Reddy B, Suryakalavathi M. Hybrid Tabu search- doi.org/10.1016/j.asoc.2018.11.042.
simulated annealing method to solve optimal reactive power problem. Int J Electr [81] Xiong H, Cheng H, Li H. Optimal reactive power flow incorporating static voltage
Power Energy Syst 2016;82:87–91. https://doi.org/10.1016/j. stability based on multi-objective adaptive immune algorithm. Energy Convers
ijepes.2016.03.007. Manag 2008;49:1175–81. https://doi.org/10.1016/j.enconman.2007.09.005.
[56] Lakshmi K, Vasantharathna S. Gencos wind-thermal scheduling problem using [82] Honório LM, Leite da Silva AM, Barbosa DA, Delboni LFN. Solving optimal power
Artificial Immune System algorithm. Int J Electr Power Energy Syst 2014;54: flow problems using a probabilistic α-constrained evolutionary approach. IET
112–22. https://doi.org/10.1016/j.ijepes.2013.06.036. Gener, Transm Distrib 2010;4:674–82. https://doi.org/10.1049/iet-
[57] Bastani M, Thanos AE, Damgacioglu H, Celik N, Chen CH. An evolutionary gtd.2009.0208.
simulation optimization framework for interruptible load management in the [83] Nualhong D, Chusanapiputt S, Phomvuttisarn S, Jantarang S. Reactive tabu
smart grid. Sustain Cities Soc 2018;41:802–9. https://doi.org/10.1016/j. search for optimal power flow under constrained emission dispatch. IEEE Reg 10
scs.2018.06.007. Conf 2004;3:327–30.
[58] El Sehiemy RA, Selim F, Bentouati B, Abido MA. A novel multi-objective hybrid [84] Abujarad SY, Mustafa MW, Jamian JJ. Recent approaches of unit commitment in
particle swarm and salp optimization algorithm for technical-economical- the presence of intermittent renewable energy resources: a review. Renew Sustain
environmental operation in power systems. Energy 2020;193:116817. https:// Energy Rev 2017. https://doi.org/10.1016/j.rser.2016.11.246.
doi.org/10.1016/j.energy.2019.116817. [85] Jordehi AR. Optimisation of demand response in electric power systems, a
[59] Zhao B, Guo CX, Cao YJ. Improved particle swam optimization algorithm for OPF review. Renew Sustain Energy Rev 2019;103:308–19. https://doi.org/10.1016/j.
problems. In: IEEE PES power syst conf expo 2004; 2004. p. 933–8. https://doi. rser.2018.12.054.
org/10.1109/psce.2004.1397582. [86] Hemmati R, Saboori H, Jirdehi MA. Stochastic planning and scheduling of energy
[60] Vlachogiannis JG, Lee KY. A comparative study on particle swarm optimization storage systems for congestion management in electric power systems including
for optimal steady-state performance of power systems. IEEE Trans Power Syst renewable energy resources. Energy 2017;133:380–7. https://doi.org/10.1016/J.
2006;21:1718–28. https://doi.org/10.1109/TPWRS.2006.883687. ENERGY.2017.05.167.
[61] Mo N, Zou ZY, Chan KW, Pong TYG. Transient stability constrained optimal [87] Kheshti M, Kang X, Li J, Regulski P, Terzija V. Lightning flash algorithm for
power flow using particle swarm optimisation. IET Gener, Transm Distrib 2007;1: solving non-convex combined emission economic dispatch with generator
476–83. https://doi.org/10.1049/iet-gtd. constraints. IET Gener, Transm Distrib 2018;12:104–16. https://doi.org/
[62] Xia S, Chan KW, Bai X, Guo Z. Enhanced particle swarm optimisation applied for 10.1049/iet-gtd.2017.0257.
transient angle and voltage constrained discrete optimal power flow with flexible [88] Yang Y, Wei B, Qin Z. Sequence-based differential evolution for solving economic
AC transmission system. IET Gener, Transm Distrib 2015;9:61–74. https://doi. dispatch considering virtual power plant. IET Gener, Transm Distrib 2019;13:
org/10.1049/iet-gtd.2014.0038. 3202–15. https://doi.org/10.1049/iet-gtd.2018.6432.
[63] Luo J, Shi L, Ni Y. A solution of optimal power flow incorporating wind [89] Nosratabadi SM, Hooshmand RA, Gholipour E. A comprehensive review on
generation and power grid uncertainties. IEEE Access 2018;6:19681–90. https:// microgrid and virtual power plant concepts employed for distributed energy
doi.org/10.1109/ACCESS.2018.2823982. resources scheduling in power systems. Renew Sustain Energy Rev 2017. https://
[64] Yang HT, Liao JT. MF-APSO-Based multiobjective optimization for PV system doi.org/10.1016/j.rser.2016.09.025.
reactive power regulation. IEEE Trans Sustain Energy 2015;6:1346–55. https:// [90] Esmaeili S, Anvari-Moghaddam A, Jadid S, Guerrero JM. Optimal simultaneous
doi.org/10.1109/TSTE.2015.2433957. day-ahead scheduling and hourly reconfiguration of distribution systems
[65] Ullah Z, Wang S, Radosavljevic J, Lai J. A solution to the optimal power flow considering responsive loads. Int J Electr Power Energy Syst 2019;104:537–48.
problem considering WT and PV generation. IEEE Access 2019;7:46763–72. https://doi.org/10.1016/j.ijepes.2018.07.055.
https://doi.org/10.1109/ACCESS.2019.2909561. [91] Marzband M, Azarinejadian F, Savaghebi M, Guerrero JM. An optimal energy
[66] Panda A, Tripathy M. Security constrained optimal power flow solution of wind- management system for islanded microgrids based on multiperiod artificial bee
thermal generation system using modified bacteria foraging algorithm. Energy colony combined with Markov chain. IEEE Syst J 2017;11:1712–22. https://doi.
2015;93:816–27. https://doi.org/10.1016/j.energy.2015.09.083. org/10.1109/JSYST.2015.2422253.
[67] Sureshkumar K, Ponnusamy V. Power flow management in micro grid through [92] Nikmehr N, Najafi Ravadanegh S. Optimal power dispatch of multi-microgrids at
renewable energy sources using a hybrid modified dragonfly algorithm with bat future smart distribution grids. IEEE Trans Smart Grid 2015;6:1648–57. https://
search algorithm. Energy 2019;181:1166–78. https://doi.org/10.1016/j. doi.org/10.1109/TSG.2015.2396992.
energy.2019.06.029. [93] Saber AY, Venayagamoorthy GK. Resource scheduling under uncertainty in a
[68] Bachtiar Nappu M, Arief A, Bansal RC. Transmission management for congested smart grid with renewables and plug-in vehicles. IEEE Syst J 2012;6:103–9.
power system: a review of concepts, technical challenges and development of a https://doi.org/10.1109/JSYST.2011.2163012.
new methodology. Renew Sustain Energy Rev 2014. https://doi.org/10.1016/j. [94] Dabhi D, Pandya K. Enhanced velocity differential evolutionary particle swarm
rser.2014.05.089. optimization for optimal scheduling of a distributed energy resources with
[69] Moradi MH, Foroutan VB, Abedini M. Power flow analysis in islanded Micro- uncertain scenarios. IEEE Access 2020;8:27001–17. https://doi.org/10.1109/
Grids via modeling different operational modes of DGs: a review and a new ACCESS.2020.2970236.
approach. Renew Sustain Energy Rev 2017. https://doi.org/10.1016/j. [95] Shukla A, Singh SN. Multi-objective unit commitment using search space-based
rser.2016.11.156. crazy particle swarm optimisation and normal boundary intersection technique.
[70] Tripathy M, Mishra S. Bacteria foraging-based solution to optimize both real IET Gener, Transm Distrib 2015;10:1222–31. https://doi.org/10.1049/iet-
power loss and voltage stability limit. IEEE Trans Power Syst 2007;22:240–8. gtd.2015.0806.
https://doi.org/10.1109/TPWRS.2006.887968.

15
M. Papadimitrakis et al. Renewable and Sustainable Energy Reviews 145 (2021) 111072

[96] Xu Z, Deng T, Hu Z, Song Y, Wang J. Data-driven pricing strategy for demand-side [121] Lau TW, Chung CY, Wong KP, Chung TS, Ho SL. Quantum-inspired evolutionary
resource aggregators. IEEE Trans Smart Grid 2018;9:57–66. https://doi.org/ algorithm approach for unit commitment. IEEE Trans Power Syst 2009;24:
10.1109/TSG.2016.2544939. 1503–12. https://doi.org/10.1109/TPWRS.2009.2021220.
[97] Nikzad M, Samimi A. Integration of designing price-based demand response [122] Canizes B, Soares J, Lezama F, Vale Z. Complex large-scale energy resource
models into a stochastic bi-level scheduling of multiple energy carrier microgrids management optimization considering demand flexibility. 2020 IEEE Congr Evol
considering energy storage systems. Appl Energy 2021. https://doi.org/10.1016/ Comput CEC 2020 - Conf Proc 2020. https://doi.org/10.1109/
j.apenergy.2020.116163. CEC48606.2020.9185617.
[98] Anand H, Narang N, Dhillon JS. Multi-objective combined heat and power unit [123] Dhaliwal JS, Dhillon JS. Profit based unit commitment using memetic binary
commitment using particle swarm optimization. Energy 2019;172:794–807. differential evolution algorithm. Appl Soft Comput J 2019;81:105502. https://
https://doi.org/10.1016/j.energy.2019.01.155. doi.org/10.1016/j.asoc.2019.105502.
[99] Hossain MA, Pota HR, Squartini S, Abdou AF. Modified PSO algorithm for real- [124] Li YF, Pedroni N, Zio E. A memetic evolutionary multi-objective optimization
time energy management in grid-connected microgrids. Renew Energy 2019;136: method for environmental power unit commitment. IEEE Trans Power Syst 2013;
746–57. https://doi.org/10.1016/j.renene.2019.01.005. 28:2660–9. https://doi.org/10.1109/TPWRS.2013.2241795.
[100] Jeong YW, Park JB, Jang SH, Lee KY. A new quantum-inspired binary PSO: [125] Damousis IG, Bakirtzis AG, Dokopoulos PS. A solution to the unit-commitment
application to unit commitment problems for power systems. IEEE Trans Power problem using integer-coded genetic algorithm. IEEE Trans Power Syst 2004;19:
Syst 2010;25:1486–95. https://doi.org/10.1109/TPWRS.2010.2042472. 1165–72. https://doi.org/10.1109/TPWRS.2003.821625.
[101] Faria P, Soares J, Vale Z, Morais H, Sousa T. Modified particle swarm [126] Rasheed MB, Qureshi MA, Javaid N, Alquthami T. Dynamic pricing mechanism
optimization applied to integrated demand response and DG resources with the integration of renewable energy source in smart grid. IEEE Access 2020;
scheduling. IEEE Trans Smart Grid 2013;4:606–16. https://doi.org/10.1109/ 8:16876–92. https://doi.org/10.1109/ACCESS.2020.2967798.
TSG.2012.2235866. [127] Lujano-Rojas JM, Dufo-López R, Bernal-Agustín JL, Catalão JPS. Optimizing daily
[102] Roy K, Krishna Mandal K, Chandra Mandal A, Narayan Patra S. Analysis of energy operation of battery energy storage systems under real-time pricing schemes.
management in micro grid – a hybrid BFOA and ANN approach. Renew Sustain IEEE Trans Smart Grid 2017;8:316–30. https://doi.org/10.1109/
Energy Rev 2018. https://doi.org/10.1016/j.rser.2017.07.037. TSG.2016.2602268.
[103] Kanwar N, Gupta N, Niazi KR, Swarnkar A. Impact of optimal scheduling of DRs [128] Furukakoi M, Adewuyi OB, Matayoshi H, Howlader AM, Senjyu T. Multi objective
and network reconfiguration on the performance of active distribution systems. unit commitment with voltage stability and PV uncertainty. Appl Energy 2018;
In: India int. Conf. Power electron. IICPE. vol. 2018. IEEE Computer Society; 228:618–23. https://doi.org/10.1016/j.apenergy.2018.06.074.
2018. https://doi.org/10.1109/IICPE.2018.8709534. [129] Ramli MAM, Bouchekara HREH. Solving the problem of large-scale optimal
[104] Roy K, Mandal KK, Mandal AC. Ant-Lion Optimizer algorithm and recurrent scheduling of distributed energy resources in smart grids using an improved
neural network for energy management of micro grid connected system. Energy variable neighborhood search. IEEE Access 2020. https://doi.org/10.1109/
2019;167:402–16. https://doi.org/10.1016/J.ENERGY.2018.10.153. ACCESS.2020.2986895.
[105] Jha R, Meena NK, Swarnkar A, Gupta N, Niazi KR. Dynamic economic dispatch of [130] Murty VVVSN, Kumar A. Optimal DG integration and network reconfiguration in
micro-grid using harmony search algorithm. In: 12th IEEE int. Conf. Electron. microgrid system with realistic time varying load model using hybrid
Energy, environ. Commun. Comput. Control (E3-C3). Institute of Electrical and optimisation. IET Smart Grid 2019;2:192–202. https://doi.org/10.1049/iet-
Electronics Engineers Inc.; 2016. https://doi.org/10.1109/ stg.2018.0146.
INDICON.2015.7443330. INDICON 2015. [131] Ben Hamida I, Salah SB, Msahli F, Mimouni MF. Optimal network reconfiguration
[106] Moghaddam MM, Marzband M, Parhizi N. Optimal energy scheduling for a grid and renewable DG integration considering time sequence variation in load and
connected Microgrid based on Multi-period Imperialist competition algorithm. In: DGs. Renew Energy 2018;121:66–80. https://doi.org/10.1016/j.
Pecon 2016 - 2016 IEEE 6th int. Conf. Power energy, conf. Proceeding; 2017. renene.2017.12.106.
p. 44–9. https://doi.org/10.1109/PECON.2016.7951470. [132] Dahalan WM, Mokhlis H. Network reconfiguration for loss reduction with
[107] Ebrahimi J, Hosseinian SH, Gharehpetian GB. Unit commitment problem solution distributed generations using PSO. PECon 2012 - 2012. IEEE Int Conf Power
using shuffled frog leaping algorithm. IEEE Trans Power Syst 2011;26:573–81. Energy 2012. https://doi.org/10.1109/PECon.2012.6450331.
[108] Zeng X, Nazir MS, Khaksar M, Nishihara K, Tao H. A day-ahead economic [133] Sedghi M, Ahmadian A, Aliakbar-Golkar M. Assessment of optimization
scheduling of microgrids equipped with plug-in hybrid electric vehicles using algorithms capability in distribution network planning: review, comparison and
modified shuffled frog leaping algorithm. J Energy Storage 2021. https://doi.org/ modification techniques. Renew Sustain Energy Rev 2016. https://doi.org/
10.1016/j.est.2020.102021. 10.1016/j.rser.2016.08.027.
[109] Chandrasekaran K, Simon SP. Optimal deviation based firefly algorithm tuned [134] Gazijahani FS, Salehi J. Stochastic multi-objective framework for optimal
fuzzy design for multi-objective UCP. IEEE Trans Power Syst 2013;28:460–71. dynamic planning of interconnected microgrids. IET Renew Power Gener 2017;
https://doi.org/10.1109/TPWRS.2012.2201963. 11:1749–59. https://doi.org/10.1049/iet-rpg.2017.0278.
[110] Rameshkumar J, Ganesan S, Abirami M, Subramanian S. Cost, emission and [135] Eltamaly AM, Alotaibi MA, Alolah AI, Ahmed MA. A novel demand response
reserve pondered pre-dispatch of thermal power generating units coordinated strategy for sizing of hybrid energy system with smart grid concepts. IEEE Access
with real coded grey wolf optimisation. IET Gener, Transm Distrib 2016;10: 2021. https://doi.org/10.1109/ACCESS.2021.3052128.
972–85. https://doi.org/10.1049/iet-gtd.2015.0726. [136] Karimyan P, Gharehpetian GB, Abedi M, Gavili A. Long term scheduling for
[111] Mahmoud K, Hussein MM, Abdel-Nasser M, Lehtonen M. Optimal voltage control optimal allocation and sizing of DG unit considering load variations and DG type.
in distribution systems with intermittent PV using multiobjective grey-wolf-lévy Int J Electr Power Energy Syst 2014;54:277–87. https://doi.org/10.1016/j.
optimizer. IEEE Syst J 2020;14:760–70. https://doi.org/10.1109/ ijepes.2013.07.016.
JSYST.2019.2931829. [137] Gutiérrez-Alcaraz G, Tovar-Hernández JH. Two-stage heuristic methodology for
[112] Wang B, Wang S, Zhou X, Watada J. Multi-objective unit commitment with wind optimal reconfiguration and Volt/VAr control in the operation of electrical
penetration and emission concerns under stochastic and fuzzy uncertainties. distribution systems. IET Gener, Transm Distrib 2017;11:3946–54. https://doi.
Energy 2016;111:18–31. https://doi.org/10.1016/j.energy.2016.05.029. org/10.1049/iet-gtd.2016.1870.
[113] Luo F, Dong ZY, Chen Y, Zhao J. Natural aggregation algorithm: a new efficient [138] Jena S, Chauhan S. Solving distribution feeder reconfiguration and concurrent dg
metaheuristic tool for power system optimizations. In: 2016 IEEE int. Conf. Smart installation problems for power loss minimization by multi swarm cooperative
grid commun. SmartGridComm 2016. Institute of Electrical and Electronics PSO algorithm. Proc IEEE Power Eng Soc Trans Distrib Conf 2016. https://doi.
Engineers Inc.; 2016. p. 186–92. https://doi.org/10.1109/ org/10.1109/TDC.2016.7520021. 2016- July.
SmartGridComm.2016.7778759. [139] Mukhopadhyay B, Das D. Multi-objective dynamic and static reconfiguration with
[114] Ramli MAM, Bouchekara HREH, Alghamdi AS. Efficient energy management in a optimized allocation of PV-DG and battery energy storage system. Renew Sustain
microgrid with intermittent renewable energy and storage sources. Sustainability Energy Rev 2020. https://doi.org/10.1016/j.rser.2020.109777.
2019;11:1–28. [140] Atteya II, Ashour H, Fahmi N, Strickland D. Radial distribution network
[115] Juste KA. An evolutionary programming solution to the unit commitment reconfiguration for power losses reduction using a modified particle swarm
problem. IEEE Trans Power Syst 1999;14:1452–9. https://doi.org/10.1109/ optimisation. CIRED - Open Access Proc J 2017;2017:2505–8. https://doi.org/
59.801925. 10.1049/oap-cired.2017.1286. Institution of Engineering and Technology.
[116] Logenthiran T, Srinivasan D, Shun TZ. Demand side management in smart grid [141] Tandon A, Saxena D. A comparative analysis of SPSO and BPSO for power loss
using heuristic optimization. IEEE Trans Smart Grid 2012;3:1244–52. https://doi. minimization in distribution system using network reconfiguration. In: Proc. Int.
org/10.1109/TSG.2012.2195686. Conf. Innov. Appl. Comput. Intell. Power, energy control. With their impact
[117] Salinas S, Li M, Li P. Multi-objective optimal energy consumption scheduling in humanit. CIPECH 2014; 2014. p. 226–32. https://doi.org/10.1109/
smart grids. IEEE Trans Smart Grid 2013;4:341–8. https://doi.org/10.1109/ CIPECH.2014.7019093.
TSG.2012.2214068. [142] Farh HMH, Al-Shaalan AM, Eltamaly AM, Al-Shamma’A AA. A novel crow search
[118] Xie P, Cai Z, Liu P, Li X, Zhang Y, Xu D. Microgrid system energy storage capacity algorithm auto-drive PSO for optimal allocation and sizing of renewable
optimization considering multiple time scale uncertainty coupling. IEEE Trans distributed generation. IEEE Access 2020;8:27807–20. https://doi.org/10.1109/
Smart Grid 2018:1. https://doi.org/10.1109/TSG.2018.2879520. ACCESS.2020.2968462.
[119] Peng C, Xie P, Pan L, Yu R. Flexible robust optimization dispatch for hybrid wind/ [143] Chang GW, Cong Chinh N. Coyote optimization algorithm-based approach for
photovoltaic/hydro/thermal power system. IEEE Trans Smart Grid 2015:1–12. strategic planning of photovoltaic distributed generation. IEEE Access 2020;8:
[120] Chung CY, Yu H, Wong KP. An advanced quantum-inspired evolutionary 36180–90. https://doi.org/10.1109/ACCESS.2020.2975107.
algorithm for unit commitment. IEEE Trans Power Syst 2011;26:847–54. https:// [144] Gazijahani FS, Salehi J. Robust design of microgrids with reconfigurable topology
doi.org/10.1109/TPWRS.2010.2059716. under severe uncertainty. IEEE Trans Sustain Energy 2018;9:559–69. https://doi.
org/10.1109/TSTE.2017.2748882.

16
M. Papadimitrakis et al. Renewable and Sustainable Energy Reviews 145 (2021) 111072

[145] Daneshvar M, Babaei E. Exchange market algorithm for multiple DG placement [167] De Oliveira LW, Seta FDS, De Oliveira EJ. Optimal reconfiguration of distribution
and sizing in a radial distribution system. J Energy Manag Technol 2018;2:54–65. systems with representation of uncertainties through interval analysis. Int J Electr
https://doi.org/10.22109/JEMT.2018.116625.1059. Power Energy Syst 2016;83:382–91. https://doi.org/10.1016/j.
[146] Kumar BV, Srikanth NV. A hybrid approach for optimal location and capacity of ijepes.2016.04.020.
UPFC to improve the dynamic stability of the power system. Appl Soft Comput J [168] Scheidler A, Thurner L, Braun M. Heuristic optimisation for automated
2017;52:974–86. https://doi.org/10.1016/j.asoc.2016.09.031. distribution system planning in network integration studies. IET renew. Power
[147] Tan WS, Hassan MY, Majid MS, Rahman HA. Allocation and sizing of DG using gener., vol. 12. Institution of Engineering and Technology; 2018. p. 530–8.
Cuckoo Search algorithm. In: PECon 2012 - 2012 IEEE int. Conf. Power energy; https://doi.org/10.1049/iet-rpg.2017.0394.
2012. https://doi.org/10.1109/PECon.2012.6450192. [169] Papalexopoulos AD, Imparato CF, Wu FF. Large-scale optimal power flow: effects
[148] Ali ES, Abd Elazim SM, Abdelaziz AY. Optimal allocation and sizing of renewable of initialization, decoupling & discretization. IEEE Power Eng Rev 1989;9:72–4.
distributed generation using ant lion optimization algorithm. Electr Eng 2018; https://doi.org/10.1109/MPER.1989.4310717.
100:99–109. https://doi.org/10.1007/s00202-016-0477-z. [170] Zohrizadeh F, Josz C, Jin M, Madani R, Lavaei J, Sojoudi S. A survey on conic
[149] Nekooei K, Farsangi MM, Nezamabadi-Pour H, Lee KY. An improved multi- relaxations of optimal power flow problem. Eur J Oper Res 2020. https://doi.org/
objective harmony search for optimal placement of DGs in distribution systems. 10.1016/j.ejor.2020.01.034.
IEEE Trans Smart Grid 2013;4:557–67. https://doi.org/10.1109/ [171] Jordehi AR. How to deal with uncertainties in electric power systems? A review.
TSG.2012.2237420. Renew Sustain Energy Rev 2018. https://doi.org/10.1016/j.rser.2018.07.056.
[150] Jordehi AR. Optimal allocation of FACTS devices for static security enhancement [172] Jordehi AR. A review on constraint handling strategies in particle swarm
in power systems via imperialistic competitive algorithm (ICA). Appl Soft Comput optimisation. Neural Comput Appl 2015. https://doi.org/10.1007/s00521-014-
J 2016. https://doi.org/10.1016/j.asoc.2016.07.014. 1808-5.
[151] Zheng Y, Meng K, Luo F, Qiu J, Zhao J. Optimal integration of MBESSs/SBESSs in [173] Clerc M, Kennedy J. The particle swarm-explosion, stability, and convergence in a
distribution systems with renewables. IET Renew Power Gener 2018;12:1172–9. multidimensional complex space. IEEE Trans Evol Comput 2002;6:58–73. https://
https://doi.org/10.1049/iet-rpg.2017.0236. doi.org/10.1109/4235.985692.
[152] Xiao H, Pei W, Dong Z, Kong L. Bi-level planning for integrated energy systems [174] Yang XS. Metaheuristic optimization: algorithm analysis and open problems. Lect
incorporating demand response and energy storage under uncertain Notes Comput Sci (including Subser Lect Notes Artif Intell Lect Notes
environments using novel metamodel. CSEE J Power Energy Syst 2018;4:155–67. Bioinformatics) 2011;6630:21–32. https://doi.org/10.1007/978-3-642-20662-7_
https://doi.org/10.17775/cseejpes.2017.01260. 2. LNCS.
[153] Daryani N, Zare K, Tohidi S, Guerrero JM. Dominated GSO algorithm for optimal [175] Yassin IM, Taib MN, Adnan R, Salleh MKM, Hamzah MK. Effect of swarm size
microgrid construction to improve consumer side properties in a distribution parameter on Binary Particle Swarm optimization-based NARX structure
system. Int J Electr Power Energy Syst 2020. https://doi.org/10.1016/j. selection. ISIEA 2012 - 2012 IEEE Symp Ind Electron Appl 2012:219–23. https://
ijepes.2020.106232. doi.org/10.1109/ISIEA.2012.6496632.
[154] Nayak MR, Behura D, Kasturi K. Optimal allocation of energy storage system and [176] Kanwar N, Gupta N, Niazi KR, Swarnkar A. Optimal allocation of DGs and
its benefit analysis for unbalanced distribution network with wind generation. reconfiguration of radial distribution systems using an intelligent search-based
J Comput Sci 2021;51:101319. https://doi.org/10.1016/j.jocs.2021.101319. TLBO. Elec Power Compon Syst 2017;45:476–90. https://doi.org/10.1080/
[155] Belmino LM, Souto Soares F, Sampaio RF, Pastora Saraiva Leão R, De Sousa 15325008.2016.1266714.
Braga AP, Silveira Melo L, et al. Placement and sizing of distributed generation in [177] Rezaee Jordehi A. Optimal setting of TCSCs in power systems using
distribution system. In: 2019 IEEE PES conf. Innov. Smart grid technol. ISGT lat. teaching–learning-based optimisation algorithm. Neural Comput Appl 2015.
Am. 2019. Institute of Electrical and Electronics Engineers (IEEE); 2019. p. 1–6. https://doi.org/10.1007/s00521-014-1791-x.
https://doi.org/10.1109/ISGT-LA.2019.8894981. [178] Skakov ES, Malysh VN. Parameter meta-optimization of metaheuristics of solving
[156] Farag HEZ, El-Saadany EF. Optimum shunt capacitor placement in distribution specific NP-hard facility location problem. J Phys Conf Ser 2018;973. https://doi.
networks with high penetration of renewable energy resources using genetic org/10.1088/1742-6596/973/1/012063.
algorithms. In: IEEE PES innov. Smart grid technol. Conf. Eur., vol. 2015. IEEE [179] Kamel S, Selim A, Jurado F, Yu J, Xie K, Yu C. Multi-objective whale optimization
Computer Society; 2015. https://doi.org/10.1109/ISGTEurope.2014.7028929. algorithm for optimal integration of multiple DGs into distribution systems. In:
Janua. 2019 IEEE PES innov. Smart grid technol. Asia, ISGT 2019; 2019. p. 1312–7.
[157] Cavalheiro EMB, Vergílio AHB, Lyra C. Optimal configuration of power https://doi.org/10.1109/ISGT-Asia.2019.8881761.
distribution networks with variable renewable energy resources. Comput Oper [180] Debnath KB, Mourshed M. Forecasting methods in energy planning models.
Res 2018;96:272–80. https://doi.org/10.1016/j.cor.2017.09.021. Renew Sustain Energy Rev 2018;88:297–325. https://doi.org/10.1016/j.
[158] Chittur Ramaswamy P, Deconinck G. Smart grid reconfiguration using simple rser.2018.02.002.
genetic algorithm and NSGA-II. IEEE PES Innov Smart Grid Technol Conf Eur [181] Karamichailidou D, Kaloutsa V, Alexandridis A. Wind turbine power curve
2012. https://doi.org/10.1109/ISGTEurope.2012.6465615. modeling using radial basis function neural networks and tabu search. Renew
[159] Agrawal P, Kanwar N, Gupta N, Niazi KR, Swarnkar A. Resiliency in active Energy 2021. https://doi.org/10.1016/j.renene.2020.10.020.
distribution systems via network reconfiguration. Sustain Energy, Grids Networks [182] Alba E, Luque G, Nesmachnow S. Parallel metaheuristics: recent advances and
2021. https://doi.org/10.1016/j.segan.2021.100434. new trends. Int Trans Oper Res 2013;20:1–48. https://doi.org/10.1111/j.1475-
[160] Das S, Das D, Patra A. Reconfiguration of distribution networks with optimal 3995.2012.00862.x.
placement of distributed generations in the presence of remote voltage controlled [183] Alexandridis A, Chondrodima E, Sarimveis H. Cooperative learning for radial
bus. Renew Sustain Energy Rev 2017. https://doi.org/10.1016/j. basis function networks using particle swarm optimization. Appl Soft Comput J
rser.2017.01.055. 2016;49:485–97. https://doi.org/10.1016/j.asoc.2016.08.032.
[161] Das S, Das D, Patra A. Operation of distribution network with optimal placement [184] Sun H, Guo Q, Qi J, Ajjarapu V, Bravo R, Chow J, et al. Review of challenges and
and sizing of dispatchable DGs and shunt capacitors. Renew Sustain Energy Rev research opportunities for voltage control in smart grids. IEEE Trans Power Syst
2019. https://doi.org/10.1016/j.rser.2019.06.026. 2019;34:2790–801.
[162] Li B, Roche R, Paire D, Miraoui A. Optimal sizing of distributed generation in gas/ [185] Chatzigiannis DI, Dourbois GA, Biskas PN, Bakirtzis AG. European day-ahead
electricity/heat supply networks. Energy 2018;151:675–88. https://doi.org/ electricity market clearing model. Elec Power Syst Res 2016;140:225–39. https://
10.1016/j.energy.2018.03.080. doi.org/10.1016/j.epsr.2016.06.019.
[163] Sattarpour T, Nazarpour D, Golshannavaz S, Siano P. A multi-objective hybrid GA [186] Csercsik D, Sleisz A, Sores PM. Increasing the flexibility of European type
and TOPSIS approach for sizing and siting of DG and RTU in smart distribution electricity auctions via a novel bid class. Int. Conf Eur Energy Mark EEM 2019;
grids. J Ambient Intell Humaniz Comput 2018;9:105–22. https://doi.org/ 2019. https://doi.org/10.1109/EEM.2019.8916306. Septe.
10.1007/s12652-016-0418-8. [187] Divenyi D, Polgari B, Raisz D, Sleisz A, Sores P. Special session on proposal of a
[164] Almehizia AA, Al-Masri HMK, Ehsani M. Integration of renewable energy sources new European co-optimized energy and ancillary service market design - part i - t.
by load shifting and utilizing value storage. IEEE Trans Smart Grid 2018:1. Int. Conf Eur Energy Mark EEM 2016. https://doi.org/10.1109/
https://doi.org/10.1109/TSG.2018.2871806. EEM.2016.7521363. 2016- July.
[165] Lim KZ, Lim KH, Wee X Bin, Li Y, Wang X. Optimal allocation of energy storage [188] Bedoya JC, Liu CC, Krishnamoorthy G, Dubey A. Bilateral electricity market in a
and solar photovoltaic systems with residential demand scheduling. Appl Energy distribution system environment. IEEE Trans Smart Grid 2019;10:6701–13.
2020. https://doi.org/10.1016/j.apenergy.2020.115116. https://doi.org/10.1109/TSG.2019.2910216.
[166] De Oliveira LW, De Oliveira EJ, Gomes FV, Silva IC, Marcato ALM, Resende PVC. [189] Gu Y, Jiang H, Zhang JJ, Zhang Y, Wu H, Muljadi E. Multi-timescale three-phase
Artificial Immune Systems applied to the reconfiguration of electrical power unbalanced distribution system operation with variable renewable generations.
distribution networks for energy loss minimization. Int J Electr Power Energy Syst IEEE Trans Smart Grid 2019;10:4497–507. https://doi.org/10.1109/
2014;56:64–74. https://doi.org/10.1016/j.ijepes.2013.11.008. TSG.2018.2861736.

17