Selecting A Meta-Heuristic Technique For Smart Micro-Grid Optimization Problem: A Comprehensive Analysis
Selecting A Meta-Heuristic Technique For Smart Micro-Grid Optimization Problem: A Comprehensive Analysis
Selecting A Meta-Heuristic Technique For Smart Micro-Grid Optimization Problem: A Comprehensive Analysis
net/publication/318515916
CITATIONS READS
19 308
2 authors:
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Baseem Khan on 06 August 2017.
ABSTRACT In current epoch, the economic operation of micro-grid under soaring renewable energy
integration has become a major concern in the smart grid environment. There are several meta-heuristic
optimization techniques available under different categories in literature. One of the most difficult tasks
in cost minimization of micro-grid is to select the best suitable optimization technique. To resolve the
problem of selecting a suitable optimization technique, a rigorous review of six meta-heuristic algorithms
(Whale Optimization, Fire Fly, Particle Swarm Optimization, Differential Evaluation, Genetic Algorithm,
and Teaching Learning-based Optimization) selected from three categories (Swarm Intelligence,
Evolutionary Algorithms, and Teaching Learning) is conducted. It presents, a comparative analysis using
different performance indicators for standard benchmark functions and proposed a smart micro-grid (SMG)
operation cost minimization problem. A proposed SMG is modeled which incorporates utility connected
power resources, e.g., wind turbine, photovoltaic, fuel cell, micro-turbine, battery storage, electric vehicle
technology, and diesel power generator. The proposed work will help researchers and engineers to select
an appropriate optimization method to solve micro-grid optimization problems with constraints. This
paper concludes with a detailed review of micro-grid operation cost minimization techniques based on an
exhaustive survey and implementation.
INDEX TERMS Smart micro-grid, meta-heuristic optimization techniques, electric vehicle technology,
fuel cell.
2169-3536
2017 IEEE. Translations and content mining are permitted for academic research only.
VOLUME 5, 2017 Personal use is also permitted, but republication/redistribution requires IEEE permission. 13951
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
B. Khan, P. Singh: Selecting a Meta-Heuristic Technique for Smart Micro-Grid Optimization Problem
NOMENCLATURE SUC MT ,t , SUC FC,t , start up cost for MT, FC, and
Bid BES,t , Bid grid,t , Bid of BES, utility, FC, MT, SUC FCEV ,t FCEV at time step t,
Bid FC,t , Bid MT ,t , WT, PV, and FCEV at time correspondingly in ACct
Bid iWT ,t , Bid iPV ,t , step t, correspondingly in AC TCPDBEV , TCPDBES Per day overall cost of BEV, BES
Bid FCEV ,t ct/kW h TCPDPHEV and PHEV, correspondingly
MC BES , FX BES Repair and constant cost for in ACct
MC BEV , FX BEV BES, BEV and PHEV, uMT ,t , uBES,t , uBEV ,t On/Off status of MT, BES, BEV,
MC PHEV , FX PHEV correspondingly in A Cct/kW h uFC,t , uPHEV ,t FC, PHEV and DiG attime
IR Interest-rate for battery uDiG,t step t, correspondingly
installation on loan t tth time step (h)
LT Life span of the batteries in years
NT Operation duration in hours
ORt Required generating backup
I. INTRODUCTION
minutes in kW
As world transit from the conventional grid system to the
OM DG Constant repair and operation
smart grid system, renewable energy sources’ incorpora-
cost of DGs in A Cct
tion has become the key issue in the present environment.
OM MT , OM FC , Constant repair and operation
In accordance with the International Energy Agency predic-
OM iWT , OM iPV , cost of MT, FC, WT, PV, FCEV
tion, power production by renewable power resources will
OM FCEV correspondingly in A Cct/kW h
be almost three times in between 2010 to 2035. It will
Pgrid,max , Pgrid,min Maximum and minimum
contribute 31% of the globe’s entire power production, in
PBES,max , PBES,min generation of power for utility,
which solar, wind and hydro will provide 7.5%, 25% and
PBEV ,max , PBEV ,min BES, BEV, PHEV, FCEV, FC
50% respectively, of the overall renewable power produc-
PPHEV ,max , PPHEV ,min and MT in (kW)
tion by 2035. The intermittency and climate dependency of
PFCEV ,max , PFCEV ,min
renewable power resources make their interconnection more
PFC,max , PFC,min
complex and difficult. Various energy storage devices are
PMT ,max , PMT ,min
used to solve above mentioned problems of intermittency
PDemand,t Power demand at time step t
and weather dependency with renewable. Hence inside the
in kW
smart grid environment, the development of micro grid is a
Cf Diesel Fuel Price (ACct/l)
great solution for integration of renewable energy sources.
PDiG Power output of diesel generator
It has numerous advantages such as energy loss-reduction,
tax Rate of tax for grid
reliability and enhancement of energy management.
ηcharge , ηdischarge Charging and discharging
Micro-grid consists of different renewable power resources
efficiency of different batteries like wind, SPV and micro turbines. It also incorporates latest
CBES,max , CBES,min Maximum and minimum generation technologies such as fuel cell technologies and
capacity of BES, combined heat and power (CHP) technology. To solve the
CBEV ,max , C BEV ,min BEV and PHEV in kW h above-discussed intermittency problem of renewable energy,
CPHEV ,max , C PHEV ,min storage devices, for example, battery energy storage system,
CBES,t CBEV ,t Stored energy in BES, BEV and electric vehicle technology and flywheel storage system can
CPHEV ,t PHEV be used. Micro-gird provides a better solution as compared to
Cost grid,t Supply cost with the grid in the distributed generation sources due to their better coordi-
upstream mode at time step t in nation and control. It can be used as islanded mode and gird
A
Cct connected mode as per requirement. Hence inside micro-grid,
Cost DG,t , Cost BEV ,t Cost of operation and fuel of the operation, control and coordination problem are of great
Cost BES,t , Cost PHEV ,t DGs, BEV, BES, PHEV and DiG importance. Further, similar to the conventional grid, micro-
Cost DiG,t at time step t, correspondingly grid also required some cost which is related to its generation,
in A
Cct maintenance and operation; consequently, many researches
C Overall cost in ACct are focused on the micro-grid cost minimization problem.
Pgrid,t , PMT ,t , PBES,t Generated power of grid, MT, Various meta-heuristic techniques are developed by
BES, PV, researchers to solve the micro-grid cost minimization
PiPV ,t , PFC,t , PFCEV ,t , FC, FCEV,WT, PHEV, problem. Population dependent meta-heuristic optimization
PiWT ,t , PPHEV ,t BEV and DiG correspondingly techniques have two main classifications: swarm intelli-
PBEV ,t , PDiG,t in kW gence (SI) and evolutionary algorithms (EA). A number
PBEV ,t ; PBEV ,t Maximum discharging and of renowned evolutionary techniques are as follows: Evo-
PBES,t ; PBES,t charging rates of BEV, BES lution Strategy (ES), Genetic Algorithm (GA), Differential
PPHEV ,t ; PPHEV ,t and PHEV at time t, Evolution (DE), Evolution Programming (EP), etc. Various
correspondingly in kW swarm intelligence dependent techniques are as follows:
13952 VOLUME 5, 2017
B. Khan, P. Singh: Selecting a Meta-Heuristic Technique for Smart Micro-Grid Optimization Problem
Firefly (FF), Shuffled Frog Leaping (SFL), Particle Swarm TABLE 1. List of meta-heuristic algorithms (1975-2016) [79].
Optimization (PSO), Artificial Bee Colony (ABC), Ant
Colony Optimization (ACO), etc. In addition to above
mentioned meta-heuristic optimization techniques, different
natural phenomena based methods are also available, e.g.,
Gravitational Search (GSA), Harmony Search (HS) algo-
rithm, Flower Pollination (FPA), Biogeography-Based Opti-
mization (BBO), etc.
Meta-heuristic techniques need no gradient information.
Meta-heuristics have the capability to recover from local
optima due to their inherent stochasticity; consequently, it
can better tackle uncertainties in objectives. It can tackle
multiple objectives with only a few algorithmic changes.
Normally, meta-heuristics techniques are probabilistic in
nature and controlled by common parameters, e.g., popula-
tion, elite population size, the number of generations, etc.
In addition to these parameters, different methods require
specific control parameters, e.g., GA utilizes the proba-
bility of mutation and crossover, operator selection, etc;
PSO employs weight of inertia, cognitive and social fac-
tors; ABC makes use of the number of different type of
bees i.e. onlooker, employed, scout and their limits. In the
same way, different techniques require separate tuning of
their specific parameters. In these techniques, parameter’s
tuning is an extremely critical issue, because it’s directly
affects the performance of techniques. Improper tuning may
result in increased computation time or local optima. Table 1
presented the time line for the evaluation of different meta-
heuristic techniques.
In this paper, the authors have focused on the problem of
choosing the appropriate meta-heuristic optimization tech-
nique for the minimization of operation cost. To identify
the best algorithm, different meta-heuristic techniques (PSO,
GA, FF, DE, TLBO and WO) from three different categories
(swarm intelligence, evolutionary algorithms and teaching
learning) are considered and compared. These optimiza-
tion methods are compared by using nineteen standard test
functions from three different categories (uni-model, multi-
model, composite) as well as a micro-grid frame work. MG
consists of different energy sources such as PV, WT, BES,
MT, FC, DiG and EVT. The problem of micro-grid operation
cost minimization is solved for two different cases (charging
mode of batteries and discharging mode of batteries) using
these optimization techniques. The comparison of different
optimization methods for various standard test functions is
accomplished using the following parameters: mean and stan-
dard deviation of fitness value, average fitness value for
different population, convergence characteristics for different
population, trajectory (fitness value of the first search agents increase in population size and power generated by different
with respect to the number of iterations for any algorithm) energy sources for different population size.
and the capability to explore the search space. In comparative The structure of the paper is as follows. The next section
analysis of SMG, two different cases are considered and provides the literature review. In section 3, the problem is
each case is observed for the following parameters: average, formulated along with its constraint. Section 4 describes a
best, worst and standard deviation of optimized operation set of nineteen standard benchmark functions from three
cost (fitness value), convergence characteristics for different different categories. The proposed SMG system and the two
population, variation of best fitness value with respect to specific cases are described in the section 5. The comparative
analysis of different optimization methods using standard long term storages and ultra-capacitors as short-term storages
test functions and proposed SMG system is provided in the for the frequency control of utility connected micro-grid.
section 6, which is followed by the concluding remarks. Alharbi and Bhattacharya [10] developed a model to calculate
the optimal power rating and energy capacity of BES for
II. LITERATURE REVIEW coordinated operation of micro-grid. Graditi et al. [80] carried
Micro-grid optimization problem is a complex and real world out the technical and economical evaluation for installing
problem. Generally, micro-grid is the combination of various different types of battery technologies to lower the electricity
renewable energy sources (solar, wind) along with energy cost for a customer-side application. Ippolito et al. [81]
storage system (BES, EVT) and diesel generator. Mathemat- developed a bidirectional converter to connect and control the
ical methods such as linear and integer programming are utility grid with renewable energy sources and battery storage
cumbersome and require more time to provide the optimal systems. Silvestre et al. [82] presented a multi-objective gen-
solution of real world problem, whereas meta-heuristic tech- eralized framework for optimal sizing of distributed energy
niques provide the optimal solutions for practical problems resources in micro-grids by using an indicator based swarm
in less time. Thus, this section presents the comprehensive approach. Graditi et al. [83] presented an optimal energy
review of micro-grid system with different technologies such dispatch problem, which is having directly controlled and
as battery energy storage, electric vehicle technology, and shiftable loads. It is solved by glow worm swarm particles
diesel generator. Additionally, it also focuses on the current optimization algorithm. Takeuchi et al. [84] described an
development in utilization of GA, PSO, DE, TLBO, FF and optimal scheduling methodology to determine the operating
WO meta-heuristic algorithms in power system optimization schedule of an energy network for minimizing CO2 emis-
problems. sion and energy costs. Favuzza et al. [85] applied an ant
colony search for electrical distribution systems management
A. MICRO-GRID TECHNOLOGIES problem. Gamarra and Guerrero [86] reviewed the technical
There are various technologies that can be incorporated in literature about optimization techniques applied to micro-grid
MG to enhance the performance and stability of the MG sys- planning.
tem. These technologies are as follows:
cost management. For confirming the precision and validity an upgraded PSO technique with adaptive weight and accel-
of the mathematical modelling of a new environmental and eration coefficients for solving the economic, environmental
economic dispatch of SMG, Liao et al. [39] used the quantum and health dispatch model of a micro-grid. An economic
genetic algorithm. Changsong et al. [40] proposed a novel operation of micro-grid under an uncertain framework is
micro-grid power trading model to find out an optimum studied by Liang et al. [51]. A micro-grid scheduling strategy
schedule for all available units over a planning horizon. is developed on the basis of Roulette Wheel Mechanism and
The METM utilized genetic algorithm (GA) to assist the Probability Density Functions. Further, PSO is used to find
micro-grid scheduling. A control scheme is presented by the optimum solution. Yang et al. [52] proposed an enhanced
Zolfaghari et al. [41] to enhance the load sharing among PSO technique for HOME-EMS, which incorporates load
inverter-based DGs in MG. GA tuned proportional-integral response in a smart grid. Hao et al. [53] studied a distinctive
(PI) controller is used for this purpose. Shi et al. [42] pro- utility connected MG, which contains WT, hydro power,
posed a multi-objective optimization problem for construc- BES and local demand. First its mathematical algorithm is
tion of energy sources, storage and interruptible load in MG. constructed and then PSO is applied to solve the optimal oper-
The problem is solved by employing improved NSGA-II. ation problem. An et al. [54] proposed a novel operating cost
Deng et al. [43] studied and modelled a micro-grid including optimization method for a building with an integrated MG
a WT, PV and a CHP system with FCs and MT. GA is linked to the utility. Further, a piecemeal decision algorithm
used to solve the optimum model and an operation strat- and a PSO algorithm were utilized to produce a charging
egy. For cost effective and reliable micro-grid, Nasser and and discharging rate’s schemes for BESs. Elamine et al. [55]
Reji [44] proposed an optimum design scheme. For opti- presented a multi-agent structural design for SMG, which
mization, a hybrid Genetic Particle Swarm Optimization is is based on wind power forecast. It utilizes neural network,
utilized. Eldessouky and Gabbar [45] presented micro-grid which is trained by hybrid PSO and back-propagation tech-
(MG) optimization using GA. The algorithm’s aim is to find nique. Liang et al. [56] studied the multi-objective optimum
out the optimum size of combined wind and gas generator to scheduling problem in a vague structure. The probability
satisfy a given key performance indices (KPIs). Shariatzadeh distribution function and roulette wheel mechanism are used
et al. [46] applied GA and PSO for reconfiguration of SMPS. to develop the scenarios and multi-objective PSO is used to
coordinate between them.
2) PARTICLE SWARM OPTIMIZATION
The PSO consists of a population of particles, known as a 3) DIFFERENTIAL EVOLUTION
swarm, with each member of the swarm being associated with Storn and Price [90] proposed a robust and easily paralleliz-
a position vector xt and a velocity vector vt . The size of these able technique DE to solve the global optimization problems.
vectors is equal to the dimension of the search space. The DE is population based meta-heuristic technique, which starts
term velocity (vt ) at any iteration t indicates the directional with randomly initialized solution vectors. To modify an
distance that the particle has covered in the (t − 1)th iteration. existing solution in the population, DE utilized the differ-
The directional velocity of any particle is calculated on the ence vector of two randomly chosen members. The weight
basis of a particle personal best ‘pbest’ (pl ) and swarm’s of a difference vector is a user-defined constant parameter
global best ‘gbest’ (pg ) [89]. (F > 0):
The following equations describe the velocity and position
update for ith particle at any iteration t: vi,t+1 = x(r1 ,t) + F.(xr2 ,t − xr3 ,t )
planned scheduling for economic energy sharing in a CHP Veltman et al. [62] proposed a prediction interval mod-
based MG using DE technique. elling tuned by an improved TLBO for load forecasting in
MGs. Dixit and Roy [63] presented the impact of PEVs
4) TEACHER LEARNING BASED OPTIMIZATION (TLBO) on automatic generation control using the TLBO technique.
TLBO simulates the teaching-learning process of the class Yammani et al. [64] developed a modified TLBO technique
room. It is a population-based algorithm and does not require to find out the optimum placement and size of distributed
any algorithm-dependent parameters. The common control RESs units in the distribution network. Rani et al. [65] applied
parameters required by TLBO are the number of generations TLBO to resolve a multi-objective problem of the economic
and population size [91]. and emission scheduling.
The function of TLBO is categorised into two phases,
‘Teacher phase’ and ‘Learner phase’. 5) FIREFLY OPTIMIZATION
Firefly Algorithm (FF) is a nature inspired meta-heuristic
a: TEACHER PHASE optimization algorithm, which is motivated from the conduct
On the basis of his/her capabilities, teacher attempts to of fireflies [92]. FF depends on three basic rules:
enhance the mean performance of the group of students i) All fireflies are attracted to each other with disregard
in their concerned subject. At ith iteration, let there be n to gender.
number of learners, m number of subjects, and Mj,i be the ii) Attractiveness is correlated with brightness (light emis-
mean performance of the learners in a particular subject j. sion) such that bright flies attract less bright flies, and
Xtotal−kbest,i is the best overall performance by considering in absence of brighter flies they move randomly.
all the subjects together. kbest is the result of best learner. iii) The brightness is proportional to the objective function.
The learner having best performance is selected as a teacher. The movement of a firefly i is attracted to another more
The difference between the performance of each subject of attractive (brighter) firefly j is determined by
a teacher and current mean performance of corresponding 2
xi = xi + β0 e−γ rij xj − xi + αi
subject is formulated as:
Where β0 is attractiveness at distance zero, rij =
xi − xj
Difference_Meanj,k,i = ri (Xj,kbest,i − TF Mj,i )
is the distance between any two fireflies i and j at distance
Where, Xj,kbest,i is the best learner’s result in subject j. TF is xi and xj , respectively, i is a vector of random numbers
the teaching factor, and ri is the random number in the drawn from a Gaussian distribution or uniform distribution
range [0, 1]. The value of TF is given as, and α being the randomization parameter. The FF is applied
on power system optimization problems as follows:
TF = round[1 + rand(0, 1){2 − 1}] Odeim et al. [17] investigated the optimization of a power
In the teaching phase, the existing solution is updated as: management scheme of a battery/super-capacitor/fuel cell
hybrid vehicular system.
X 0 j,k,i = Xj,k,i + Difference_Meanj,k,i
6) WHALE OPTIMIZATION
Where, X 0 j,k,i is the updated value of Xj,k,i .
Whale Optimization Algorithm (WO) is a nature inspired
meta-heuristic optimization algorithm, which is encouraged
b: LEARNER PHASE
from the behaviour of Humpback whales [93]. The WO
In the learner phase, learners enhance their knowledge by
algorithm is working on the following rules:
interacting randomly with other learners. A less knowledge-
able learner learns new things from more knowledgeable
a: ENCIRCLING PREY
learners. For n population size, the learning process is as
Humpback whales can identify the position of prey and
follows:
encircle them. As per the activities of humpback whales the
P and Q learners are selected randomly such that
WO assumes that the current best solution is the objective
X 0 total−P,i 6 = X 0 total−Q,i , where, X 0 total−P,i and X 0 total−Q,i are
prey or near to the optimum. Other search agents will renew
the updated function values of Xtotal−P,i and Xtotal−Q,i
their position towards the best agent, as represented by the
of P and Q respectively at the end of teacher phase:
following equations:
X 00 j,P,i = X 0 j,P,i + ri (X 0 j,P,i − X 0 j,Q,i ), −
→ − →− → −
→
D = C .X ∗ (t) − X (t)
if X 0 total−P,i < X 0 total−Q,i
−
→ −
→ −
→− →
X 00 j,P,i = X 0 j,P,i + ri (X 0 j,Q,i − X 0 j,P,i ), X (t + 1) = X ∗ (t) − A . D
−
→
if X 0 total−Q,I < X 0 total−P,i A = 2−→a .−
→r −− →a
−
→ −
→
C = 2. r
X 00 j,P,I is accepted if it provides a better value of function.
−
→ −
→
The TLBO is applied on various power system optimization where t indicates the current iteration, A and C are coef-
problems as follows: ∗
ficient vectors, X is the position vector of the best solution
−
→
obtained so far, X is the position vector, || is the absolute ct is the summation of the supply cost of grid, operation and
value, and ‘•’ is an element-by-element multiplication, −
→a is fuel cost of DG, BES, BEV, PHEV, as well as start up cost of
linearly decreased from 2 to 0 over the course of iterations FC, MT and FCEV, as shown in
and −
→r is a random vector in [0, 1].
ct = Cost grid,t + Cost DG,t + Cost BES,t + Cost BEV ,t
b: BUBBLE-NET ATTACKING METHOD + Cost PHEV ,t + SUC FC,t + SUC MT ,t + SUC FCEV ,t
Humpback whales swim about the prey within a shrink- (2)
ing circle and along a spiral-shaped path, concurrently. The Supply cost of the grid is defined by
authors have assumed that there is a 50% probability (p) of
either the shrinking encircling mechanism or the spiral model Bid grid,t Pgrid,t
if Pgrid,t > 0
to update the location of whales through optimization. The Cost grid,t = (1 − tax) Bid grid,t Pgrid,t if Pgrid,t < 0
mathematical model is as follows:
0 if Pgrid,t = 0
−→ −
→− →
−
→ X ∗ (t) − A . D ifp < 0.5 (3)
X (t + 1) = − →
D0 .ebl . cos (2πl) + −
→
X ∗ (t) ifp ≥ 0.5 The operation and fuel cost of the distributed generators are
presented by
Where p is a random number in [0, 1], b is a constant to define
the shape of the logarithmic spiral and l is a random number Cost DG,t = Bid MT ,t PMT ,t uMT ,t + Bid FC,t PFC,t uFC,t
in [−1, 1]. + Bid FCEV ,t PFCEV ,t uFCEV ,t + Bid PVi,t PPVi,t
+ Bid WTi,t PWTi,t (4)
c: SEARCH FOR PREY
This method highlights exploration and allows the WO algo- The start up cost of FC, FCEV and MT are provided, respec-
rithm to achieve a global search. The mathematical model is tively, in
as follows: SUC FC,t = Start FC ∗ max 0, uFC,t − uFC,t−1
(5)
−
→ −
→ −−→ − →
D = | C .Xrand − X (t) |
SUC FCEV ,t = Start FCEV ∗ max 0, uFCEV ,t − uFCEV ,t−1
−
→ −−→ − →− →
X (t + 1) = Xrand − A . D (6)
−−→
Where Xrand is a random position vector. SUC MT ,t = Start MT ∗ max 0, uMT ,t − uMT ,t−1 (7)
The WO algorithm starts with a set of random solutions. The constant repair and operation cost of distributed genera-
In every iteration, the search agents revise their positions with tors are presented by
regard to either a randomly chosen search agent or the best
solution obtained so far. The WO is applied on power system OM DG = (OM MT + OM FC + OM FCEV + OM PVi
optimization problems as follows: + OM WTi ) ∗ NT (8)
Reddy et al. [68] utilized whale optimization algorithm
The overall charges of the SMG consist of the operation
(WOA) to determine the optimal distributed generation size.
charges of BES, BEV, PHEV, FCEV and utility, fuel and OM
Trivedi et al. [69] presented the solution of an emission
charges of DGs and DiGs, the start-up charges of FCEV, MT
constraint environment dispatch problem with MG by using
and FC in addition cumulative per day overall cost of batter-
whale optimization algorithm.
ies, which is used in BES, BEV and PHEV (CTCPD). The
cost of batteries contains the single-time constant cost (FX)
III. PROBLEM FORMULATION
and the yearly repair cost (MC). Overall cost of the battery is
Selecting a suitable method to solve the cost minimization
(FC + MC)∗Cmax , where Cmax is the battery’s size. The time
problem of SMG, the proposed problem is formulated as
window selected for this work is a day; therefore, the opera-
follows:
tion cost is computed over 24 hours and TCPD is required in
The formulation of the economic operation problem of
A
Cct/day. The TCPDs of mounted batteries in A Cct/day can be
SMG (EOSMG) is illustrated as follows:
achieved using the following equations [70]:
Minimization of total cost of SMG is given by Eq.1.
NT CBES,max IR (1+IR)LT
X TCPDBES = ( FC BES + MC BES )
MinC (X ) = ct + OM DG + CTCPD 365 (1+IR)LT − 1
t=1 (9)
N
X CBEV ,max IR (1 + IR)LT
+ Cf aP2DiG(i) + bPDiG(i) + c (1) TCPDBEV = ( FC BEV +MC BEV )
365 (1 + IR)LT − 1
i=1
(10)
where cumulative total cost per day for batteries is the sum- CPHEV ,max IR (1 + IR) LT
mation of total cost per day for BES, BEV and PHEV. TCPDPHEV = (
365 (1 + IR)LT − 1
CTCPD = TCPDBES + TCPDBEV + TCPDPHEV ×FC PHEV + MC PHEV ) (11)
V. SYSTEM DESCRIPTION
To select the suitable algorithm for the proposed problem, six the proposed SMG system consists of EVTs (BEV, PHEV,
algorithms are applied on an advanced SMG object system and FCEV) technology. In this problem formulation, FCEV
as shown in Fig. 1. There are different DGs (MT, FC, PV, is considered as a DG source. Appendix A provides the details
WT), Li-ion BES and DiG in the adopted SMG. In addition, of generation limits and coefficients used in the present work.
If the SMG under study has MT, FC, FCEV, PV, WT, BES,
BEV, PHEV and DiG then the location of mth search agent
Xm can be characterized as (37), shown at the bottom of
this page. More information regarding the implementation
of SMG can be referred from [70], [71], [75]. All DGs FIGURE 2. Forecasted load demand (FLD) and operating reserve (OR).
generate active power at the unity power factor in the current
work. The FX and MT charges, for mounting and operation To verify the performance of different techniques, they are
of batteries used in BES, BEV and PHEV, are assumed to frequently applied on the considered problem of EOSMG for
be 465 (A Cct/ kWh) and 15 (A Cct/ kWh) [13]. The LT and 30 independent trials. The important variables of different
IR for funding the batteries of BES, BEV and PHEV are algorithms include the maximum number of iterations and
correspondingly 3 and 0.06. The tax is considered as 10% in population size. In this work, four different population num-
this work. The charging and discharging rate of batteries are bers are selected i.e. 25, 50, 100 and 200. The maximum
equal and kept at 90%. Let the minimum capacities to be 10% number of iterations is 500. For comparative analysis, the
of the maximum capacities of the batteries. The maximum algorithms WO, FF, PSO, DE, GA and TLBO are considered.
capacities of batteries are set to 500 kWh. For a time horizon The sequential optimization strategy (i.e., one-by-one param-
of one day with hourly time step, the EOSMG studies are eter) is used to tune one parameter at a time, and its optimal
performed. A PC with 2.4 GHz Intel i5-4210U CPU and value is determined empirically. For FF, attraction coeffi-
4 GB RAM is used to simulate the techniques in MATLAB. cient, mutation coefficient and mutation coefficient damping
The operation reserve and forecasted values of load demand ratio are 2, 0.2 and 0.98, respectively. For PSO local learn-
are given in Fig 2. This figure shows the forecasted value ing coefficient and global learning coefficient are selected
CBESmax,1 , C BESmax,2 , · · · , C BESmax,s1 , C BEVmax,1 , C BEVmax,2 , · · · , CBEVmax,s2
CPHEVmax,1 , C PHEVmax,2 , · · · , C PHEVmax,s3 , Pm MT ,1 , PMT ,2 , . . . , PMT ,T ,
m m
PFC,1 , PFC,2 , . . . , PFC,T , PFCEV ,1 , PFCEV ,2 , . . . , PFCEV ,T
m m m m m m
PV ,1 , PPV ,2 , . . . , PPV ,T , PWT ,1 , PWT ,2 , . . . , PWT ,T
Pm m m m m m
PBES,1 , PBES,2 , . . . , PBES,T , PBEV ,1 , PBEV ,2 , . . . , Pm
m m m m m
BEV ,T
Xm = [xm,1 xm,2 . . . . . . xm,D ] ×
PPHEV ,1 , Pm
m
PHEV ,2 , . . . , PPHEV ,T , PGrid,1 , PGrid,2 , . . . , PGrid,T
m m m m
(37)
Pm DiG,1 , PDiG,2 , . . . , PDiG,T , uMT ,1 , uMT ,2 , . . . , uMT ,T ,
m m m m m
uFC,1 , uFC,2 , . . . , uFC,T
m m m
FCEV ,1 FCEV ,2 , . . . , uFCEV ,T , uPV ,1 , uPV ,2 , . . . , uPV ,T
um ,
um m m m m
uWT ,1 , uWT ,2 , . . . , uWT ,T , uBES,1 , uBES,2 , . . . , uBES,T
m m m m m m
um BEV ,1 , uBEV ,2 , . . . , uBEV ,T , uPHEV ,1 , uPHEV ,2 , . . . , uPHEV ,T
m m m m m
uGrid,1 , uGrid,2 , . . . , uGrid,T , uDiG,1 , uDiG,2 , . . . , uDiG,T
m m m m m m
TABLE 3(A). Standard function optimal values (30 trials). TABLE 3(C). Standard function optimal values (30 trials).
FIGURE 3(A). Convergence curve of selected algorithm on different FIGURE 3(B). Convergence curve of selected algorithm on different
population. (a1) Population:25. (a2) Population:25. (b1) Population:50. population. (c1) Population:100. (c2) Population:100. (d1) Population:200.
(b2) Population:50. (d2) Population:200.
B. CASE 2: DISCHARGING MODE OF BATTERIES in Fig. 9(B). Due to economical power supplied by batteries,
In this case, all the batteries are fully charged, the optimal it is beneficial for EOSMG to purchase power from BES and
power output of MT, FC, BES, BEV, PHEV, FCEV, PV, WT, EVTs. In this case, the system considers batteries of opti-
DiG and power grid for six algorithms in the SMG are shown mum size 50 kWh. A comparative analysis among different
TABLE 4. Case I.
FIGURE 5(A1). Trajectory of F1 (a1-a6) for different algorithm. (a1) WO. (a2) FF. (a3) PSO. (a4) DE. (a5) GA. (a6) TLBO.
for the third category (i.e. composite benchmark functions): second best is TLBO as well as their performance are stable;
the standard deviations of functions F14, F16 and F17 are for the composite benchmark functions the performance of
obtained by DE; F15 and F19 are obtained by GA; F18 is WO is best, but their standard deviations are not least, i.e. it’s
obtained by TLBO. From the above analysis, it is observed performance is not stable (there is a lot of fluctuation in the
that for the uni-modal benchmark functions the performance computed mean values).
of TLBO and DE is better and stable; for the multi-modal Fig. 3(A) and (B) presents the comparison of convergence
benchmark function the performance of DE is best and the characteristics for different algorithms for four standard
FIGURE 5(A2). Trajectory of F2 (b1-b6) for different algorithm. (b1) WO. (b2) FF. (b3) PSO. (b4) DE. (b5) GA. (b6) TLBO.
functions (F1, F2, F8 and F9) under two categories (uni- fastest convergence of F1, F2 and F9 is provided by the TLBO
modal and multi modal). To increase the clarity and under- algorithm, which is followed by the WO algorithm. Similarly,
standability of the convergence of different algorithms, two from Figs. 3 (B-c1) and 3 (B-c2), it is clear that the fastest
different types of convergence graphs are generated: the first, convergence of F8 is obtained by the TLBO algorithm, which
normal plot of fitness value with respect to the number is followed by the DE algorithm.
of iterations; the second, semi-log of fitness with respect Fig. 4(a) to 4(d) represents the comparison of average fit-
to the number of iterations. From Figs. 3(A-a1), 3(A-a2), ness value for different algorithms on four standard functions
3(A-b1), 3(A-b2), 3(B-d1) and 3(B-d2), it is clear that the (F1, F2, F8 and F9) under two categories (uni-modal and
FIGURE 5(B1). Trajectory of F8 (c1-c6) for different algorithm. (c1) WO. (c2) FF. (c3) PSO. (c4) DE. (c5) GA. (c6) TLBO.
multi modal). To show the comparative analysis among the FF algorithm. This reflects that all the search agents of TLBO
average fitness value of four functions (F1, F2, F8 and F9) as well as FF are competent to search the optimal search space
with respect to the number of iterations for six algorithms, rather than other algorithms. Similarly, from Fig. 4(b) the
five search agents are considered. From Figs. 4(a) and 4(d), minimum average fitness value for function F2 is obtained
it is apparent that the minimum average value for F1 and F9 by the PSO algorithm, which is followed by FF algorithm.
is provided by the TLBO algorithm, which is followed by the This reflects that all the search agents of PSO as well as
FIGURE 5(B2). Trajectory of F9 (d1-d6) for different algorithm. (d1) WO. (d2) FF. (d3) PSO. (d4) DE. (d5) GA. (d6) PSO.
FF are capable to search the optimal search space rather from the above analysis that all the search agents of TLBO
than other algorithms. Similarly, for F8, Fig.4(c) presents the are best capable to explore the optimal search space.
minimum average fitness value obtained by the TLBO, which Fig. 5(A1), 5(A2), 5(B1) and 5(B2) provides the compari-
is followed by the DE. This reflects that all the search agents son among six algorithms based on the trajectory of the first
of TLBO as well as DE are capable to search the optimal search agent, which is obtained by using standard benchmark
search space rather than other algorithms. It is concluded function F1, F2, F8 and F9. Sub graphs (a1) to (a6) of
From the observations of Figs. 3–6 it is concluded that: in charged condition. Hence all batteries worked as energy
• It is observed from the analysis of Fig. 3 that fastest con- sources.
vergence for most of the standard function is obtained by Case 1: For case one, the results are summarized in Table 4.
the TLBO From the results it is extracted that on most of the population
• It is clear from the analysis of Fig. 4 that all the search sizes, i.e. 25, 50 and 100, FF algorithm provided the best
agents of TLBO are best capable to explore the optimal results, while on the population size of 200, TLBO computed
search space the best results. Table 4 presents a comparison among the
• It is derived from the analysis of Fig. 5 that the effort results obtained by different techniques for EOSMG under
made by the first search agent of TLBO is least four criteria, i.e. average value, the best value, the worst value
• As per the observation of Fig. 6, WO explores a wide and standard deviation. From the results it is clear that at
range of search space than other algorithms. population size 25, 50 and 100, FF algorithm provides the
On the basis of the above aggregation it is concluded that best values i.e. 1310.512, 1255.266 and 1243.399 A Cct/day,
the performance of TLBO is the best in most of the cases respectively, while at population size 200, TLBO generates
among the selected six meta-heuristic techniques. the optimum solution (1262.903 A Cct/day). By comparing the
Apart from the comparison on different standard functions, results obtained for standard test functions and EOSMG, one
another comparison is made on EOSMG problem under four can observe that on standard functions WO and/or TLBO
different population sizes i.e. 25, 50, 100 and 200. For this compute the best results, but for the complex optimization
comparative analysis, a SMG cost minimization problem is problem (EOSMG) with a large number of variables and
proposed and different algorithms are applied on this problem constraints rather than WO and TLBO, FF gives the most
under two different cases. For the first case, all the batteries optimum results.
are connected to the system at no charge condition. In this Fig. 7(A) presents the comparison of convergence char-
case, these batteries are connected as loads to the micro acteristics of different algorithms for the EOSMG problem
grid. The Second case considered all connected batteries under case 1. From the sub Figures (a), (b), (c) and (d) of
Fig. 7(A), the analysis made from the convergence sub graphs
at different population sizes (25, 50, 100 and 200) are as TLBO is better than the other algorithms
follows: • For the population sizes 50 and 100, the convergence of
• For the population size 25, the convergence of FF and FF and GA is better than the other algorithms
FIGURE 8(A). Variation of best cost with respect to population for case 1.
FIGURE 8(B). Variation of best cost with respect to population for case 2.
cost for micro-grid operation, therefore the generation by TABLE 7. Constraints limit and Bids of the DGs, Utility, BES, Dig and EVTs.
different energy sources computed by FF algorithm is optimal
generation. Hence for the population of 25, 50, 100 and
200, the optimal generations are 751.931, 812.585, 769.386,
and 654.67 MW with respect to the optimal costs 1625.751,
1590.586, 1501.22, and 1502.00 A Cct/day.
Apart from analytical analysis, Table 6 provides the the-
oretical comparison of different properties of six meta–
heuristic techniques. This comparison is based on different
factors such as input parameters, convergence, intensification
and diversification component, advantages, drawbacks and
applications.
VII. CONCLUSION
Currently, there is a continuously increasing demand of
TABLE 8(A). Uni-modal benchmark functions.
electrical energy. Therefore, along with conventional energy
sources, the renewable energy sources have also been inte-
grated in the system to fulfill the electrical energy demand.
Micro-gird technology provides a platform to integrate all
types of renewable and non-conventional energy sources. For
the efficient utilization of MG technology, the economic oper-
ation and control problem of micro-grid should be optimized.
Due to a large availability of optimization techniques, the
selection process of appropriate technique is cumbersome.
A comprehensive analysis of optimal economic operation
and control problem of SMG using different meta-heuristics
techniques (WO, PSO, FF, DE, GA and TLBO) is performed of different algorithms on various standard functions that
to select an appropriate optimization technique. the overall performance of TLBO is better than the other
To analyze available meta-heuristic techniques, nineteen algorithms.
standard test functions under three different categories, Further, the comparisons of different algorithms for two
i.e. uni-modal, multi model and composite, are selected. separate cases of developed EOSMG optimization problem
Additionally, a smart micro-grid system with MT, FC, PV, framework are presented.
WT, BES, EVTs and Diesel Generator is proposed to show For case 1, at population sizes 25, 50 and 100, FF algo-
the comparison of an EOSMG problem. Optimization meth- rithm provides the best values i.e. 1310.512, 1255.266 and
ods from three different categories (SI, EA and TL) are 1243.399 A Cct/day, respectively, while at population size 200,
compared on different performance parameters to show their TLBO generates the optimum solution (1262.903 A Cct/day).
effectiveness. Two different comparisons are presented, the The convergence of FF is faster than other algorithms for the
first, for standard functions and the second, for a developed different population sizes. It is observed that with the incre-
problem. Individual source generations are also provided. ment in population size, the performance of DE is impairing
The percentage of minimum mean values of fitness, for while the performance of GA is improving. For WO and FF,
different nineteen standard functions, obtained by the WO, the performance is improved for increment in the population
FF, PSO, DE, GA and TLBO are 21.05, 5.26, 0, 36.84, 5.26 size from 25 to 100, but it deteriorates at 200. The perfor-
and 36.84 respectively. The percentage of minimum standard mances of PSO and TLBO are unstable with an increase in
deviation of fitness, for different nineteen standard functions, population size.
obtained by the WO, FF, PSO, DE, GA and TLBO are 5.26, For case2, at population sizes 25, 50, 100 and 200, FF
5.26, 0, 47.36, 15.78 and 31.57 respectively. It is clear from algorithm provides the best values i.e. 1625.751, 1590.586,
the comparison of mean value, standard deviation, percentage 1501.22 and 1502.0 A Cct/day, respectively. The convergence
of minimum mean values and standard deviation of fitness for of WO and FF is faster than other algorithms for the differ-
different standard functions that the performance of TLBO ent population sizes. It is clear that with the increment in
and DE is better and stable with respect to other algorithms. population size, the performances of PSO, GA and TLBO
As per the convergence characteristic’s analysis of differ- are improving, while for WO and FF the performance are
ent algorithms, it is observed that TLBO technique has the improved from 25 to 100, but it deteriorates at 200. The
fastest convergence. All the search agents of TLBO are best performance of DE is unstable with an increase in popula-
capable to explore the optimal search space. It is observed tion size.
that WO explores a wide range of search space and easily get The comparison of different algorithms for two sep-
stuck in the local minima. It is concluded from the analysis arate cases of developed EOSMG optimization problem
TABLE 8(B). Composite benchmark functions. TABLE 8(C). Multimodal benchmark functions.
APPENDIX A
Constraints limits used in this study are shown in Table 7.
APPENDIX B
Standard functions used in this study are shown
in Table 8(A)–(c).
REFERENCES
[1] C. C. Thompson, P. E. K. Oikonomou, A. H. Etemadi, and V. J. Sorger,
‘‘Optimization of data center battery storage investments for microgrid cost
savings, emissions reduction, and reliability enhancement,’’ IEEE Trans.
Ind. Appl., vol. 52, no. 3, pp. 2053–2060, May/Jun. 2016.
[2] S. Sharma, S. Bhattacharjee, and A. Bhattacharya, ‘‘Grey wolf optimi-
sation for optimal sizing of battery energy storage device to minimise
operation cost of microgrid,’’ IET Generat., Transmiss. Distrib., vol. 10,
no. 3, pp. 625–637, Feb. 2016.
[3] V. Krishnamurthy and A. Kwasinski, ‘‘Effects of power electronics, energy
storage, power distribution architecture, and lifeline dependencies on
framework reflects that the performance of FF is superior microgrid resiliency during extreme events,’’ IEEE J. Emerg. Sel. Topics
to other methods. It is concluded that algorithms of swarm Power Electron., vol. 4, no. 4, pp. 1310–1323, Dec. 2016.
[4] X. Xu, M. Bishop, D. G. Oikarinen, and C. Hao, ‘‘Application and mod-
intelligence category are better fitted to solve such cost min- eling of battery energy storage in power systems,’’ CSEE J. Power Energy
imization problems. Syst., vol. 2, no. 3, pp. 82–90, Sep. 2016.
[5] Z. Liu, Y. Chen, Y. Luo, G. Zhao, and X. Jin, ‘‘Optimized plan- [26] J. Morales-Morales, I. Cervantes, and U. Cano-Castillo, ‘‘On the design
ning of power source capacity in microgrid, considering combina- of robust energy management strategies for FCHEV,’’ IEEE Trans. Veh.
tions of energy storage devices,’’ Appl. Sci., vol. 6, no. 12, p. 416, Technol., vol. 64, no. 5, pp. 1716–1728, May 2015.
2016. [27] X. Sun, B. Liu, Y. Cai, H. Zhang, Y. Zhu, and B. Wang, ‘‘Frequency-
[6] R. Khodabakhsh and S. Sirouspour, ‘‘Optimal control of energy storage in a based power management for photovoltaic/battery/fuel cell-electrolyser
microgrid by minimizing conditional value-at-risk,’’ IEEE Trans. Sustain. stand-alone microgrid,’’ in IET Power Electron., vol. 9, no. 13,
Energy, vol. 7, no. 3, pp. 1264–1273, Jul. 2016. pp. 2602–2610, Oct. 2016.
[7] J. Shen, C. Jiang, Y. Liu, and X. Wang, ‘‘A microgrid energy management [28] H. Ramírez-Murillo, C. Restrepo, J. Calvente, A. Romero, and R. Giral,
system and risk management under an electricity market environment,’’ ‘‘Energy management of a fuel-cell serial–parallel hybrid system,’’
IEEE Access, vol. 4, pp. 2349–2356, 2016. IEEE Trans. Ind. Electron., vol. 62, no. 8, pp. 5227–5235,
[8] Y. Guo, J. Xiong, S. Xu, and W. Su, ‘‘Two-stage economic operation of Aug. 2015.
microgrid-like electric vehicle parking deck,’’ IEEE Trans. Smart Grid, [29] S. S. Thale, R. G. Wandhare, and V. Agarwal, ‘‘A novel recon-
vol. 7, no. 3, pp. 1703–1712, May 2016. figurable microgrid architecture with renewable energy sources and
[9] H. Hassanzadehfard, S. M. Moghaddas-Tafreshi, and S. M. Hakimi, ‘‘Opti- storage,’’ IEEE Trans. Ind. Appl., vol. 51, no. 2, pp. 1805–1816,
mization of grid connected micro-grid consisting of PV/FC/UC with con- Mar./Apr. 2015.
sidered frequency control,’’ Turkish J. Electr. Eng. Comput. Sci., vol. 23, [30] K. V. Vidyanandan and N. Senroy, ‘‘Frequency regulation in a wind-
no. 1, pp. 1–16, 2015. diesel powered microgrid using flywheels and fuel cells,’’ IET Generat.,
[10] H. Alharbi and K. Bhattacharya, ‘‘Optimal sizing of battery energy stor- Transmiss. Distrib., vol. 10, no. 3, pp. 780–788, 2016.
age systems for microgrids,’’ in Proc. IEEE Electr. Power Energy Conf., [31] Q. Tang, N. Liu, and J. Zhang, ‘‘Optimal operation method for microgrid
Calgary, AB, Canada, Nov. 2014, pp. 275–280. with wind/PV/diesel generator/battery and desalination,’’ J. Appl. Math.,
[11] F. Y. Melhem, N. Moubayed, and O. Grunder, ‘‘Residential energy man- vol. 2014, Jun. 2014, Art. no. 857541.
agement in smart grid considering renewable energy sources and vehicle- [32] K. Hajar, A. Hably, A. Elrafhi, Z. Obeid, and S. Bacha, ‘‘Optimization
to-grid integration,’’ in Proc. IEEE Electr. Power Energy Conf. (EPEC), of a microgrid with renewable energy and distributed generation: A case
Ottawa, ON, Canada, Oct. 2016, pp. 1–6. study,’’ in Proc. 19th Int. Conf. Syst. Theory, Control Comput. (ICSTCC),
[12] R. Yu, W. Zhong, S. Xie, C. Yuen, S. Gjessing, and Y. Zhang, ‘‘Bal- Romania, Balkans, Oct. 2015, pp. 662–665.
ancing power demand through EV mobility in vehicle-to-grid mobile [33] F. A. Mohamed and H. N. Koivo, ‘‘System modelling and online optimal
energy networks,’’ IEEE Trans. Ind. Informat., vol. 12, no. 1, pp. 79–90, management of MicroGrid using multiobjective optimization,’’ in Proc.
Feb. 2016. Int. Conf. Clean Electr. Power, Capri, Italy, 2007, pp. 148–153.
[13] F. Laureri, L. Puliga, M. Robba, F. Delfino, and G. O. Bultò, ‘‘An optimiza- [34] H. Afshar, Z. Moravej, and M. Niasati, ‘‘Modeling and optimization
tion model for the integration of electric vehicles and smart grids: Problem of microgrid considering emissions,’’ in Proc. Smart Grid Conf. (SGC),
definition and experimental validation,’’ in Proc. IEEE Int. Smart Cities Tehran, Iran, 2013, pp. 225–229.
Conf. (ISC), Trento, Italy, Sep. 2016, pp. 1–6. [35] V. Jayadev and K. S. Swarup, ‘‘Optimization of microgrid with
demand side management using Genetic Algorithm,’’ in Proc. IET
[14] N. G. Paterakis, O. Erdinç, I. N. Pappi, A. G. Bakirtzis, and J. P. S. Catalão,
Conf. Power Unity, Whole Syst. Approach, London, U.K., Oct. 2013,
‘‘Coordinated operation of a neighborhood of smart households compris-
pp. 1–6.
ing electric vehicles, energy storage and distributed generation,’’ IEEE
[36] P. P. Vergara, R. Torquato, and L. C. P. da Silva, ‘‘Towards a real-time
Trans. Smart Grid, vol. 7, no. 6, pp. 2736–2747, Nov. 2016.
energy management system for a microgrid using a multi-objective genetic
[15] G. Li, D. Wu, J. Hu, Y. Li, M. S. Hossain, and A. Ghoneim, ‘‘HELOS:
algorithm,’’ in Proc. IEEE Power Energy Soc. General Meet., Denver, CO,
Heterogeneous load scheduling for electric vehicle-integrated microgrids,’’
USA, Jul. 2015, pp. 1–5.
IEEE Trans. Veh. Technol., vol. 66, no. 7, pp. 5785–5796, Jul. 2017.
[37] C. Siqi et al., ‘‘Optimal coordinated operation for microgrid with hybrid
[16] L. Yao, W. H. Lim, and T. S. Tsai, ‘‘A real-time charging scheme for
energy storage and diesel generator,’’ in Proc. Int. Conf. Power Syst.
demand response in electric vehicle parking station,’’ IEEE Trans. Smart
Technol., Chengdu, China, Oct. 2014, pp. 3207–3212.
Grid, vol. 8, no. 1, pp. 52–62, Jan. 2017.
[38] E. D. Santis, A. Rizzi, A. Sadeghiany, and F. M. F. Mascioli, ‘‘Genetic
[17] F. Odeim, J. Roes, and A. Heinzel, ‘‘Power management optimization of a optimization of a fuzzy control system for energy flow management
fuel cell/battery/supercapacitor hybrid system for transit bus applications,’’ in micro-grids,’’ in Proc. Joint IFSA World Congr. NAFIPS Annu.
IEEE Trans. Veh. Technol., vol. 65, no. 7, pp. 5783–5788, Jul. 2016. Meet. (IFSA/NAFIPS), Edmonton, AB, Canada, Jun. 2013, pp. 418–423.
[18] Y. Xu, ‘‘Optimal distributed charging rate control of plug-in electric vehi- [39] G. C. Liao, ‘‘Using chaotic quantum genetic algorithm solving envi-
cles for demand management,’’ IEEE Trans. Power Syst., vol. 30, no. 3, ronmental economic dispatch of smart microgrid containing distributed
pp. 1536–1545, May 2015. generation system problems,’’ in Proc. Int. Conf. Power Syst. Technol.,
[19] E. Vinot, V. Reinbold, and R. Trigui, ‘‘Global optimized design of an elec- Hangzhou, China, 2010, pp. 1–7.
tric variable transmission for HEVs,’’ IEEE Trans. Veh. Technol., vol. 65, [40] C. Changsong, D. Shanxu, C. Tao, L. Bangyin, and Y. Jinjun, ‘‘Energy trad-
no. 8, pp. 6794–6798, Aug. 2016. ing model for optimal microgrid scheduling based on genetic algorithm,’’
[20] M. Patterson, N. F. Macia, and A. M. Kannan, ‘‘Hybrid microgrid model in Proc. IEEE 6th Int. Power Electron. Motion Control Conf., Wuhan,
based on solar photovoltaic battery fuel cell system for intermittent load China, May 2009, pp. 2136–2139.
applications,’’ IEEE Trans. Energy Convers., vol. 30, no. 1, pp. 359–366, [41] M. Zolfaghari, H. B. Habil, H. A. Abyaneh, and M. Abedi, ‘‘Load shar-
Mar. 2015. ing improvement between parallel-connected inverter based DGs using
[21] K. Ettihir, L. Boulon, and K. Agbossou, ‘‘Energy management strategy a GA based optimization control strategy in microgrids,’’ in Proc. IEEE
for a fuel cell hybrid vehicle based on maximum efficiency and maxi- PES Asia–Pacific Power Energy Eng. Conf. (APPEEC), Xi’an, China,
mum power identification,’’ in IET Electr. Syst. Transp., vol. 6, no. 4, Oct. 2016, pp. 320–323.
pp. 261–268, Dec. 2016. [42] Z. Shi, Y. Peng, and W. Wei, ‘‘Optimal sizing of DGs and storage
[22] Y. Zhang, Y. Mou, and Z. Yang, ‘‘An energy management study on hybrid for microgrid with interruptible load using improved NSGA-II,’’ in
power of electric vehicle based on aluminum air fuel cell,’’ IEEE Trans. Proc. IEEE Congr. Evol. Comput. (CEC), Beijing, China, Jul. 2014,
Appl. Supercond., vol. 26, no. 7, pp. 1–6, Oct. 2016. pp. 2108–2115.
[23] D. Chakraborty, E. Breaz, A. K. Rathore, and F. Gao, ‘‘Parasitics-assisted [43] Q. Deng, X. Gao, H. Zhou, and W. Hu, ‘‘System modeling and optimization
soft-switching and secondary modulated snubberless clamping current- of microgrid using genetic algorithm,’’ in Proc. 2nd Int. Conf. Intell.
fed bidirectional voltage doubler for fuel cell vehicles,’’ IEEE Trans. Veh. Control Inf. Process., Harbin, China, Jul. 2011, pp. 540–544.
Technol., vol. 66, no. 2, pp. 1053–1062, Feb. 2017. [44] A. Nasser and P. Reji, ‘‘Optimal planning approach for a cost effec-
[24] X. Hu, N. Murgovski, L. M. Johannesson, and B. Egardt, ‘‘Optimal tive and reliable microgrid,’’ in Proc. Int. Conf. Cogeneration, Small
dimensioning and power management of a fuel cell/battery hybrid bus via Power Plants District Energy (ICUE), Bangkok, Thailand, Sep. 2016,
convex programming,’’ IEEE/ASME Trans. Mechatronics, vol. 20, no. 1, pp. 1–6.
pp. 457–468, Feb. 2015. [45] A. S. Eldessouky and H. A. Gabbar, ‘‘Micro grid renewables dynamic
[25] X. Hu, J. Jiang, B. Egardt, and D. Cao, ‘‘Advanced power-source integra- and static performance optimization using genetic algorithm,’’ in Proc.
tion in hybrid electric vehicles: Multicriteria optimization approach,’’ IEEE IEEE Int. Conf. Smart Energy Grid Eng. (SEGE), Oshawa, ON, Canada,
Trans. Ind. Electron., vol. 62, no. 12, pp. 7847–7858, Dec. 2015. Aug. 2015, pp. 1–6.
[46] F. Shariatzadeh, N. Kumar, and A. K. Srivastava, ‘‘Optimal control algo- [65] S. Rani, S. Roy, K. Bhattacharjee, and A. Bhattacharya, ‘‘Teaching learning
rithms for reconfiguration of shipboard microgrid distribution system using based optimization to solve economic and emission scheduling problems,’’
intelligent techniques,’’ IEEE Trans. Ind. Appl., vol. 53, no. 1, pp. 474–482, in Proc. 2nd Int. Conf. Control, Instrum., Energy Commun. (CIEC),
Jan./Feb. 2017. Kolkata, India, 2016, pp. 546–550.
[47] A. Y. Saber and G. K. Venayagamoorthy, ‘‘Smart micro-grid optimization [66] O. Penangsang, A. Soeprijanto, A. A. Fitriana, and E. S. Ningrum,
with controllable loads using particle swarm optimization,’’ in Proc. IEEE ‘‘Operation optimization stand-alone microgrid using firefly algo-
Power Energy Soc. General Meet., Vancouver, BC, Canada, Jul. 2013, rithm considering lifetime characteristics of battery,’’ in Proc. Int.
pp. 1–5. Seminar Intell. Technol. Appl. (ISITIA), Lombok, Indonesia, 2016,
[48] H. Cao et al., ‘‘Economic dispatch of micro-grid based on improved pp. 565–570.
particle-swarm optimization algorithm,’’ in Proc. North Amer. Power [67] S. Mohammadi, B. Mozafari, and S. Soleymani, ‘‘Optimal operation
Symp. (NAPS), Denver, CO, USA, 2016, pp. 1–6. management of microgrids using the point estimate method and firefly
[49] Y. Chen, J. Zhang, Q. Tang, and S. Lin, ‘‘The implementation of algorithm while considering uncertainty,’’ Turkish J. Electr. Eng. Comput.
micro-grid economic dispatch based on particle swarm optimization,’’ Sci., vol. 22, no. 3, pp. 735–753, 2014.
in Proc. Chin. Autom. Congr. (CAC), Wuhan, China, Nov. 2015, [68] P. D. P. Reddy, V. C. V. Reddy, and T. G. Manohar, ‘‘Whale optimization
pp. 1310–1315. algorithm for optimal sizing of renewable resources for loss reduction in
[50] J. Chen, J. Wang, Q. Chen, and D. Wu, ‘‘Optimal dispatch of medium- distribution systems,’’ Renewables: Wind, Water, and Solar, vol. 4, p. 3,
voltage microgrid using an adaptive PSO algorithm,’’ in Proc. 7th Int. Dec. 2017.
Conf. Intell. Human-Mach. Syst. Cybern., Hangzhou, China, Aug. 2015, [69] I. N. Trivedi, M. Bhoye, R. H. Bhesdadiya, P. Jangir, N. Jangir,
pp. 324–329. and A. Kumar, ‘‘An emission constraint environment dispatch prob-
[51] G. Liang, P. Liyuan, L. Ruihuan, Z. Fen, L. Jinhui, and W. Xin, ‘‘Study lem solution with microgrid using whale optimization algorithm,’’
on economic operation for micro-grid based on scenario and PSO,’’ in Proc. 19th IEEE Nat. Power Syst. Conf. (NPSC), Dec. 2016,
in Proc. Int. Conf. Power Syst. Technol., Chengdu, China, Oct. 2014, pp. 1–6.
pp. 3152–3156. [70] S. X. Chen, H. B. Gooi, and M. Q. Wang, ‘‘Sizing of energy storage
[52] H.-T. Yang, C.-T. Yang, C.-C. Tsai, G.-J. Chen, and S.-Y. Chen, ‘‘Improved for microgrids,’’ IEEE Trans. Smart Grid, vol. 3, no. 1, pp. 142–151,
PSO based home energy management systems integrated with demand Mar. 2012.
response in a smart grid,’’ in Proc. IEEE Congr. Evol. Comput. (CEC), [71] A. A. Moghaddam, A. Seifi, T. Niknam, and M. R. A. Pahlavani, ‘‘Multi-
Sendai, Japan, May 2015, pp. 275–282. objective operation management of a renewable MG (micro-grid) with
[53] G. Hao, R. Cong, and H. Zhou, ‘‘PSO applied to optimal oper- back-up micro-turbine/fuel cell/battery hybrid power source,’’ Energy,
ation of a micro-grid with wind power,’’ in Proc. 6th Int. Symp. vol. 36, no. 6, pp. 490–507, 2011.
Parallel Architectures, Algorithms Programm., Beijing, China, 2014, [72] A. A. Moghaddam, A. Seifi, and T. Niknam, ‘‘Multi-operation man-
pp. 46–51. agement of a typical micro-grids using Particle Swarm Optimization: A
[54] P. Q. An, M. D. Murphy, M. C. Breen, and T. Scully, ‘‘Economic optimisa- comparative study,’’ Renew. Sust. Energy Rev., vol. 16, no. 12, pp. 68–81,
tion for a building with an integrated micro-grid connected to the national Feb. 2012.
grid,’’ in Proc. World Congr. Sustain. Technol. (WCST), London, U.K.,
[73] X. Tan, Q. Li, and H. Wang, ‘‘Advances and trends of energy storage
2015, pp. 140–144.
technology in Microgrid,’’ Int. J. Electr. Power Energy Syst., vol. 44,
[55] D. O. Elamine, E. H. Nfaoui, and B. Jaouad, ‘‘Multi-agent architecture
pp. 179–191, Jan. 2013.
for smart micro-grid optimal control using a hybrid BP-PSO algorithm for
[74] L. Gao, S. Liu, and R. A. Dougal, ‘‘Dynamic lithium-ion battery model
wind power prediction,’’ in Proc. 2nd World Conf. Complex Syst. (WCCS),
for system simulation,’’ IEEE Trans. Compon. Packag. Manuf. Technol.,
Agadir, Morocco, Nov. 2014, pp. 554–560.
vol. 25, no. 3, pp. 495–505, Sep. 2002.
[56] G. Liang, P. Liyuan, L. Ruihuan, Z. Fen, and W. Xin, ‘‘Multi-objective
[75] T. Niknam and F. Golestaneh, ‘‘Probabilistic multiobjective operation
stochastic optimal day-ahead scheduling for micro-grid based on scenario
management of microgrids with hydrogen storage and polymer exchange
and PSO,’’ in Proc. China Int. Conf. Electr. Distrib. (CICED), Shenzhen,
fuel cell power plants,’’ Fuel Cells, vol. 12, no. 5, pp. 809–826,
China, Sep. 2014, pp. 204–208.
2012.
[57] N. Tiwari and L. Srivastava, ‘‘Generation scheduling and micro-grid
energy management using differential evolution algorithm,’’ in Proc. Int. [76] W. Al-Saedi, S. W. Lachowicz, D. Habibi, and O. Bass, ‘‘Power
Conf. Circuit, Power Comput. Technol. (ICCPCT), Nagercoil, India, 2016, quality enhancement in autonomous microgrid operation using particle
pp. 1–7. swarm optimization,’’ Int. J. Electr. Power Energy Syst.,vol. 42, no. 1,
[58] D. Shuai, L. Nian, and C. Yingmiao, ‘‘Optimal operation of the island pp. 139–149, 2013.
microgrid with renewable energy and desalination,’’ in Proc. Int. Conf. [77] W. Gu et al., ‘‘Modeling, planning and optimal energy management of
Mech. Sci., Electric Eng. Comput. (MEC), Shengyang, China, Dec. 2013, combined cooling, heating and power microgrid: A review,’’ Int. J. Electr.
pp. 3718–3722. Power Energy Syst., vol. 54, pp. 26–37, Jan. 2014.
[59] L. Wu, Y. Wang, X. Yuan, and Z. Chen, ‘‘Multiobjective optimization [78] G. Li, P. Niu, and X. Xiao, ‘‘Development and investigation of
of HEV fuel economy and emissions using the self-adaptive differen- efficient Artificial Bee Colony algorithm for numerical function
tial evolution algorithm,’’ IEEE Trans. Veh. Technol., vol. 60, no. 6, optimization,’’ Appl. Soft Comput., vol. 12, no. 1, pp. 320–332,
pp. 2458–2470, Jul. 2011. 2012.
[60] W. Fan, N. Liu, and J. Zhang, ‘‘Multi-objective optimization model for [79] Z. Beheshti and S. M. H. Shamsuddin, ‘‘A review of population-based
energy mangement of household micro-grids participating in demand meta-heuristic algorithms,’’ Int. J. Adv. Soft Comput. Appl., vol. 5, no. 1,
response,’’ in Proc. IEEE Innov. Smart Grid Technol.-Asia (ISGT ASIA), pp. 1–35, Mar. 2013.
Bangkok, Thailand, Nov. 2015, pp. 1–6. [80] G. Graditi, M. G. Ippolito, E. Telaretti, and G. Zizzo, ‘‘Technical and
[61] A. K. Basu, A. Bhattacharya, S. Chowdhury, and S. P. Chowdhury, economical assessment of distributed electrochemical storages for load
‘‘Planned scheduling for economic power sharing in a CHP-based micro- shifting applications: An Italian case study,’’ Renew. Sustain. Energy Rev.,
grid,’’ IEEE Trans. Power Syst., vol. 27, no. 1, pp. 30–38, Feb. 2012. vol. 57, pp. 515–523, May 2016.
[62] F. Veltman, L. G. Marin, D. Sáez, L. Guitierrez, and A. Núñez, ‘‘Predic- [81] M. G. Ippolito, E. Telaretti, G. Zizzo, and G. Graditi, ‘‘A new device for the
tion interval modeling tuned by an improved teaching learning algorithm control and the connection to the grid of combined RES-based generators
applied to load forecasting in microgrids,’’ in Proc. IEEE Symp. Ser. and electric storage systems,’’ in Proc. Int. Conf. Clean Electr. Power
Comput. Intell., Cape Town, South Africa, Dec. 2015, pp. 651–658. (ICCEP), Alghero, Italy, Jun. 2013, pp. 262–267.
[63] M. Dixit and R. Roy, ‘‘Impact of PEVs on automatic generation control [82] M. L. Di Silvestre, G. Graditi, and E. R. Sanseverino, ‘‘A generalized
using TLBO algorithm,’’ in Proc. IEEE Int. Conf. Signal Process., Inf., framework for optimal sizing of distributed energy resources
Commun. Energy Syst. (SPICES), Kozhikode, India, Jul. 2015, pp. 1–5. in micro-grids using an indicator-based swarm approach,’’
[64] C. Yammani, G. Sowjanya, S. Maheswarapu, and S. K. Matam, ‘‘Opti- IEEE Trans Ind. Informat., vol. 10, no. 1, pp. 152–162,
mal placement and sizing of DER’s with load models using a modified Feb. 2014.
teaching learning based optimization algorithm,’’ in Proc. Int. Conf. Green [83] G. Graditi, M. L. D. Silvestre, R. Gallea, and E. R. Sanseverino, ‘‘Heuristic-
Comput. Commun. Electr. Eng. (ICGCCEE), Coimbatore, India, 2014, based shiftable loads optimal management in smart micro-grids,’’ IEEE
pp. 1–6. Trans. Ind. Informat., vol. 11, no. 1, pp. 271–280, Feb. 2015.
[84] A. Takeuchi, T. Hayashi, Y. Nozaki, and T. Shimakage, ‘‘Optimal schedul- BASEEM KHAN (M’16) received the B.E. degree
ing using metaheuristics for energy networks,’’ IEEE Trans. Smart Grid, in electrical engineering from Rajiv Gandhi Tech-
vol. 3, no. 2, pp. 968–974, Jun. 2012. nological University in 2008, and the M.Tech.
[85] S. Favuzza, G. Graditi, and E. R. Sanseverino, ‘‘Adaptive and and Ph.D. degrees in electrical engineering from
dynamic ant colony search algorithm for optimal distribution systems the Maulana Azad National Institute of Technol-
reinforcement strategy,’’ Appl. Intell., vol. 24, no. 1, pp. 31–42, ogy, India, in 2010 and 2014, respectively. Since
Feb. 2006. 2015, he has been an Assistant Professor with
[86] C. Gamarra and M. J. Guerrero, ‘‘Computational optimization techniques the Hawassa Institute of Technology, Hawassa
applied to microgrids planning: A review,’’ Renew. Sustain. Energy Rev.,
University, Awasa, Ethiopia. His research interest
vol. 48, pp. 413–424, Aug. 2015.
includes power system restructuring, power sys-
[87] Z. Fan, ‘‘A distributed demand response algorithm and its application to
tem planning, smart grid, meta-heuristic optimization techniques, and renew-
PHEV charging in smart grids,’’ IEEE Trans. Smart Grid, vol. 3, no. 3,
pp. 1280–1290, Sep. 2012. able energy integration.
[88] J. H. Holland, Adaptation in Natural and Artificial Systems. Ann Arbor,
MI, USA: Univ. Michigan Press, 1975. PAWAN SINGH received the B.E. degree in
[89] J. Kennedy and R. C. Eberhart, ‘‘Particle swarm optimization,’’ in computer science and engineering from Chaud-
Proc. IEEE Int. Conf. Neural Netw., Perth, WA, Australia, Jul. 1995, hary Charan Singh University, Meerut, India, the
pp. 1942–1948. M.Tech. degree in information technology from
[90] R. Storn and K. Price, ‘‘Differential evolution—A simple and efficient Guru Gobind Singh Indraprastha University, New
heuristic for global optimization over continuous spaces,’’ J. Global
Delhi, India, and the Ph.D. degree in computer sci-
Optim., vol. 11, no. 4, pp. 341–359, 1997.
ence from Magadh University, Bodh Gaya, India,
[91] R. V. Rao, V. J. Savsani, and D. P. Vakharia, ‘‘Teaching–learning-
in 2013. He is currently with the School of Infor-
based optimization: A novel method for constrained mechanical
design optimization problems,’’ Comput.-Aided Des., vol. 43, no. 3, matics, Hawassa Institute of Technology, Hawassa
pp. 303–315, 2011. University, Awasa, Ethiopia. He has authored or
[92] X.-S. Yang, ‘‘Firefly algorithms for multimodal optimization,’’ in Stochas- co-authored number of research papers in the journals of international
tic Algorithms: Foundations and Applications (SAGA) (Lecture Notes in reputation. His current research interests include software metrics, soft-
Computer Science), vol. 5792, O. Watanabe and T. Zeugmann, Eds. Berlin, ware testing, software cost estimation, web structure mining, energy aware
Germany: Springer, 2009, pp. 169–178. scheduling, and nature inspired meta-heuristic optimization techniques and
[93] S. Mirjalili and A. Lewis, ‘‘The whale optimization algorithm,’’ Adv. Eng. its applications.
Softw., vol. 95, pp. 51–67, May 2016.