This article has been accepted for publication in a future issue of this journal, but has not been

fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2017.2767618, IEEE Access

An Innovative Hybrid Wind-Solar and Battery-Supercapacitor

Microgrid System—Development and Optimization
Umer Akram, Muhammad Khalid, and Saifullah Shafiq

Abstract—This paper presents a methodology for the joint B. Indices

capacity optimization of renewable energy (RE) sources, i.e,
wind and solar, and state-of-the-art hybrid energy storage system
i Index of PV
(HESS) comprising of battery energy storage (BES) and super- j Index of WT
capacitor (SC) storage technology, employed in a grid-connected t Time
microgrid (MG). The problem involves multiple fields, i.e., RE,
battery technology, SC technology, and control theory, and u Case number
requires an efficient and precise co-ordination between sub-fields w Iteration number
to harness the full benefits, this makes the problem labyrinthine.
The optimization problem is formulated, and it involves variety
of realistic constraints from both hybrid generation and storage, C. Parameters
and an objective function is proposed to (i) minimize the cost, A Area of PV array
(ii) improve the reliability, and (iii) curtail greenhouse gases
(GHG) emissions. The complex optimization problem is solved Blim Minimum battery utilization limit
innovatively in piece-wise fashion to decrease the complexity and c Scale factor
computational time. First, sizes of solar photovoltaic (PV) and
e Allowable tolerance
wind turbine (WT) are determined using an innovative search
algorithm, and in the second step, the size of HESS is calculated, FORPV Forced outage rate of PV
finally the optimal solution is determined. A comparison based FORW T Forced outage rate of WT
upon cost, reliability, and GHG emissions is presented which
plainly shows the effectiveness of the proposed methodology. The I Solar irradiation
technique is also applied to determine the size of a MG employing n Total number of intervals
PV, WT and BES operating in grid-connected mode. And a brief
N Total number of renewable sources types
cost analysis, reliability assessment, and emission reduction are
given for three scenarios, 1) MG with HESS, 2) MG with BES, NPV Number of PVs
and 3) MG with conventional generation. It is shown that a MG NW T Number of WTs
with HESS is not only economical but also more reliable and has
lower GHG emissions. Prated Rated power of WT
To Atmospheric temperature
Index Terms—Hybrid energy storage, microgrid, optimization,
renewable power. vr Rated speed of WT
vci Cut-in speed of WT
N OMENCLATURE vco Cut-out speed of WT
A. Abbreviations ηPV Efficiency of PV
ηc Charging efficiency of battery
BES Battery energy storage
ηd Discharging efficiency of battery
DG Distributed generator
η Efficiency of supercapacitor
ERBC Emission reduction benefit cost
ξ Self discharge rate of supercapacitor
ESS Energy storage system
∆t Time step
GHG Greenhouse gases
σ Shape factor
HESS Hybrid energy storage system
MG Microgrid
D. Variables
PSO Particle swarm optimization
Bmax Maximum energy capacity of battery
PV Photovoltaic
Bcap Required energy capacity of battery
RE Renewable energy
Bopt Optimal energy capacity of battery
SC Supercapacitor
Bchg Battery charging factor
UG Utility grid
BCS Battery corrected size
WT Wind turbine
BDV Battery decision variable
This work was supported by the Deanship of Research at King Fahd Cmax Maximum energy capacity of supercapacitor
University of Petroleum & Minerals (KFUPM).
The authors are with the Electrical Engineering Department (EED) at King Ccap Required energy capacity of supercapacitor
Fahd University of Petroleum & Minerals, Dhahran 31261, Saudi Arabia.
Corresponding author: Umer Akram (g201512930@kfupm.edu.sa),
Copt Optimal energy capacity of supercapacitor
Phone:+966 53 048 1960, Fax: +966 13 860 3535 Cinv−sc Investment cost of renewable energy sources

2169-3536 (c) 2017 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2017.2767618, IEEE Access

Cc,sr Capital cost of renewable energy sources I. I NTRODUCTION

Cinv−stg Investment cost of storage
Cc,stg
Crep−stg
Capital cost of storage
Replacement cost of storage
D URING the past few decades, rising concerns for global
warming and volatile fossil fuels prices have made
renewable energy (RE) sources an attractive alternative. This
Com Operation and maintenance cost trend has been further underpinned by rapid advancements in
Com, f Fixed operation and maintenance cost the power electronics field, which enabled full controllability
Com,v Variable operation and maintenance cost of RE sources, within the constraints inflicted by the natural
CMG−U Total cost of energy supplied by MG to UG phenomenon [1]. The integration of RE sources in electric
power grid has evolved into the concept of microgrid (MG).
CU−MG Total cost of energy supplied by UG to MG
MGs are state-of-the-art active distribution networks consisting
Cexchg Per-unit cost of the exchanged power of distributed generators (DGs), energy storage system (ESS),
CERBC Emission reduction benefit cost and flexible loads, operated grid-connected or islanded, in a
Cg Per-unit cost of microgrid controlled, coordinated way [2]. Due to the propinquity of
ES Energy served DGs to the loads in MGs and the utilization of RE sources,
ENS Energy not served MGs are trusted to supply its customers with more efficient
and eco-friendly energy, reduced power losses and network
EBES Energy stored in battery
congestion, and improved power quality and reliability com-
min
EBES Minimum stored energy limit of battery pared to the energy supplied by conventional power plants.
max
EBES Maximum stored energy limit of battery MGs are contemplated to be an integral part of smart grids in
ESC Energy stored in supercapacitor the future electric power system.
max
ESC Maximum stored energy limit of supercapacitor Solar and wind are two expeditiously emerging RE sources,
min
ESC Minimum stored energy limit of supercapacitor especially solar has gained more popularity due to significant
decline in its cost over the past few years. Since, such sources
NDES Net discounted energy served
are intermittent, uncontrollable, stochastic, and highly variable,
PL Load power demand their integration in the electric power grid poses challenges to
PGT Microgrid total power generation its effective operation, especially at higher penetration levels.
PCap Power supplied by supercapacitor For example, load mismatch, poor load following, voltage
PBat Power supplied by battery instability, frequency deviation, inferior power quality, and
c
PBES Battery charging power reliability problems are some of the detrimental impacts that
d RE sources introduce in electric power network [3]. New, in-
PBES Battery discharging power
novative technologies and novel ideas are required to alleviate
PPV Power supplied by PV the aforementioned problems, to increase the penetration of
PW T Power supplied by WT RE sources in the electric power grid.
Pgap Difference between generation and demand A potential candidate solution to the aforesaid problems is
Pgap−L Low frequency component of Pgap to store energy during surplus generation hours using ESS and
Pgap−H High frequency component of Pgap redispatch it appropriately later when needed [4]–[8]. Several
PG Power generated by wind and solar types of ESS are available and among them battery energy
storage (BES) system is most frequently utilized [9]–[11]. But
∆P Cumulative error matrix
batteries are only efficient at supplying low steady loads, while
cmax
PBES Battery charging power limit outputs of RE sources are highly fluctuating, which are not
dmax
PBES Battery discharging power limit suitable for them. It is difficult for batteries to recover from
c
PBES Battery charging power rapid power swings without a significant reduction in their
d
PBES Battery discharging power lifetime [12]. An ideal ESS must have a high power density
max
PSC Supercapacitor power limit to follow rapid power fluctuations, a high energy density to
give autonomy to the electric power grid, and longer life. As
PMG−U Power supplied by microgrid to utility
a sole energy storage technology is unlikely to deliver these
PU−MG Power supplied by utility to microgrid essentials effectively and economically, it is vital to couple
PW F Present worth factor multifarious energy storage technologies, creating a hybrid
r Uniform random number energy storage system (HESS).
RSspace Reduced search space Most recently, HESS has become an emergig storage tech-
Sspace Search space nology as it combines the benefits of multiple technologies.
For example, BES and SC can be combined to build HESS.
SS Solution space
The BES systems have high specific energy, low specific
To Atmospheric temperature power, low self-discharge, low cycle life, long charge times,
v Wind speed and relatively lower cost per watt-hour. On the other hand,
ωo Cut-off frequency the SC storage systems have low specific energy, high specific
power, high cycle life, very high self-discharge, short charge

2169-3536 (c) 2017 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2017.2767618, IEEE Access

Generation
Bus
Load Bus energy storage, and optimize their capacities. This optimiza-
tion exploits the benefits from each individual element, and
therefore the solution achieved is more cost efficient, highly
PSC PBES reliable and eco-friendlier.
Energy Management Bidirectional
System and Control Power As a case study, the proposed methodology is validated
SOCBES
SOCSC
Converter using real-world data of wind speed, solar irradiation and
Power Signal
power demand from Dammam city in Saudi Arabia.
Control Signal
The remainder of the paper is organized as follows. Section
Communication
Signal
BES SC
II presents the related work and detailed MG modeling is
Hybrid Energy Storage System discussed in Section III. The proposed methodology is demon-
strated in Section IV. The information of the databases is
Fig. 1. A hybrid energy storage system. provided in Section V. The Section VI presents results and
discussions and conclusion is given in Section VII.

times, and high cost per watt-hour [13], [14]. The HESS makes
use of complementary properties of BES and SC and provides II. R ELATED W ORK
large energy supply, high power, and fast dynamic response Different methods have been proposed in literature for
at the same time economically and effectively [15]–[18]. sizing of MG and we will briefly review some of them here. In
Nevertheless, to optimize the lifetime of both BES and SC, [20], sizing of PV, WT, and ESS is done based on minimization
it is vital to ensure that both operate within their operational of total planning cost. In [21], genetic algorithm is used to
constraints. BES must operate within its state of charge and determine the optimal sizes of PV, WT, diesel, and BES based
current bounds and SC within its voltage and current bounds. upon cost, carbon emissions, and dump energy. In [22], sizing
At the same time the SC should respond to rapid large current of WT, PV, BES, and fuel cell is done based on the cost and
signals in order to maximize the lifespan of BES [19]. It is reliability. In [23], optimal sizes of PV, WT, diesel generator,
also important to note that the cost of energy storage units per BES, and pumped storage are determined based upon min-
kilowatt is a strong function of their capacity, and too high imizing both initial investment and operational/maintenance
cost is prohibitive to commercial and industrial acceptance, a costs. In [24], sizes of PV, WT, and BES are determined
method for optimizing the size and operation of such HESS to using mixed integer linear programming. In [25], capacity
fit application constraints is a crucial task. In this paper, BES optimization of PV, WT, diesel generator, BES, fuel cell,
and SC storage technologies will be combined to build up a electrolyzer, and hydrogen tank is done based upon cost
HESS as shown in Fig. 1. minimization, GHG emissions, and reliability using particle
This paper proposes a methodology for the capacity op- swarm optimization (PSO). In [26], optimal sizes of PV,
timization of a residential MG employing hybrid PV-WT WT, and BES are determined considering multiple objectives,
and BES-SC. The proposed method benefits from multi- i.e., cost minimization, and higher supply reliability. In [27],
disciplinary fields. For example, battery storage technology, optimal sizing of MG comprising of WT, PV, BES, and
SC technology, RE technology, control theory with an efficient biomass is done considering two alternative objectives, i.e.,
co-ordination within the subsystems. For instance, very precise minimization of total annual energy losses and cost of energy.
co-ordination is required for the operation of HESS specifi- In [28], optimal sizing and sitting of renewable DGs is done
cally at the switching instance otherwise the full benefits of based upon the minimization of annual investment cost and
HESS will not be harnessed. The sizing problem is formulated operation cost. In [29], planning of PV, WT, and BES grid-
and it is a complex optimization problem, subjected to variety connected MG is done based upon cost minimization and
of realistic constraints both from hybrid generation and stor- customer satisfaction maximization, and mixed integer linear
age, and an objective function is proposed to (i) minimize the programming is used to solve the optimization problem. In
cost, (ii) improve the reliability, and (iii) curtail green house [30], optimal sizing of PV, WT, and BES based residential
gases (GHG) emissions. The complex optimization problem MG is done considering the cost minimization objective. In
is solved innovatively in piece-wise fashion to reduce the [31], capacity optimization of PV and BES is done based
complexity and computational time. First, combinations of upon levelized cost of energy. In [32], capacity optimization
optimal sizes of PV and WT are determined out of several of PV, WT, tidal turbine, and BES is done based upon the
possible combinations using an advanced constraint based net present cost and reliability, the optimization problem is
innovative search algorithm, and in the second step, size of solved using crow search algorithm. In [33], optimal sizes of
HESS is calculated for each combination. Finally, the optimal PV, WT, diesel generator, biodiesel generator, fuel cell, and
solution is determined based upon the minimum value of the BES are determined based on minimization of cost of energy.
cost function. The main contributions of our work is that In [34], optimization of BES-SC is done based upon initial
we propose a technique for combined capacity optimization investment cost and simulated annealing PSO is used to solve
of hybrid generation, i.e., WT-PV system and the HESS the optimiztion problem. In [35], a statistical approach is used
comprising of BES and SC employed in grid-connected MG. to optimize the size of the BES-SC hybrid storage system.
As, instead of a single energy storage or renewable energy It is important to understand that the size of ESS depends
source, we consider hybrid power generation with hybrid upon the behavior of load, wind power, and solar power. So,

2169-3536 (c) 2017 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2017.2767618, IEEE Access

optimal size of ESS determined for a particular geographical historical data. Using the wind speed the power output of a
location cannot be considered as optimal for any other location WT can be calculated as [36].
even with same installed capacities of RE sources. Thus, 
to harvest maximum benefits it is necessary to optimize 
 0 v < vci

RE sources and ESS altogether. Also, BES systems have

Prated × v−vci

vci ≤ v < vr
vr −vci
limited number of cycles. One of the primary purposes of PW T (v) = (4)
hybridization of BES system with SC storage technology is


 Prated vr ≤ v < vco

to prolong the lifespan of BES system. Therefore, it is also

0 v ≥ vco
vital to consider the cycles of BES system while optimizing
BES-SC system. Moreover, initial investment cost, operation where vci , vr , and vco are cut-in, rated, and cut-out speeds
and maintenance cost, replacement cost, reliability, and GHG respectively, PW T is the output power of WT, and Prated is
emissions should be considered in the objective function to get the rated power of WT. From (4) the power output of WT is
more optimized results. It is clearly evident from the existing zero below vci and above vco , and the output power increases
literature, the researchers have considered these aforemen- linearly with the increase in the wind speed between vci and
tioned parameters in their formulations, however, considering vr , and it generates rated power between vr and vco .
sum of parameters is important but in most of the cases,
formulations are based upon one parameter only. For example, B. Solar Photovoltaic Generation System
[34] has optimized the size of BES-SC energy storage based A solar PV generation system converts solar energy to elec-
upon initial investment cost. trical energy. The output power of a PV system depends upon
In general, the literature deals with the capacity optimization solar irradiation, atmospheric temperature, area and efficiency
of MGs employing single or multiple RE sources, conventional of the PV array. It is assumed that a maximum power point
DGs, ESS, and variety of loads, operating in grid-connected or tracker is installed to extract the maximum of available power.
islanded mode based upon cost or cost-reliability. And to date, The hourly output power of a PV system is calculated as
to the best knowledge of authors no methodology has been
developed for the joint capacity optimization of the emerging PPV (t) = ηPV AI (t) (1 − 0.005 (To (t) − 25)) (5)
hybrid PV-WT and ultramodern BES-SC system employed by
a grid-connected MG system. where ηPV is the efficiency and A is the area in m2 of the
solar cell array, I is the solar irradiation in kW /m2 , and To is
the atmospheric temperature in oC.
III. M ICROGRID M ODELLING
The block diagram of the MG considered in this study
C. Battery Energy Storage System
is shown in Fig. 2, it utilizes PV-WT for hybrid power
generation and BES-SC for hybrid energy storage. A hybrid A BES system consists of series and parallel strings of
RE system, employing two or more RE sources, mitigates batteries. In this work, a sodium sulphur (NaS) battery is
the intermittent nature of RE resources to some extent and considered. NaS is one of the batteries used for commercial
also improves the system efficiency [21]. The PV and WT electrical energy storage in electric utility distribution grid
are coupled to generation bus through DC/AC and AC/AC support, wind power integration, and high-value grid services.
converters respectively, and HESS is connected to AC bus via Its applications include load levelling, peak shaving, and
bidirectional DC/AC converter. The residential load is taped power quality as well as renewable energy management and
from the load bus via step down coupling transformers, and integration. A BES model, as given in [37], is calculated as
the utility grid (UG) is also connected to the load bus at point c
of common coupling (PCC) through a controlled switch. Charge : EBES (t + ∆t) = EBES (t) + ∆tPBES ηc (6)

d
PBES
A. Wind Power Generation System Disharge : EBES (t + ∆t) = EBES (t) − ∆t (7)
ηd
A WT generates electrical energy from kinetic energy of
wind. Weibull distribution function can be used to estimate Charging/discharging constraints are
the wind speed as c
0 ≤ PBES cmax
≤ PBES (8)
  v σ 
F (v) = 1 − exp − (1) d dmax
c 0 ≤ PBES ≤ PBES (9)
1
v = c [−ln (1 − r)] σ (2) Stored energy bounds are
min max
1
EBES ≤ EBES (t) ≤ EBES (10)
v = c [−ln (r)] σ (3)
where EBES is the energy stored in the battery, i.e., state-of-
where σ is shape factor, c is scale factor, v is wind speed c
charge, PBES d
and PBES are charging and discharging powers
in m/sec and, r is uniform random number. The values of σ respectively, and ηc is charging efficiency and ηd is discharg-
and c of any geographical location can be obtained from the ing efficiency of the battery.

2169-3536 (c) 2017 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2017.2767618, IEEE Access

Microgrid
Generation Load
Communication signal E: Error
Bus Bus

Residential Demand
L: Low
Control signal Hybrid Frequency
Power Demand Component
Power signal
H: High
E Frequency

Low Pass Filter

Component
L
Switching Control Utility Grid
PCC
H

Energy Switch
PBES SOCBES
Management Control
System
BES SC PSC SOCSC
Hybrid Power Hybrid Energy
Generation System Storage Constraints

Fig. 2. Block diagram of grid-connected MG system.

D. Supercapacitor IV. P ROPOSED M ETHODOLOGY

Energy stored in the SC at any instant is modeled as The proposed methodology is further divided into five
sub-sections. Section IV-A presents methodology for sizing
of RE sources, and HESS sizing strategy is discussed in
ESC (t + ∆t) = ESC (t) + η∆tPSC − ξ ESC (t) (11) detail in Section IV-B. Reliability and economic modeling is
discussed in Section IV-C, while modeling of GHG emissions
subjected to the following constraints is presented in Section IV-D. Finally, the objective function
min max formulation is given in Section IV-E.
ESC ≤ ESC (t) ≤ ESC (12)

A. Renewable Energy Sources Sizing

max
0 ≤ PSC (t) ≤ PSC (13) Consider a hybrid PV-WT generation system as shown in
Fig. 2. Power generated by the system is calculated as follows
where ESC is the energy stored in SC, η is charg-
ing/discharging efficiency, ξ is self-discharge rate, and PSC (i, j)
PG i
(t) = NPV i
PPV (t) PVstatus j
(t) + NWj T PW T (t)W Tstatus (t)
is the power supplied/drawn to/from the SC. During charging (15)
period PSC is positive while it is negative during discharg- ∀i ∈ [1, imax ] , j ∈ [1, jmax ] ,t > 0
ing period. The (12) represents the stored energy constraint
whereas (13) represents the bounds for power supplied/drawn where
to/from the SC.
(
i
0 GiPV (t) < FORPV
PVstatus (t) = ∀i,t > 0 (16)
1 otherwise
E. Utillity Grid
GiPV (t) = rand ()
Utility grid serves two important objectives, (i) ensures load
and generation always to be equal by supplying the demand and
0 GWj T (t) < FORW T
(
whenever MG generation is lower than demand, and (ii) buys j
energy from the MG during surplus generation hours to make W Tstatus (t) = ∀i,t > 0 (17)
1 otherwise
the system more economical. The power of UG at any instant
of time can be modeled as GWj T (t) = rand ()
subjected to the following constraints
Pgrid (t) = PL (t) − ∑ (PW T (t) , PPV (t) , PBES (t) , PSC (t)) (14)
min i max
NPV ≤ NPV ≤ NPV
where Pgrid is the power supplied to/from the UG, PL is load j
NWmin max
T ≤ NW T ≤ NW T
demand, PBES is the power supplied by BES and PSC is the
power supplied by SC. During surplus generation hours Pgrid where PG is the renewable generated power, NPV and NW T are
is negative while positive during inadequate supply hours. the number of PVs and WTs, PPV and PW T are the powers

2169-3536 (c) 2017 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2017.2767618, IEEE Access

generated by single PV and WT, and PVstatus and W Tstatus where Sspace is the search space which contains all possible
are the statuses of PVs and WTs which decide whether they combinations of NPV and NW T and their corresponding cumu-
would generate power or not. When the value of PVstatus of a lative errors. The Sspace is reduced by selecting the minimum
solar PV is 0 that means the solar PV cannot generate power value of the ∆P from its each column as follows
because of some fault or any other reason. As given in (16) and h i
1 jmax
(17), values of PVstatus and W Tstatus are calculated using forced ∆ P min = ∆Pmin · · · ∆Pmin (28)
1× jmax
outage rates FORPV and FORW T . GPV and GW T are random
numbers which are generated using the rand () command of where
j
MATLAB. NPV min , N min , N max , and N max are minimum and ∆Pmin = min (Sspace (z, j)) ∀j (29)
WT PV WT
maximum number of PVs and WTs which are calculated using z = 2, ......, imax + 1
following relations:
where ∆ P min is the vector which contains the minimum values
min ∑n αPL (t) of cumulative errors , i.e., ∆Pmin extracted from the Sspace . The
NPV = t=1
n (18)
∑t=1 PPV (t) values of NPV and NW T that correspond to the minimum values
n
β PL (t) of the cumulative errors are stored in the following vectors as
∑t=1
NWmin
T = n (19)
∑t=1 PW T (t)
h iT
1 jmax
NPVmin = NPV min
· · · NPV min
(30)
n jmax ×1
max ∑t=1 γPL (t)
NPV = n (20)
∑t=1 PPV (t)
h i
NWTmin = NW1 Tmin · · · NWjmax
Tmin (31)
n 1× jmax
∑t=1 ρPL (t)
NWmax
T = n (21) A reduced search space is formulated using (28)-(31) as
∑t=1 PW T (t)
where α, β , γ, and ρ are scaling factors and n is the total RSspace = ∆ P Tmin NWTmin T
 
NPVmin jmax ×3
(32)
number of intervals. The instantaneous error between load and
generation is calculated as follows where RSspace is the reduced search space. There are jmax
combinations of NPVmin and NW Tmin in RSspace for which there
(i, j)
∆p(i, j) (t) = PL (t) − PG (t) ∀i, j,t > 0 (22) will be jmax optimal storage sizes, one for each combination.
where ∆p is the instantaneous error. Cumulative error ∆P is
the sum of absolute values of all instantaneous errors. B. Hybrid Energy Storage System Sizing
n   Power generated by any combination selected from the
∆P(i, j) (t) = ∑ |∆p(i, j) (t)| ∀i, j,t > 0 (23) reduced search space is calculated using following equation
t=1

where ∆P(i, j)
is the cumulative error corresponding to NPVi
PGu (t) = NPV
min u
PPV (t) + NWmin
T PW T (t)
u
∀t > 0 (33)
j
and NW T . A smaller value of the cumulative error means
that intermittent generation follows load demand effectively. subjected to following constraints
j
While a larger value of the cumulative error shows that there min 1 min u
minmax
NPV ≤ NPV ≤ NPV
is a significant difference between load and generation. All
1 u j
cumulative errors are stored in a matrix as NWmin min minmax
T ≤ NW T ≤ NW T

where PG is the power generated by the MG and u is the case

∆P(1,1) ∆P(1, jmax )
 
···
number. The difference between power generated and power

∆P =  .. .. .. 
(24) demanded for case u at any instant of time is denoted by pugap
 . . .


and calculated as follows
∆P(imax ,1) ··· ∆P(imax , jmax ) (imax × jmax )

where ∆ P is the matrix that contains the values of the pugap (t) = PL (t) − PGu (t) ∀t > 0 (34)
cumulative errors for every possible combination of NPV and The pgap is divided into two components, i.e., high frequency
NW T . The values of NPV and NW T that correspond to each ∆P and low frequency components as shown in Fig. 2. The high
are stored in the vectors N PV and N W T as follows frequency component is used for the sizing of SC and the
 min max
T low frequency component is used for the sizing of BES.
NPV = NPV · · · NPV (25) The algorithm for dividing pgap into high and low frequency
components can be realized by low pass energy filter whose
NWT = NWmin · · · NWmax
 
T T (26) transfer function is given below
A search space has been formulated by using (24)-(26) as Kωo2
H (s) = (35)
" # s2 + (ωo /Q) + ωo2
0 NWT
Sspace = (27)
NPV ∆ P
(imax +1)×( jmax +1)
pugap = pugap−H + pugap−L (36)

2169-3536 (c) 2017 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2017.2767618, IEEE Access

where ωo is cut-off frequency, pgap−H is high frequency where Smin and Smax are the minimum and maximum bound-
component, and pgap−L is low frequency component. The aries of solution, ST EPa is the difference between the mini-
maximum capacities of BES and SC that can be employed mum and maximum boundaries, and w is the iteration number.
to store all of the excess energy are determined by using Initially, Smin and Smax are set equal to 0 and Bmax respectively.
the algorithms given below. These algorithms represent the Both of them get updated in an iterative fashion until ST EPa
operation of BES and SC based upon which their maximum becomes less than or equal to the allowable tolerance e. Similar
capacities, that can be installed, have been calculated. calculations that are given in (43)-(47) can be repeated (just
with a little modification in the constraints while calculating
(
u u x) to find the required size of SC (Ccap ). The Bcap and Ccap are
u
Pgap−H (h̄) Pgap−H (h̄) > 0
Pgap−H (h̄) = ∀h̄ (37) the required sizes but both are not optimal. If they are installed,
0 otherwise the system will be reliable but not efficient and economical.
An important parameter that needs to be checked is utiliza-
where
h̄ tion factor; higher value of utilization factor reduces the idle
u
Pgap−H (h̄) = ∑ pugap−H (t) (38) time and ensures the maximum benefit from BES and SC.
t=1 Therefore, utilization factor is used to determine the optimal
The maximum SC size is calculated as capacities of BES and SC. In order to calculate the utilization
u u
factor, battery decision variable BDV is computed as
Cmax = max(Pgap−H ) (39)
∑ Buchg−dcg (t)
Similarly, for BES we have BDV u = ∀t > 0 (48)
( n
u
Pgap−L u
(h̄) Pgap−L (h̄) > 0
u where
Pgap−L (h̄) = ∀h̄ (40) (
0 otherwise u (t) − Pu (t − ∆t)| ≥ λ Pu
1 |Pgap gap BESmax
Buchg−dcg (t) = ∀t > 0
where 0 else
h̄ (49)
u
Pgap−L (h̄) = ∑ pugap−L (t) (41) where Bchg−dcg is the battery charging and discharging factor
t=1
which is equal to one when the charging or discharging power
The maximum capacity of BES is calculated as is greater than or equal to λ PBESmax , where λ is a constant
Bumax = max(Pgap−L
u
) (42) which can take a value btween 0 and 1. Hence, optimal size
of BES is calculated as
where Bmax and Cmax are the maximum capacities of BES and (
SC. Required capacities of BES and SC would be less than or u
Bucap BDV u ≥ Blim
Bopt = (50)
equal to Bmax and Cmax which are calculated as BCSu else
(
u
Bumax x = 1 where Bopt is the optimal capacity of BES and Blim is the min-
Bcap = (43) imum battery utilization limit. The Bopt equals to Bcap if BDV
CBSu x = 0 becomes greater than or equal to Blim otherwise Bopt equals
where Bcap is the required energy capacity of battery. The to battery corrected size BCS. The BCS is calculated using a
Bcap is equal to Bmax if the battery discharges completely region elimination iterative search algorithm as following
after the full charging period and this condition is indicated by u (w) + OCF u (w)
OCFmin max
x = 1. While the Bcap is equal to corrected battery size CBS BCSu (w) = (51)
if the battery does not discharge fully after its full charging 2
period, this condition is indicated by x = 0. The condition where
(
x = 0 implies that the battery is over-sized. An iterative region u (w) BDV u ≥ BU
OCFmin
u lim
elimination algorithm is used to calculate CBS which is defined OCFmin (w + 1) = (52)
BCSu (w) else
by following equations:
and
u Su (w) + Smax
u (w)
CBS (w) = min
(
(44) BCS(w) BDV u ≥ BUlim
2 u
OCFmax (w + 1) = (53)
u (w) else
OCFmax
where (
u (w)
Smin x=1
u
Smin (w + 1) = (45) ST EPbu = OCFmax
u u
(w) − OCFmin (w) (54)
CBS(w) x = 0
where OCFmin and OCFmax are the minimum and maximum
and ( boundaries of solution respectively and ST EPb is the error
u (w) x = 0
Smax
u
Smax (w + 1) = (46) between minimum and maximum boundaries. Initially, OCFmin
CBS(w) x=1 and OCFmax are set equal to 0 and Bucap and both of them get
updated in the iterative algorithm until ST EPb reaches below
ST EPau = Smax
u u
(w) − Smin (w) (47) the allowable tolerance e. Similar calculations that are given

2169-3536 (c) 2017 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2017.2767618, IEEE Access

in (50)-(54) are repeated (just with a little modification in energy exchanged between MG and UG are considered. The
the constraints while calculating BDV ) to find optimal size following relation is used to determine the investment cost
of SC (Copt ). A suitable value of ωo for every combination associated with RE sources
of RSspace is determined using particle swarm optimization
N
(PSO) technique to calculate the Bopt and Copt . u
Cinv−sc = k
Psrk,u (62)
∑ Cc,sr
k=1
C. Reliability and Economic Modeling k
where Cinv−sc is the total investment cost of RE sources, Cc,sr
Reliability and cost are two important parameters that can be is the capital cost of kth source in $/MW, Psrk is the installed
used to analyze the performance of a system. A power system capacity of kth source in MW, and N is the total number of RE
having lower cost, higher reliability (i.e., energy served), sources. The investment cost of storage system is calculated
and lower GHG emissions can be considered to have better as follows
performance. In this study, the optimal solution is determined
M  
on the basis of cost, reliability, and GHG emissions. u
Cinv−stg = k
Cc,stg k,u
Estg k
+Cc,pcs k,u
Pstg (63)
The total generation of MG after finding the optimal size of
∑
k=1
the HESS, i.e., Bopt and Copt for uth combination of PV and k
WT is calculated as following where Cc,stg is the capital cost of kth storage unit in $/MWh,
Estg is the energy capacity of kth storage unit in MWh, Cc,pcs
k k is
u u
the cost of power conditioning system required for the storage
u min
PGT (t) = NPV PPV (t) + NWmin u u
T PW T (t) + PBat (t) + PCap (t) in $/MW, Pstgk is the power capacity of kth storage unit in MW,
(55) and M is the total number of storage units. The replacement
where PGT is the total power generated by MG, PBat is the cost of storage is modeled as following
power supplied by battery and PCap is the power supplied by
SC. Energy served is the summation of demand that is served k E k,u
Cc,stg
k,u stg
by the MG over a period of its operation. Crep−stg =∑ s = p, 2p, 3p, ....l − p (64)
(1 + d)s
n
ESu = ∑ Du (t) (56) where d is the discount rate and p is the life of the storage
t=1 in years. The operation and maintenance costs of the MG
where consist of fixed and variable operation and maintenance costs
(
u (t) ≥ P (t) are calculated as follows
PL (t) PGT L
Du (t) = ∀t > 0 (57) u u u
u (t) otherwise = Com,
PGT Com f +Com,v (65)
where ES is energy served, PL is the load power demand, and where
n M+N
PGT is total generation of the MG. Energy not served ENS u
T k (t) Jom,
k k
Com, f = ∑ ∑ f Pr (66)
is the summation of demand that is not supplied by the MG t=1 k=1
during its operation. and
n M+N
n u k
u u Com,v =∑ ∑ Jom,v Pk (t) (67)
ENS = ∑ G (t) (58) t=1 k=1
t=1
where where Com, f is the fixed operation and maintenance cost, T k
( is the time of operation of kth element, Jom,
k
f is the operation
u (t) P (t) > Pu (t)
PL (t) − PGT
u
G (t) =
L GT
∀t > 0 and maintenance cost factor in $/MW-yr, and Prk is the rated
0 otherwise power capacity of kth element in MW. The variable operation
k
and maintenance cost is referred by Com,v , Jom,v is the variable
(59)
Net discounted energy served is calculated as operation and maintenance cost factor in $/MW, and Pk (t) is
the output power of kth element at time t in MW. Net present
NDESu = ESu PW F (60) value of Com is calculated as
where u u
" # NPVCom = Com PW F (68)
(1 + d)l − 1
PW F = (61)
d (1 + d)l where NPVCom is the present worth of Com .
During surplus generation hours, the MG sells energy to
where PW F is the present worth factor, l is the year of UG. Total cost and net present worth of the energy supplied
operation, and NDESu is the net discounted energy served by by MG to UG is calculated as follows
the MG.
n
As discussed earlier, cost is also an important parameter u u
CMG−U = ∑ Cexchg (t) PMG−U (t) (69)
that can be used to assess the performance of a system. In t=1
this work initial investment cost, fixed and variable operation
u u
and maintenance costs, storage replacement cost, and cost of NPVCMG−U = CMG−U PW F (70)

2169-3536 (c) 2017 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2017.2767618, IEEE Access

TABLE I (
G REENHOUSE G ASES E MISSION DATA X u1 , X u2 ) = 0 ` = 1, 2, , ..., m
g ` (X
s.t. (76)
X u1 , X u2 ) ≤ 0
h ı (X ı = 1, 2, , ..., q
Greenhouse gases CO2 CO SO2 NOx
Emissions kg/MWh 1000.7 1.55 9.993 6.46
where
h i
Correction Cost $/kg 0.0037 0.16 0.97 1.29 minu u
X u1 = NPV , NWmin
T , Pu
Cap , Pu
Bat (77)

and
where CMG−U is the total cost of the energy sold by MG to UG,
X u2 = ECap
 u u
, ωou

, EBat (78)
Cexchg (t) is the cost at time t, PMG−U (t) is the power supplied
by MG to UG at time t, and NPVCMG−U is the present worth The first term in the objective function is cost per unit of
of CMG−U . In the event when MG generation is insufficient to the MG. It is important to note that the initial investment
meet the demand, the MG buys power from UG. Total cost cost of sources (62), initial investment cost of storage (63),
and net present worth of energy supplied by the utility to MG operation and maintenance cost (68), replacement cost (64),
is calculated using following equations cost of energy supplied by UG to MG (71), and energy served
n
(60) are incorporated in the cost per unit of MG. Second term
u
CU−MG u
= ∑ Cexchg (t) PU−MG (t) (71) of the objective function represents the GHG emissions that are
t=1 translated in terms of cost using ERBC concept as discussed
in Section IV-D. The equality constraints are referred by g and
u u in-equality constraints are referred by h . All system constarints
NPVCU−MG = CU−MG PW F (72)
are summarized as:
where CU−MG is the total cost of energy supplied by UG to The Pimary System Constraint (Generation = Demand):
MG, Cexchg (t) is the cost at time t, PU−MG (t) is the power
supplied by UG to MG at time t, and NPVCU−MG is the u
PGT u
+ PU−MG u
− PL − PMG−U =0 (79)
present worth of CU−MG .
The hybrid power generation constraints:
1 u j
D. Modeling of GHG Emissions min
NPV min
≤ NPV min
≤ NPV (80)
When electric power is generated by burning fossil fuels,
it results in GHG emissions in the environment. There is a 1 u j
NWmin min min
T ≤ NW T ≤ NW T (81)
correction cost which is needed to mitigate the damage caused
by these emissions as shown in Table I. This correction cost BES Constraints:
would be a saving if the electric power is generated by utilizing
c cmax
RE sources instead of fossil fuels. This saving is named as 0 ≤ PBES ≤ PBES (82)
emission reduction benefit cost (ERBC), and modeled as
d dmax
4 n 0 ≤ PBES ≤ PBES (83)
u u k k
CERB = ∑∑ PGT (t) E Ecc (73)
k=1 t=1
min max
EBES ≤ EBES (t) ≤ EBES (84)
where CERB is the total ERBC, PGT (t) is the power output of
MG at time t, E k is the emission of kth type of GHG, and Ecc
k SC Constraints:
th
is the cost required to correct the damage caused by k type
min max
of GHG. The net present value of total ERBC is calculated ESC ≤ ESC (t) ≤ ESC (85)
using following equation as
max
u u 0 ≤ PSC (t) ≤ PSC (86)
NPVCERB = CERB PW F (74)
Switching Frequency Constraint:
where NPVCERB is the present worth of ERBC.
0 ≤ ωou ≤ 1 (87)
E. Cost Function Formulation A solution space is generated which contains the values of
The objective is to find an optimal combination of PV, WT, cost function that correspond to all jmax combinations of PV,
BES, and SC that must result in lower cost, higher reliability, WT, BES, and SC. Solution space is given as follows
and lower GHG emissions. Cost function of the optimization
S S = F 1 F 2 ... F jmax 1× jmax
 
problem is formulated as following (88)

q
2
ob j : F u = X u1 , X u2 ) − f2u (X
f1u (X X u1 , X u2 ) → min (75)

2169-3536 (c) 2017 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2017.2767618, IEEE Access

10

Algorithm I Optimal Capacity Sizing TABLE II

M ONTHLY S HAPE AND S CALE FACTORS
Start
Initialization: i ← 1, j ← 1, t ← 1, u ← 1, e ← 0.1 Month σ c
• Data Generation Jan 2.40 4.77
Feb 2.45 4.85
Read: c, σ , I, α, β , γ, ρ, Load Mar 2.55 5.15
Calculate: PPV , PW T , PL , NPV min , N min , N max , N max , P Apr 2.40 5.06
WT PV WT G
min
Save: PPV , PW T , PL , NPV , NWmin max max May 2.40 5.52
T , NPV , NW T , PG Jun 2.60 6.51
• Search Space Formation Jul 2.50 5.54
Aug 2.30 4.91
While i ≤ imax Do Sep 2.20 4.18
While j ≤ jmax Do Oct 2.05 4.09
Nov 2.20 4.38
Calculate: ∆P(i, j) , ∆ P Dec 2.00 4.68
Save: ∆ P , NPV , NWT
Calculate: Sspace
j ← j+1
End While
i ← i+1
End While 1

• Reduced Search Space Formation

While j ≤ jmax + 1 Do 0.95

While i ≤ imax + 1 Do
j j
Calculate: ∆Pmin , NPV , NWj Tmin
Demand (pu)

min 0.9
i ← i+1
End While
0.85
Calculate: ∆ P min , NPVmin , NWTmin
j ← j+1
End While 0.8
Calculate: RSspace
• HESS Sizing
0.75
1 5 10 15 20 24
While u ≤ jmax Do Time (hr)
Call PSO
Calculate: PGu , pugap , ωo , Pgap−H
u u
, Pgap−L Fig. 3. Normalized average daily power demand of Dammam of calendar
year 2015.
• Required size of BESS and SC

Calculate: Bumax , x, Cmax

u ,

Call Region reduction iterative search algorithm

• Optimal size of BESS and SC

Calculate: Buchg−dcg , BDV u 1

Call Region reduction iterative search algorithm 0.9

Calculate: PGTu , E u , Cu , F u
S g 0.8
u ← u+1 0.7
Solar irradiation (pu)

End While
0.6
• Determination of the optimal solution
0.5
Generate: S S
0.4
Call Minima search algorithm
0.3
Optimal solution
End 0.2

0.1
where S S is the solution space and F is value of the
0
objective function. There are jmax possible solutions each 1 5 10 15 20 24
corresponding to one combination. The best solution from the Time (hr)

S S is selected based upon minimum value of the cost function

Fig. 4. Normalized daily average solar irradiation.
by using a minima search algorithm. The pseudocode of the
optimal sizing is given in Algorithm I.

2169-3536 (c) 2017 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2017.2767618, IEEE Access

11

TABLE III
E CONOMIC AND T ECHNICAL DATA OF S OURCES

PV WT BES SC
FORPV Cinv Com FORW T Cinv Com Life Cycles DOD Cinv Com Cinv Com
(—) ($/kW) ($/kW-yr) (—) ($/kW) ($/kW-yr) (No.) (%) ($/kWh) ($/kWh-yr) ($/kWh) ($/kWh-yr)
0.04 2025 16 0.04 2346 33 2500 100 400 10 1000 10

1 8000

0.99
7000
0.98

Objective function ($/MWh)

0.97 6000
Wind speed (pu)

0.96
5000
0.95

0.94 4000

0.93
3000
0.92

0.91 2000
1 5 10 15 20 24 1 100 200 300 400 500 600 700 800
Time (hr) Index of solution space

Fig. 5. Normalized daily average wind speed. Fig. 6. Variation in the cost function w.r.t index of the solution space.

V. DATABASES operated with properly sized and suitable type of energy stor-
As mentioned earlier, the proposed methodology is tested age. The optimal size of ESS is characterized by both energy
using real-world data of wind speed, solar irradiation and storing capacity and maximum power rating. Second step of
power demand from Dammam city in Saudi Arabia. The the strategy is to determine the energy (MWh) and power
Dammam city lies in the eastern province of Kingdom of (MW) capacities of BES system and SC storage. Finally, the
Saudi Arabia and its coordinates are 26.3927N, 49.9777E. The optimal combination of PV, WT, BES, and SC is determined
wind speed is calculated using the shape and scale parameters based upon three important parameters, i.e., cost, reliability,
of Weibull distribution. The shape and scale parameters were and GHG emissions. The solution space, as given in (88), is
determined using well maintained meteorological data of wind formulated and shown in Fig. 6. It is important to note that
of 20 years [38]. Monthly shape and scale parameters are given for each index of solution space there is a combination of PV,
in Table II. The normalized daily average residential power WT, BES, and SC, and optimal combination corresponds to
demand of the calendar year 2015 is presented in Fig. 3, which the index with minimum value of objective function. A smaller
depicts that the daily peak occurs around 8 pm while minimum value of the objective function implies that the energy supplied
demand appears around 7 am. The normalized daily average by the MG is larger, cost per unit of the MG is smaller,
solar irradiation is shown in Fig. 4 showing irradiation peak GHG emissions are lesser, and emission reduction benefit cost
at around 1pm. Whereas the normalized daily average wind (as defined in (73)) is higher. In the presented case study,
speed is shown in Fig. 5. It can be observed that wind speed the minimum value of cost function appears corresponding to
is variable and fluctuates throughout the day. The economic index number 88 which can be seen in Fig. 6.
and technical data is given in Table III. The optimal capacities of PV, WT, BES, and SC are given
in Table IV. The overall cost per unit is USD 0.1552. It is
important to note that the cost per unit is the function of many
VI. S IMULATION R ESULTS AND D ISCUSSIONS
The first step in our proposed methodology is to find
combinations of PV and WT for the reduced search space TABLE IV
given in (32). The determination of the combinations is based O PTIMAL C APACITIES C OMBINATION
upon minimization of the cumulative error between load and
generation, defined in (23), so that renewable power generation Source Solar Wind BES SC
should have better load following. The algorithm given in Type (MW) (MW) (MWh,MW) (MWh,MW)
Section III-A has selected 800 combinations of PV and WT. Capacity 87 88 48 9.6 4.4 52
As mentioned earlier, RE sources perform effectively when

2169-3536 (c) 2017 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2017.2767618, IEEE Access

12

Fig. 7. Cost function vs different combinations of PV and WT.

Fig. 8. Cost function vs different combinations of SC and BES.

parameters, for example a location with higher correlation

between the solar irradiation curve, wind speed curve and
load curve would result in further reduction in cost, and our 160
optimized solution will then be more effective and justified.
In order to get more comprehensible insight in variations 140

of the cost function with RE sources and HESS capacities, 120

the cost function is plotted against the combinations of PV
and WT, and SC and BES, that corresponds to each index 100
Power (MW)

of solution space shown in Fig. 7 and Fig. 8 respectively. 80

Initially, the installed capacities of both RE sources and HESS
are small while values of the cost function are higher because 60

GHG emissions are higher, energy served by the MG is lesser,

40
and ERBC is also small. With the increase in the installed
capacities, corresponding values of the cost function decrease 20

because energy supplied by the MG increases which results in

0
lower GHG emissions and increased ERBC. Although, per 0 1000 2000 3000 4000 5000 6000 7000 8000 9000
unit cost also increases with the increase in the installed Time (hr)

capacities but its impact on the objective function is non-

Fig. 9. Hybrid power generation.
dominant for moderate capacities. However, for very large
capacities, values of the objective function are high because the
impact of cost of RE sources and HESS on objective function
becomes dominant. Although, for very large capacities energy
supplied by the MG is higher, GHG emissions are lesser and 100

ERBC is also higher but their impact on the cost function 90

is less dominant as compared to the cost of RE sources and
HESS. The cost of BES and SC is very high so for larger 80

capacities they have large impact on overall cost of the MG.

70
Power (MW)

The hybrid power generated by optimal PV-WT system

and the load power demand are shown in Fig. 9 and Fig. 60

10 respectively. While the error/gap between the generation

50
and demand is plotted in Fig. 11. This error is required to
be supplied to/by energy storage system, which is HESS in 40

our case. The error signal is supplied to/by BES and/or SC

30
depending upon its magnitude and frequency. A histogram of
the error signal is presented in Fig. 12 which shows that most 20
0 1000 2000 3000 4000 5000 6000 7000 8000 9000
of the times the magnitude of error is more than the rated Time (hr)
power capacity of BES system. However, most of the times
the error signals can be supplied by utilizing the HESS system. Fig. 10. Load power demand.
A comparison based upon cost per unit of the MG, GHG

2169-3536 (c) 2017 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2017.2767618, IEEE Access

13

Energy supplied by MG
100 5 Energy supplied by UG

80 4.5

60 4

40 3.5

Energy (GWh)
Power (MW)

20 3

0 2.5

-20 2

-40 1.5

-60 1

-80 0.5

-100 0
0 1000 2000 3000 4000 5000 6000 7000 8000 9000 1 100 200 300 400 500 600 700 800
Time (hr) Index of solution space

Fig. 11. Difference in generation and demand. Fig. 13. Energy share of MG and UG.

106
300 9
CO
8 NOx
SO2
250
CO2*

Green house gases emissions(kg*)

7
kg* = 100kg
Frequency of appearing

200 6

5
150
4

100 3

2
50
1

0 0
-100 -80 -60 -40 -20 0 20 40 60 80 100 0 100 200 300 400 500 600 700 800
Power (MW) Index of solution space

Fig. 12. Histogram of gap between generation and demand. Fig. 14. Variation in GHG emissions.

emissions, energy served by the MG, and ERBC, between four decreases. In the beginning, energy supplied by MG increases
different possible solutions selected from the solution space, and energy supplied by utility decreases rapidly and finally
is tabulated in Table V. Cost per kWh of the MG is minimum both saturates.
for Case number 2 and maximum for Case number 4, since The variation in the GHG emissions and ERBC with the
installed capacities in this case are very high. The GHG indices of solution space are shown in the Figs. 14 and 15
emissions are minimum for Case number 4 while maximum respectively. It can be depicted that GHG emissions decrease
for Case number 1, and moderate for Case number 2 and 3. while ERBC increases with the increase in the index (as overall
Energy served is highest for Case number 4 whilst lowest installed capacities of RE sources and HESS increase with the
for Case number 1 and moderate for Case number 2. As the increase in the index). It can be observed from Fig. 14 that
optimal solution is determined based upon the cost, reliability, the GHG emissions for the optimal solution are almost 55%
i.e., energy served, and GHG emissions, Table V clearly shows less as compared to the conventional generation.
that Case number 2 is optimal. The operation of the MG over a period of one year with the
As mentioned earlier that the output of RE sources is optimal parameters is shown in Fig. 16. It can be observed that
variable, so it may happen during the operation of the MG the system supplies the required demand effectively. As the
that output of the RE sources and storage system becomes system under consideration is grid-connected that is why the
inadequate to supply the required load demand, during such generation is always equal to the required load power demand
events, the MG buys power from UG in order to meet the which makes the system highly reliable.
demand. Total energy share supplied by the MG and the UG A comparison between the following three case studies is
to fulfill the required load demand is shown in Fig. 13. It presented in Table VI.
can be observed that with the increase in indices the energy • CASE-I: PV/WT/BES/SC based grid-connected MG.
supplied by the MG increases while energy supplied by UG • CASE-II: PV/WT/BES based grid-connected MG.

2169-3536 (c) 2017 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2017.2767618, IEEE Access

14

TABLE V
C OMPARISON B ETWEEN D IFFERENT P OSSIBLE S OLUTIONS

Case Index PV Capacity WT Capacity BES Capacity SC Capacity MG Generation Cost GHG Emissions Energy Served ERBC
No. No. (MW) (MW) (MWh) (MWh) ($/kWh) (kt) (GWh) (M$)
1 60 100 60 28 3.4 0.28 245 237 5.21
2 88 87 88 48 4.4 0.27 213 269 5.90
3 200 54 200 166 10 0.315 117 362 7.97
4 800 10 800 308 42 0.801 15 463 10.18

by the CASE-I is more than that of CASE-II. Similarly, the

11
emissions are minimum for CASE-I and maximum for CASE-
10 III. The emissions for CASE-I are approximately 55% less
than that of CASE-III. These less GHG emissions also result
Emission reduction benefit cost (M$)

9
in considerable savings in terms of ERBC and it can be seen
8
from the Table VI that the ERBC is maximum for CASE-I.
So, the HESS is not only economical but also more reliable
7 and cleaner as compared to BES.
6
TABLE VI
5 C OMPARISON OF D IFFERENT C ASE S TUDIES

4
Index CASE-I CASE-II CASE-III
3
1 100 200 300 400 500 600 700 800 WT (MW) 88 88 0.0
Index of solution space
PV (MW) 87 87 0.0

Fig. 15. Emission reduction benefit cost. BES (MWh) 48 49 0.0

SC (MWh) 4.4 0.0 0.0
100
Demand MG 269 266 0.0
Energy (GWh)
Generation Utility 209 212 478
90
CO 357 362 817
80 CO2 230544 233854 527274
100 Emissions (T)
NOx 1488 1510 3404
90
70 SO2 2302 2335 5265
Power (MW)

80
60
Cost ($/MWh) 155.2 155.4 100
70
ERBC (M$) 5.9 5.8 0.0
50 60

50
40
40
0 100 200 300 400 500
30

VII. C ONCLUSION
20
0 1000 2000 3000 4000 5000 6000 7000 8000 9000
Time (hr) This paper has presented a methodology for joint capacity
optimization of hybrid renewable power generation system and
Fig. 16. Actual demand vs supplied power with optimal system parameters.
energy storage in the context of a grid-connected microgrid
(MG). The hybrid generation system is comprised of solar
photovoltaic (PV) and wind turbine (WT) and hybrid energy
• CASE-III: A MG employing conventional generation. storage system (HESS) is composed of battery energy storage
The cost per unit is minimum for CASE-III but it can not (BES) system and supercapacitor (SC) technology. The com-
be selected as in this case the load demand is served by bined optimization exploits the benefits of both hybrid power
conventional generation only. The cost per unit of the CASE-I generation and HESS. The proposed strategy is primarily
is 0.13% less than that of CASE-II. This can result in savings based upon a few important factors associated with a MG
of about USD 1M per year as compared to CASE-II. So, by system such as cost minimization, greenhouse gases (GHG)
employing the HESS instead of BES system, cost per unit emissions reduction, higher emission reduction benefit cost
decreases which results in considerable savings. This decrease (ERBC), and higher reliability. The optimization problem has
in the cost is due to the fact that the SC prolongs the lifespan been formulated and solved in a piece-wise fashion to decrease
of BES system. The clean energy (energy from MG) supplied the complexity and computational time.

2169-3536 (c) 2017 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2017.2767618, IEEE Access

15

The proposed methodology has been tested using real [8] M. Khalid and A. Savkin, “A model predictive control approach to
residential power demand, solar irradiation and wind speed the problem of wind power smoothing with controlled battery storage,”
Renewable Energy, vol. 35, no. 7, pp. 1520–1526, 2010.
data. The resulted optimal solution is economical, has higher [9] M. Aneke and M. Wang, “Energy storage technologies and real life
reliability and lesser GHG emissions when compared with applications–a state of the art review,” Applied Energy, vol. 179, pp.
other possible solutions. It has also been shown that when 350–377, 2016.
[10] A. V. Savkin, M. Khalid, and V. G. Agelidis, “A constrained monotonic
the MG is operated with optimal parameters it serves the de- charging/discharging strategy for optimal capacity of battery energy
mand effectively. Moreover, a comparison between three case storage supporting wind farms,” IEEE Transactions on Sustainable
studies, i.e., PV/WT/BES/SC, PV/WT/BES, and conventional Energy, vol. 7, no. 3, pp. 1224–1231, 2016.
generation has also been presented. It has been observed that [11] M. Khalid, A. Ahmadi, A. V. Savkin, and V. G. Agelidis, “Minimiz-
ing the energy cost for microgrids integrated with renewable energy
the topology, PV/WT/BES/SC, resulted in the optimal choice resources and conventional generation using controlled battery energy
as there are multiple benefits associated with hybrid BES-SC storage,” Renewable Energy, vol. 97, pp. 646–655, 2016.
energy storage system. It is an economical and reliable solution [12] T. Ma, H. Yang, and L. Lu, “Development of hybrid battery–
supercapacitor energy storage for remote area renewable energy sys-
because the use of SC in conjunction with BES prolongs tems,” Applied Energy, vol. 153, pp. 56–62, 2015.
the BES lifespan, and supplies the demand more effectively [13] A. Schneuwly, “High reliability power backup with advanced energy
and efficiently. In addition, it results in least GHG emissions storage (white paper),” Maxwell Technologies ,San Diego, CA, USA,,
Tech. Rep., 2006.
thus increasing ERBC which makes the overall system more
[14] Q. Xu, X. Hu, P. Wang, J. Xiao, P. Tu, C. Wen, and M. Y. Lee, “A
economical and eco-friendly. decentralized dynamic power sharing strategy for hybrid energy storage
For future research, a detailed load analysis considering dif- system in autonomous DC microgrid,” IEEE Transactions on Industrial
ferent types of loads including controllable and uncontrollable Electronics, vol. 64, no. 7, pp. 5930–5941, 2017.
[15] J. Shen and A. Khaligh, “A supervisory energy management control
loads, and load shifting will be performed to determine an strategy in a battery/ultracapacitor hybrid energy storage system,” IEEE
optimum combination of RE sources along with conventional Transactions on Transportation Electrification, vol. 1, no. 3, pp. 223–
generation that fulfills the load effectively and economically. 231, 2015.
[16] Y. Liu, W. Du, L. Xiao, H. Wang, S. Bu, and J. Cao, “Sizing a hybrid
Furthermore, uncertainty associated with the availability of RE energy storage system for maintaining power balance of an isolated
sources and degradation of battery energy storage will also be system with high penetration of wind generation,” IEEE Transactions
considered for more realistic results. on Power Systems, vol. 31, no. 4, pp. 3267–3275, 2016.
[17] S. Zhang, R. Xiong, and J. Cao, “Battery durability and longevity
based power management for plug-in hybrid electric vehicle with hybrid
ACKNOWLEDGMENT energy storage system,” Applied Energy, vol. 179, pp. 316–328, 2016.
The authors would like to acknowledge the support provided [18] J. M. Blanes, R. Gutiérrez, A. Garrigós, J. L. Lizán, and J. M. Cuadrado,
by the Deanship of Research (DSR) at King Fahd University “Electric vehicle battery life extension using ultracapacitors and an
FPGA controlled interleaved buck–boost converter,” IEEE Transactions
of Petroleum & Minerals (KFUPM) for funding this work on Power Electronics, vol. 28, no. 12, pp. 5940–5948, 2013.
through project No. SR161001. Also, we would like to thank [19] B. Hredzak, V. G. Agelidis, and M. Jang, “A model predictive control
the Research Institute (RI) at KFUPM and Saudi Electricity system for a hybrid battery-ultracapacitor power source,” IEEE Trans-
actions on Power Electronics, vol. 29, no. 3, pp. 1469–1479, 2014.
Company (SEC) for providing the time series data of solar [20] H. Lotfi and A. Khodaei, “Hybrid AC/DC microgrid planning,” Energy,
irradiation, wind speed data and residential load data. vol. 118, pp. 37–46, 2017.
[21] A. Ogunjuyigbe, T. Ayodele, and O. Akinola, “Optimal allocation
R EFERENCES and sizing of PV/Wind/Split-diesel/Battery hybrid energy system for
minimizing life cycle cost, carbon emission and dump energy of remote
[1] F. Blaabjerg, R. Teodorescu, M. Liserre, and A. V. Timbus, “Overview residential building,” Applied Energy, vol. 171, pp. 153–171, 2016.
of control and grid synchronization for distributed power generation [22] R. Hosseinalizadeh, H. Shakouri, M. S. Amalnick, and P. Taghipour,
systems,” IEEE Transactions on industrial electronics, vol. 53, no. 5, “Economic sizing of a hybrid (PV–WT–FC) renewable energy system
pp. 1398–1409, 2006. (HRES) for stand-alone usages by an optimization-simulation model:
[2] E. Kuznetsova, C. Ruiz, Y.-F. Li, and E. Zio, “Analysis of robust case study of iran,” Renewable and Sustainable Energy Reviews, vol. 54,
optimization for decentralized microgrid energy management under un- pp. 139–150, 2016.
certainty,” International Journal of Electrical Power & Energy Systems,
[23] P. Yang and A. Nehorai, “Joint optimization of hybrid energy storage
vol. 64, pp. 815–832, 2015.
and generation capacity with renewable energy,” IEEE Transactions on
[3] A. Kyritsis, D. Voglitsis, N. Papanikolaou, S. Tselepis, C. Christodoulou,
Smart Grid, vol. 5, no. 4, pp. 1566–1574, 2014.
I. Gonos, and S. Kalogirou, “Evolution of PV systems in Greece and
review of applicable solutions for higher penetration levels,” Renewable [24] T. Dragičević, H. Pandžić, D. Škrlec, I. Kuzle, J. M. Guerrero, and
Energy, vol. 109, pp. 487–499, 2017. D. S. Kirschen, “Capacity optimization of renewable energy sources
[4] H. T. Le and T. Q. Nguyen, “Sizing energy storage systems for wind and battery storage in an autonomous telecommunication facility,” IEEE
power firming: An analytical approach and a cost-benefit analysis,” in Transactions on Sustainable Energy, vol. 5, no. 4, pp. 1367–1378, 2014.
Power and Energy Society General Meeting - Conversion and Delivery [25] M. Sharafi and T. Y. ELMekkawy, “Multi-objective optimal design of hy-
of Electrical Energy in the 21st Century. IEEE, 2008, pp. 1–8. brid renewable energy systems using PSO-simulation based approach,”
[5] P. Denholm and R. M. Margolis, “Evaluating the limits of solar pho- Renewable Energy, vol. 68, pp. 67–79, 2014.
tovoltaics (PV) in electric power systems utilizing energy storage and [26] L. Xu, X. Ruan, C. Mao, B. Zhang, and Y. Luo, “An improved
other enabling technologies,” Energy Policy, vol. 35, no. 9, pp. 4424– optimal sizing method for wind-solar-battery hybrid power system,”
4433, 2007. IEEE transactions on Sustainable Energy, vol. 4, no. 3, pp. 774–785,
[6] B. Yang, Y. Makarov, J. Desteese, V. Viswanathan, P. Nyeng, B. Mc- 2013.
Manus, and J. Pease, “On the use of energy storage technologies for [27] E. Sfikas, Y. Katsigiannis, and P. Georgilakis, “Simultaneous capacity
regulation services in electric power systems with significant penetration optimization of distributed generation and storage in medium voltage
of wind energy,” in 5th International Conference on European Electricity microgrids,” International Journal of Electrical Power & Energy Sys-
Market. IEEE, 2008, pp. 1–6. tems, vol. 67, pp. 101–113, 2015.
[7] A. Khatamianfar, M. Khalid, A. V. Savkin, and V. G. Agelidis, “Improv- [28] J. Zhu, W. Gu, G. Lou, L. Wang, B. Xu, M. Wu, and W. Sheng,
ing wind farm dispatch in the Australian electricity market with battery “Learning automata based methodology for optimal allocation of renew-
energy storage using model predictive control,” IEEE Transactions on able distributed generation considering network reconfiguration,” IEEE
Sustainable Energy, vol. 4, no. 3, pp. 745–755, 2013. Access, 2017.

2169-3536 (c) 2017 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2017.2767618, IEEE Access

16

[29] J. Chen, W. Zhang, J. Li, W. Zhang, Y. Liu, B. Zhao, and Y. Zhang,

“Optimal sizing for grid-tied microgrids with consideration of joint opti-
mization of planning and operation,” IEEE Transactions on Sustainable
Energy, 2017.
[30] R. Atia and N. Yamada, “Sizing and analysis of renewable energy and
battery systems in residential microgrids,” IEEE Transactions on Smart
Grid, vol. 7, no. 3, pp. 1204–1213, 2016.
[31] C. S. Lai and M. D. McCulloch, “Sizing of stand-alone solar PV and
storage system with anaerobic digestion biogas power plants,” IEEE
Transactions on Industrial Electronics, vol. 64, no. 3, pp. 2112–2121,
2017.
[32] A. Askarzadeh, “Electrical power generation by an optimised au-
tonomous PV/wind/tidal/battery system,” IET Renewable Power Gen-
eration, vol. 11, no. 1, pp. 152–164, 2016.
[33] Y. A. Katsigiannis, P. S. Georgilakis, and E. S. Karapidakis, “Hy-
brid simulated annealing–tabu search method for optimal sizing of
autonomous power systems with renewables,” IEEE Transactions on
Sustainable Energy, vol. 3, no. 3, pp. 330–338, 2012.
[34] T. Zhou and W. Sun, “Optimization of battery–supercapacitor hybrid en-
ergy storage station in wind/solar generation system,” IEEE transactions
on sustainable energy, vol. 5, no. 2, pp. 408–415, 2014.
[35] A. Abbassi, M. A. Dami, and M. Jemli, “A statistical approach for
hybrid energy storage system sizing based on capacity distributions in
an autonomous PV/Wind power generation system,” Renewable Energy,
vol. 103, pp. 81–93, 2017.
[36] A. Arabali, M. Ghofrani, M. Etezadi-Amoli, and M. S. Fadali, “Stochas-
tic performance assessment and sizing for a hybrid power system of
solar/wind/energy storage,” IEEE Transactions on Sustainable Energy,
vol. 5, no. 2, pp. 363–371, 2014.
[37] S. Chen, H. B. Gooi, and M. Wang, “Sizing of energy storage for
microgrids,” IEEE Transactions on Smart Grid, vol. 3, no. 1, pp. 142–
151, 2012.
[38] S. Rehman, T. Halawani, and T. Husain, “Weibull parameters for wind
speed distribution in Saudi Arabia,” Solar Energy, vol. 53, no. 6, pp.
473–479, 1994.

2169-3536 (c) 2017 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.