IET Renewable Power Generation

Research Article

Optimal sizing of a wind/solar/battery hybrid ISSN 1752-1416

Received on 9th January 2017

grid-connected microgrid system

Revised 7th September 2017
Accepted on 2nd October 2017
E-First on 3rd November 2017
doi: 10.1049/iet-rpg.2017.0010
www.ietdl.org

Umer Akram1 , Muhammad Khalid1, Saifullah Shafiq1

1Department of Electrical Engineering, King Fahd University of Petroleum & Minerals (KFUPM), Dhahran 31261, Saudi Arabia
E-mail: g201512930@kfupm.edu.sa

Abstract: Higher cost and stochastic nature of intermittent renewable energy (RE) resources complicate their planning,
integration and operation of electric power system. Therefore, it is critical to determine the appropriate sizes of RE sources and
associated energy storage for efficient, economic and reliable operation of electric power system. In this study, two constraint-
based iterative search algorithms are proposed for optimal sizing of the wind turbine (WT), solar photovoltaic (PV) and the
battery energy storage system (BESS) in the grid-connected configuration of a microgrid. The first algorithm, named as sources
sizing algorithm, determines the optimal sizes of RE sources while the second algorithm, called as battery sizing algorithm,
determines the optimal capacity of BESS. These algorithms are mainly based upon two key essentials, i.e. maximum reliability
and minimum cost. The proposed methodology aims to avoid over- and under-sizing by searching every possible solution in the
given search space. Moreover, it considers the forced outage rates of PV, WT and utilisation factor of BESS which makes it
more realistic. Simulation results depict the effectiveness of the proposed approach.

Nomenclature Cg per unit cost of microgrid

Crep replacement cost
Indices
ES energy served
i index of PV ENS energy not served
j index of WT EB energy stored in battery
t time NDES net discounted energy served
u case number
PL load power demand
w iteration number
PGT microgrid total power generation
Parameters PBES power supplied by battery
PPV power supplied by PV
A area of PV array PWT power supplied by WT
Blim minimum battery utilisation limit pgap difference between generation and demand
c shape factor PG power generated by wind and solar
e allowable tolerance
Pgrid power supplied by utility grid
FORPV forced outage rate of PV
ΔP cumulative error matrix
FORWT forced outage rate of WT
r uniform random number
I solar irradiation RSspace reduced search space
n total number of intervals
Sspace search space
DOD depth of discharge
NPV number of PVs SV solution vector
Δp instantaneous error between generation and demand
NWT number of WTs
ΔP cumulative error between generation and demand
Prated rated power of WT v wind speed
To atmospheric temperature
vr rated speed of WT
1 Introduction
vci cut-in speed of WT
vco cut-out speed of WT Most recently, a major proportion of electricity has been produced
by utilising fossil fuels such as coal, natural gas and oil, which
ηPV efficiency of PV
emits massive amounts of greenhouse gases in the environment.
ηB efficiency of battery Moreover, fossil fuels reserves are depleting and their prices are
σ scale factor also volatile. On the other hand, load demand is increasing and new
environmental policies are also being legislated to curtail
Variables greenhouse gases emissions. For instance, 160 countries around the
world have signed the Paris agreement to combat against the
Bmax maximum energy capacity of battery climate changes [1]. To fulfil the agreement, nations around the
Bcap required energy capacity of battery globe are planning to reduce their greenhouse gases emissions, e.g.
Bopt optimal energy capacity of battery Kingdom of Saudi Arabia has planned to reduce CO2 emissions by
Bchg − dcg battery charging/discharging factor 130 million tons a year by 2030. So, a reduction in the utilisation of
CT total cost the fossil fuels and search for their substitutes is crucial for reliable
Cc initial investment cost and eco-friendly electricity generation. Renewable energy (RE)
Com operation/maintenance cost sources can be used as a substitute for the fossil fuels as they

IET Renew. Power Gener., 2018, Vol. 12 Iss. 1, pp. 72-80 72

© The Institution of Engineering and Technology 2017

Fig. 1 Block diagram of the optimisation algorithm
generate green energy and freely available in large abundance. The
integration of RE sources in power system has evolved into the
concept of microgrid (MG). MGs are state-of-the-art active
distribution networks employing distributed generators, energy
storage system (ESS) and loads, operated in grid-connected or
islanded mode, in a controlled, coordinated way [2].
Solar photovoltaic (PV) and wind turbine (WT) are two highly
emerging RE sources, especially solar has attained more popularity
due to the dramatic reduction in its cost over the last few years.
Output powers of such sources are less controllable as compared to
conventional generators. As RE sources harvest energy from
natural resources which are uncontrollable while conventional
generators utilise fossil fuels, which are fully in control of system
operator. However, the output of the RE sources can be controlled
to some extent by utilising advance power electronic converters [3–
6]. Since the outputs of RE sources are less controllable,
intermittent, highly variable and stochastic, their integration in
power system poses threats to its effective operation, principally at
larger penetration levels, e.g. poor load following, load mismatch,
inferior power quality, frequency deviation, voltage instability and
reliability issues are the detrimental effects that RE sources
introduce in power system [7].
A potential candidate solution to the aforementioned problems
is to store electrical energy during surplus generation hours using
an ESS and discharge it appropriately later when needed [8–12].
The use of ESS provides support to RE sources, improves
reliability and power quality, reactive power control, power factor
correction, voltage control, load levelling, load following and damp
out of the power and frequency oscillations [13]. Several types of
ESSs are available and among them, the battery ESS (BESS) is the
most promising and frequently utilised [14–18]. It is also important
to note that the cost of an MG employing RE sources and BESS is
higher and too high cost is prohibitive to commercial and industrial
acceptance, so a method for optimising the size and operation of
such MG to fit application constraints is a crucial task.
Different methods have been developed for capacity
optimisation of PV, WT and BESS, and we will briefly discuss
some of them. In [19], an iterative technique based upon
minimisation of the total annual cost is proposed for sizing of a
standalone PV, WT and BESS system. In [20], optimal sizes of PV,
WT and BESS are calculated based upon multiple-objectives, i.e.
high supply reliability, minimisation of cost and full utilisation of
complementary characteristics of wind and solar. In [21], optimal
sizing of hybrid PV–WT generation system is done based upon the
reliability and cost. In [22], optimal sizes of PV, WT and BESS are
determined based upon cost, reliability and emissions, and well
known optimisation technique, i.e. particle swarm optimisation
(PSO) (see [23] for PSO) is used to solve the optimisation problem.
In [24], PSO and fuzzy logic are used to determine optimal sizes
and types of distributed generators and optimal capacity of ESS. In
[25], optimal sizes of PV, WT and BESS are determined based on
minimising the gap between load and demand. In [26], optimal
sizing of PV, WT and BESS is done based upon the minimisation
of total present cost. In [27], capacity optimisation of hybrid
system, employing PV, WT diesel generator and battery, is done
based upon the minimisation of life cycle cost, CO2 emissions and
dump energy. In [28], a methodology for capacity optimisation of
RE sources and ESS is proposed based upon the minimisation of
initial investment and operation/maintenance costs. In [29], optimal
capacities of PV, WT and BESS are determined based upon the
annualised cost minimisation. In [30], sizing of standalone MG
IET Renew. Power Gener., 2018, Vol. 12 Iss. 1, pp. 72-80
© The Institution of Engineering and Technology 2017
system employing PV and BESS is done based upon the levelised
cost of energy.
In this paper, a novel optimisation strategy is proposed based
upon two advanced constraint based iterative search algorithms for
optimal sizing of PV, WT and BESS in a grid-connected MG. The
process of finding the optimal capacities is completed in two
stages. In the first stage, the proposed algorithm generates a search
space containing every possible combination of PV modules and
WTs starting from a suitable minimum value to a reasonable
maximum value. While in the second stage, optimal BESS capacity
is calculated corresponding to each solution in the given search
space, finally, optimal capacities are determined on the basis of
reliability and cost ratio. The novelty of the proposed approach lies
in the fact that it considers cost, reliability, utilisation factor of
BESS and forced outage rates of PV and WT simultaneously to
determine the optimal solution. All chosen parameters have a
special significance in the domain of RE sources and BESS
capacity sizing methodologies, e.g. system with the lower cost and
higher reliability is considered to be efficient, while higher
utilisation factor avoids the longer idle time of BESS. Similarly,
forced outage rates associated with power conversion devices of
PV and WT are considered which affects the availability of their
output power, and such measures, in fact, make the whole system
more realistic and practical.
As a case study, the proposed methodology is tested using
hourly solar irradiation, wind speed and load demand data of
Dammam city which is situated in the eastern region of Saudi
Arabia. A detailed comparison has been provided based on the
chosen parameters and it was observed that the proposed
methodology is reliable enough as it searches each and every
possible solution.
The remaining of this paper is organised as follows. Section 2
defines the problem description while Section 3 presents the
structure and mathematical modelling of the grid-connected MG.
Section 4 discusses the proposed methodology. Section 5 contains
results and discussions while the conclusion is given in Section 6.
2 Problem description
Most of the conventional power generating stations utilise fossil
fuels to supply the load demand. These power stations have a
major role in global warming. In addition, legislation for reduction
of greenhouse gases impels the utilities to shift their generation
from fossil fuels to renewables. High cost, low efficiency,
intermittent nature and less controllability limit the utilisation of
renewables. Moreover, the storage system is also required for
reliable operation of RE sources [31]. The advantages of both
conventional and RE sources can be combined by integrating RE
sources with the conventional grid. However, sizing of RE sources
and BESS in a grid-connected system is essential for reduction of
emissions at a reasonable cost.
Block diagram of the proposed methodology for optimal sizing
of grid-connected MG is shown in Fig. 1. Initially, the MG is
assumed to be in a standalone mode for optimal sizing of PV, WT
and BESS. The problem is divided into two steps, i.e. sources
sizing and storage sizing. Sources sizing algorithm (SSA) first
forms a search space by using wind power, solar power and
demand data and then reduces it based on given constraints.
Whereas, battery sizing algorithm (BSA) finds the optimal capacity
of BESS in terms of energy (MWh) and power (MW) for the
reduced search space. The BSA determines the optimal solution
based on reliability to cost ratio.
3 MG modelling
In this study, a grid-connected MG is shown in Fig. 2. Both WT
and PV are utilised for hybrid renewable power generation while
battery technology is employed for storage of electrical energy. A
hybrid PV–WT generation topology utilises both solar and wind to
harvest maximum of the available energy. In addition, it is more
reliable and efficient and requires less storage capacity than solar
or wind alone making it more economical [14]. The WT and PV
are connected to generation bus via AC/AC and DC/AC
converters, respectively. However, BESS is connected to
73
Fig. 2 Grid-connected MG system

Table 1 Monthly shape and scale factors The cut-in speed is the minimum speed required by WT to generate
Month σ c power. However, the cut-out is the maximum speed allowed for
January 2.40 4.77 power generation beyond which WT is shutdown to avoid damage.
From (4), WT generates rated power between vr and vco whereas
February 2.45 4.85
the power output increases linearly with speed between vci and vr.
March 2.55 5.15
April 2.40 5.06
3.2 Solar power system
May 2.40 5.52
June 2.60 6.51 A solar PV system harvests electrical energy from solar energy.
July 2.50 5.54 The power output of PV module depends upon the area of PV
August 2.30 4.91 module, solar irradiation, atmospheric temperature and efficiency
of PV module. In order to extract the maximum power, it is
September 2.20 4.18
assumed that a maximum power point tracker is installed. The
October 2.05 4.09 maximum hourly power output is calculated using the following
November 2.20 4.38 equation [34]:
December 2.00 4.68
PPV t = ηPV AI t 1 − 0.005 T o t − 25 ∀t > 0 (5)

generation bus via bidirectional DC/AC converter. The load bus is
connected to generation bus via a transformer. The utility grid is
also connected to load bus via a controlled switch.
3.1 Wind power system
A wind power system converts the kinetic energy of wind into
electrical energy. Hourly wind speed can be estimated using
Weibull distribution function
σ and so on. In this work, sodium sulphur (NaS) type battery energy
v storage is considered. NaS is one of the batteries used for
F v = 1 − exp −
c commercial electrical energy storage in electric utility distribution
v = c −ln 1 − r 1/ σ
v = c −ln r 1/ σ
where σ is the scale factor, c is the shape factor, r is a uniform
random number and v is the wind speed. The values of shape and
scale factors of Weibull distribution function can be obtained from
historical data [32]. The monthly shape and scale factors of
Dammam region are shown in Table 1. The output power of WT
can be calculated using following equation [33]:
0 v < vci
v − vci
Prated × vci ≤ v < vr
PWT v = vr − vci
Prated vr ≤ v < vco
0 v ≥ vco
where Prated is the rated power of WT, PWT is the power output of
WT, vr is rated speed, vci is cut-in speed and vco is cut-out speed.
where A is the area of the solar cell array in m2, ηPV is the
efficiency of PV array, I is the solar irradiation in W/m2 and T o is
the atmospheric temperature in °C.
3.3 Battery ESS
The BESS is composed of series and parallel strings of batteries.
Several types of battery energy storage technologies are available
having different chemistry, depth of discharge, number of cycles
(1)
grid support, wind power integration and high-value grid services.
(2) Its applications include load levelling, peak shaving and power
quality as well as RE management and integration. The state of
(3) charge of the BESS system at any instant of time is calculated
using the mathematical model given below:
EB t = EB t − 1 + ηB × ∑ EG t − EL t ∀t > 0 (6)
Subjected to the following constraint:
1 − DOD ≤ EB t ≤ 1 ∀t > 0 (7)
where ηB is the charge/discharge efficiency, DOD is the depth of
discharge, EB is the energy of BESS, EG is the energy supplied by
(4) both RE sources and BESS and EL is the load energy demand.
3.4 Grid model
As mentioned earlier, the output power of RE sources is sporadic,
so a condition can arise during which output power of the RE
sources and BESS becomes inadequate to supply the required load

74 IET Renew. Power Gener., 2018, Vol. 12 Iss. 1, pp. 72-80

© The Institution of Engineering and Technology 2017
 j
0 GWT (t) < FORWT
j
WTstatus t = ∀t > 0 (11)
1 otherwise

where
j
GWT(t) = rand

Subjected to following capacities constraints:

min
NPV i
≤ NPV max
≤ NPV (12)

min
NWT i
≤ NWT max
≤ NWT (13)

where PG is the hybrid power generated by PV–WT system, NPV

Fig. 3 Normalised average daily load demand of Dammam of calendar
year 2015
demand. During such events, utility grid acts as a backup and sells
power to MG in order to meet the demand. The power supplied by
the utility grid is modelled as follows:
Pgrid t = PL t − ∑ PWT t , PPV t , PBES t (8)
and NWT are the number of PVs and WTs, PPV and PWT are the
powers generated by single PV and WT, PVstatus and WTstatus are
the statuses of PVs and WTs which decide whether they would
generate power or not. When the value of PVstatus of a solar PV is 0
that means the solar PV cannot generate power because of some
fault or any other reason. As given in (10) and (11), the values of
PVstatus and WTstatus are calculated using the forced outage rates
FORPV and FORWT. GPV and GWT are the random numbers which
are generated using rand() command of MATLAB. NPV min
, NWT
min
,
NPV and NWT are the minimum and maximum number of PVs and
max max
WTs which are calculated using following relations:

where Pgrid is the power supplied by utility grid, PL is the load n

∑t = 1 αPL t
min
power demand and PBES is the power supplied by BESS. NPV = n (14)
∑t = 1 PPV t
3.5 System demand n
min ∑t = 1 βPL t
Dammam city lies in Eastern province of Saudi Arabia. The NWT = n (15)
residential load demand of this city is used to test the proposed ∑t = 1 PWT t
methodology. The normalised daily average demand of Dammam,
KSA in the year 2015 is shown in Fig. 3. The minimum average max
n
∑t = 1 γPL t
daily power demand occurs from 6 am to 8 am, and it increases NPV = n (16)
afterward, and peak power demand appears at 8 pm. It should be ∑t = 1 PPV t
noted that the residential load demand data is considered that is
why peak load demand occurs in the evening. max
n
∑t = 1 ρPL t
NWT = n (17)
∑t = 1 PWT t
4 Proposed methodology
The proposed methodology is divided into two sections. The first where α, β, γ and ρ are scaling factors and n is the total number of
section discusses the sizing of RE sources, i.e. solar PV and WT intervals. The instantaneous error/gap between load and generation
using the SSA, whereas Section 2 determines the optimal capacity is calculated as follows:
of BESS, i.e. MW and MWh and the optimal solution using the
BSA. Δp(i, j) t = PL t − PG(i, j)(t) ∀t > 0 (18)

4.1 Sources sizing algorithm where Δp is the instantaneous error. The sum of absolute values of
all instantaneous errors is referred as cumulative error ΔP and
The hybrid power generated by an MG utilising solar PV and WT
is modelled as follows: calculated as
n
PG(i, j) t = NPV
i j
PPV t PVistatus t + NWT j
PWT t WTstatus t ΔP(i, j) t = ∑ |Δp(i, j) t | ∀t > 0 (19)
(9) t=1
∀i ∈ 1, imax , j ∈ 1, jmax , t > 0

where where ΔP(i, j) is the cumulative error corresponding to NPV i

and
NWT. A smaller value of cumulative error reflects that the
j

0 i
GPV (t) < FORPV intermittent generation follows load demand effectively, while a
PVistatus t = ∀t > 0 (10) larger value of cumulative error implies that there is a significant
1 otherwise
error/gap between hybrid renewable power generation and load
demand. The cumulative error for every possible combination of
where PV and WT is calculated and stored in a matrix as follows:
i
GPV (t) = rand
ΔP(1, 1) ⋯ ΔP(1, jmax)
and ΔP = ⋮ ⋱ ⋮ (20)
(imax, 1) (imax, jmax)
ΔP ⋯ ΔP (imax × jmax)

IET Renew. Power Gener., 2018, Vol. 12 Iss. 1, pp. 72-80 75

© The Institution of Engineering and Technology 2017
where ΔP is the error matrix that contains the values of all possible
cumulative errors. The values of NPV and NWT that correspond to
each cumulative error are stored in the vectors NPV and NWT as

T
min
NPV = NPV ⋯ max
NPV (imax × 1) (21)

min max
NWT = NWT ⋯ NWT (1 × jmax) (22)

A search space is generated using ΔP, NPV and NWT

0 NWT
Sspace = (23)
NPV ΔP (imax + 1) × ( jmax + 1)

where Sspace is the search space. The Sspace contains all possible
combinations of PV and WT and the cumulative errors that
correspond to each combination. The Sspace is reduced by choosing
the minimum value of ΔP from its every column. Mathematically

j
1
ΔPmin = ΔPmin ⋯ max
ΔPmin (1 × jmax) (24)

where
j
ΔPmin = min Sspace z, j ∀j
(25)
z = 2, . . . , imax + 1

where ΔPmin is the vector which contains the minimum values

(selected from the Sspace) of cumulative errors. The values of NPV
and NWT that correspond to each ΔPmin are given as
Fig. 4 Algorithm 1: SSA
T
(26)
1 j
max
NPVmin = NPV min
⋯ NPV min ( jmax × 1)
the excess energy is determined by using the algorithm given
jmax below. This algorithm represents the operation of BESS based
(27)
1
NWTmin = NWTmin ⋯ NWT min (1 × jmax) upon which maximum capacity, that can be installed, has been
calculated
A reduced search space is generated using ΔPmin, NWTmin and NPVmin
u u
given below: Pgap ℏ Pgap ℏ >0
u
Pgap ℏ = ∀ℏ > 0 (31)
T T
0 otherwise
RSspace = ΔPmin NPVmin NWTmin jmax × 3 (28)
where
where RSspace is the reduced search space. There are jmax
ℏ
combinations in RSspace for which there will be jmax optimal u
Pgap ℏ = ∑ pgap
u
t (32)
storage units, one for each combination. The SSA is explained in t=1
Algorithm 1 (see Fig. 4).
The maximum storage size required to store the maximum of the
4.2 Battery sizing algorithm excess energy is calculated as
The power generated by case number u selected from RSspace can u
Bmax u
= max Pgap (33)
be calculated using following equation
where Bmax is the maximum energy capacity of the BESS. The
PGu (t) = NPV
u u
P (t) + NWT
min PV
P (t)
min WT
(29) required battery storage capacity is less than or equal to Bmax which
∀u ∈ 1, jmax , t > 0 is calculated using the following relation:

Subjected to the following constraints: u

Bmax x=1
u
Bcap = (34)
u
1
NPV u
≤ NPV max
≤ NPV
j CBS x=0
min min min

1
NWT u
≤ NWT max
≤ NWT
j where Bcap is referred as required energy capacity of BESS. The
min min min
Bcap equals to Bmax if the battery discharges fully after its full
where PG is the power generated by the PVs and WTs that are charging period, this condition is indicated by x = 1. The Bcap is
selected for RSspace. The difference in generation and demand is equal to corrected battery size (CBS) if the battery does not
calculated as follows: discharge fully after its full charging period, this condition is
indicated by x = 0. The condition x = 0 implies that the battery is
over-sized. An iterative region reduction algorithm is used to
u
pgap(t) = PL(t) − PGu (t) ∀u, t > 0 (30)
calculate CBS which is given by the following equations:
where pgap is the gap between generation and demand. The
maximum capacity of BESS that can be employed to store all of

76 IET Renew. Power Gener., 2018, Vol. 12 Iss. 1, pp. 72-80

© The Institution of Engineering and Technology 2017
 u
Smin u
(w) + Smax (w) STEPub = OCFumax(w) − OCFumin(w) (45)
CBSu(w) = (35)
2
where STEPb is the error, OCFmin is the lower boundary of
where solution, while OCFmax is the upper boundary of solution. Initially,
OCFmin and OCFmax are set to 0 and Bcap, respectively, both get
u
Smin (w) x=1 updated in an iterative fashion until STEPb becomes less than the
u
Smin (w + 1) = (36)
CBS(w) x=0 allowable tolerance e. The process of calculation of Bopt is repeated
for every selected combination of PV and WT.
and Cost and reliability are two important parameters that can be
used to assess the performance of a system. An electric power
u
Smax (w) x=0 system having higher reliability (energy served) and lower cost can
u
Smax (w + 1) = (37) be considered to have better performance. The optimal solution in
CBS(w) x=1 this study is determined on the basis of cost and reliability. The
total generation after finding the suitable size of the BESS for
STEPua = Smax
u u
(w) − Smin (w) (38) combination number u is calculated as follows:

where Smin and Smax are the minimum and maximum boundaries of u
PGT u
(t) = NPV u
P (t) + NWT u
P (t) + PBES (t) (46)
min PV min WT
the solution, respectively, STEPa is the difference between the
minimum and maximum boundaries and w is the iteration number. where PBES is the power supplied by BESS and PGT is the total
Initially, Smin and Smax are set equal to 0 and Bmax, respectively. power generated by the MG. A reliability index, i.e. energy served,
Both of them get updated in an iterative fashion until STEPa is the total demand that system serves during its operation
becomes less than or equal to the allowable tolerance e. The Bcap is
the required size but it is not optimal. If it is installed, the system n

will be reliable but not efficient and economical. ESu = ∑ D(t)u (47)
An important parameter that is required to be checked is the t=1

utilisation factor; a higher value of the utilisation factor decreases

idle time and guarantees maximum benefit from BESS. Therefore, where
utilisation factor is used to determine the optimal capacity of u
BESS. To compute utilisation factor, battery decision variable PL(t) PGT (t) ≥ PL(t)
(BDV) is calculated as follows: D(t)u =
u
∀t > 0 (48)
PGT otherwise
u
∑ Bchg − dcg(t) where ES is the energy served. Similarly, another important
BDVu = ∀t > 0 (39)
n reliability index is energy not served and it is the sum of total
demand that is not served by a system during its operation.
where Mathematically, energy not served is modelled as
u u
1 |Pgap (t) − Pgap (t − 1) | ≥ λPbumax n
u
Bchg − dcg(t) = ∀t > 0 (40) u
ENS = ∑ G(t)u (49)
0 else t=1

where Bchg − dcg is the BESS charging/discharging factor, it equals to where

1 when the charging/discharging power is ≥ λPbmax, λ is a constant
and its value is between 0 and 1. Finally, optimal energy capacity u
PL(t) − PGT u
PL(t) > PGT
of BESS is calculated as follows: G(t)u = ∀t > 0 (50)
0 otherwise
u
Bcap BDVu ≥ Blim where ENS is energy not served. The present worth factor (PWF) to
u
Bopt = (41)
BCSu else calculate the net discounted energy (NDE) served can be defined as

where Bopt is the optimal energy capacity of BESS and Blim is the 1+d l−1
PWF = (51)
minimum BESS utilisation limit. The Bopt equals to battery d 1+d l
corrected size (BCS) if BDV is less than Blim, otherwise Bopt equals
to Bcap. The BCS is determined using a region reduction iterative and NDE served is calculated as
algorithm as given below:
NDEuS = ESuPWF (52)
OCFumin(w) + OCFumax(w)
u
BCS (w) = (42)
2 where NDES is the NDE served, d is the discount rate and l is the
year of operation. The total cost CT is the sum of investment cost,
where operation/maintenance cost and replacement cost

OCFumin(w) BDVu ≥ Blim CTu = Ccu + Com

u u
+ Crep (53)
OCFumin(w + 1) = (43)
u
BCS (w) else
where Cc is the initial investment cost, Com is operation/
and maintenance cost and Crep is the replacement cost. The operation/
maintenance cost consists of fixed and variable operation/
maintenance costs. The economic data of RE sources and BESS is
BCSu(w) BDVu ≥ Blim
OCFumax(w + 1) = (44) tabulated in Table 2. The cost of generation is calculated as
OCFumax(w) else

IET Renew. Power Gener., 2018, Vol. 12 Iss. 1, pp. 72-80 77

© The Institution of Engineering and Technology 2017
Table 2 Economic data
Source Cc, $/kWh*, CM (fixed), CM (variable), Life, y
type $/kW $/kWh/y $/kWh/y
WT 2346 33 0 20
PV 2025 16 0 20
BESS* 450 10 0 10

Fig. 7 SV

where SV is the solution vector. There is an optimal combination

of solar PV, WT and BESS capacities corresponding to every index
of SV. While the optimal solution corresponds to the index of the
maximum value of the SV, as the maximum value of an SV implies
that the MG has higher reliability at a relatively lower cost. From
(55), it seems that the value of an ODV will also be maximum for
lower cost and lower reliability, but in fact for lower investments,
i.e. for smaller capacities of PV, WT and BESS, the value of
reliability is very small and with the increase in the investment
cost, i.e. increasing the capacities of PV, WT and BESS the
percentage increase in the reliability is more as compared to
percentage increase in the investment cost. The BSA is explained
in Algorithm 2 (see Fig. 5).

Fig. 5 Algorithm 2: BSA 5 Results and discussions

The BSA has selected 1000 different possible combinations of PV
and WT for reduced search space. The BESS capacity for each
selected combination is calculated using the BSA. Solar PV, WT
and BESS capacities combinations are shown in Fig. 6. As
mentioned earlier, reliability and cost are the two main parameters
that can be used to assess the performance of an MG. The BSA has
carried out reliability and economic analyses of the combinations
selected for reduced search space to determine the optimal
solution. The SV as in (56) generated by the BSA is shown in
Fig. 7. It is important to note that for each index of SV there is a
combination of WT, PV and BESS, and overall installed capacities
of RE sources and BESS increase with the increase in the index.
The optimal combination corresponds to the index with a
maximum value of ODV. Because a higher value of ODV implies
relatively higher reliability at a reasonable lower cost, while a
smaller value of ODV shows higher reliability at a very high cost
or low reliability at a lower cost. It can be observed from Fig. 7
that the combination of PV, WT and BESS corresponding to the
Fig. 6 Combinations of WT, solar PV and BESS capacities index number 187 of the SV is the optimal solution. The capacities
of the PV, WT and BESS are 57 MW, 187 MW and 63 MWh (20
N
∑k = 1 NPVuk MW), respectively.
Cgu = (54) There is an increase at the beginning until the maximum point
NDEuS and then decreasing trend is observed for the values of ODV.
Smaller values of ODV in the beginning are due to the fact that
where Cg is the cost of generation of MG, NPV is the net present energy served is very little for smaller capacities of PV, WT and
value of total cost. The optimal decision variable (ODV) is defined BESS. Similarly, values of ODV are also small for higher indices
as follows: because for largely installed capacities the cost becomes very high
which makes the solution un-economical. Although, reliability
Esu (energy served) is also higher for larger capacities its impact on
ODVu = (55) ODV is non-dominant as compared to cost, as costs of RE sources
Cgu
and storage are high and for very large capacities overall cost
becomes, even more, higher which makes the solution very
The solution vector (SV) is generated as
expensive. Moreover, from Figs. 6 and 7 it can also be observed
that appropriate limits of minimum and maximum capacities of PV
SV = ODV1 ⋯ ODV jmax − 1 ODV jmax (1 × jmax) (56) and WT should be selected while determining the optimal size
using any technique. The narrow limits may lead the algorithms to

78 IET Renew. Power Gener., 2018, Vol. 12 Iss. 1, pp. 72-80

© The Institution of Engineering and Technology 2017
operate in the region where the values of ODV s are low and result
in an un-economical solution.
As mentioned earlier, that the output of the RE sources is
intermittent, so during the operation of the MG it may happen that
the output of the MG becomes insufficient to meet the required
demand, in such events, the MG buys power from the utility grid to
supply the load. Total energy served by the MG and the utility grid
to meet the required load is shown in Fig. 8. It can be seen that
with the increase in the indices the energy served by the MG
increases while energy served by the utility grid decreases. In the
beginning, energy served by the MG rises and energy served by the
utility grid reduces rapidly and finally both saturate.
The variation in demand and generation throughout a year is
shown in Fig. 9. It can be observed that the generation is always
equal to the demand. As the system under study is grid-connected,
and utility grid is serving as a backup. So, whenever the output
power of MG becomes inadequate to supply the required load
demand, MG buys power from the utility grid and in this way the
generation remains always equal to demand was made the overall
system highly reliable.
Fig. 8 Energy share of MG and main grid
To show that the global optimum can be reached, the sizing
problem is formulated and solved using genetic algorithm also. The
optimal capacities of PV, WT and BESS determined by genetic
algorithm are 57.3 MW, 187 MW and 63 MWh. Hence, both
algorithms have found the same solution, the small deviation in the
size of PV is due to randomness factor. Although both algorithms
resulted in the same solution computational time of the proposed
algorithm is lesser and guarantees global maxima. Moreover,
unlikely to most of the conventional optimisation algorithms, in the
proposed algorithm user can visualise that how the solution is
being determined, and the user can communicate to the algorithm
to determine the required solution in an easy and understandable
way. Also, there is no parameter tuning involved in the proposed
algorithm like the most of the conventional optimisation
approaches, which are based upon hit and trial methods.
To show the effectiveness of the proposed methodology, a
comparison based upon per unit cost, clean energy supplied
(energy supplied by MG) and CO2 emissions between different
possible solutions is presented in Table 3. Case I represents that all
energies are supplied by conventional generation system. The per
unit cost is the lowest while emissions are the highest for this case.
The remaining cases represent that energy is supplied by both MG
and utility. The overall cost for case II is reasonable and emissions
are also moderate for this case. The clean energy served by case III
is more as compared to case II. We can see from Table 3 that while
moving from cases II to III the percentage increase in the clean
energy served is more as compared to the percentage increase in
overall cost. Also, the percentage decrease in CO2 emissions is
more as compared to the percentage increase in cost. Hence, we
can conclude that case III is better than case II. The CO2 emissions
of case III are almost 70% less as compared to case I, i.e.
conventional generation. The CO2 emissions are the lowest for
case V whereas per unit cost is very high for this case due to
largely installed capacities of RE sources and BESS. The overall
cost per unit for case IV is also higher and CO2 emissions are
lower. It is clear from the above discussion that case III is an
optimal solution as it serves a considerable amount of clean energy
and reduces CO2 emissions by a substantial amount at reasonable
per unit cost, i.e. 16.18 c/kWh. It is important to note that the cost
is the factor/function of many other parameters, e.g. a location with
a higher correlation between the solar irradiation, wind speed curve
and load curve would result in further reduction in cost, and our
optimised solution will then be more effective and justified.
MATLAB is used for the modelling and simulations. Core i7,
6th generation, 2.6 GHz, 16 GB RAM, platform is used to run the
simulations.

6 Conclusion
In this paper, a technique has been proposed for the capacity
optimisation of RE sources, i.e. WT and solar PV and associated
BESS in grid-connected MG system. The optimal capacities are
determined based on the energy served to cost per unit ratio. It has
been shown that the optimal solution is economical and produces
less CO2 emissions. A comparison based on the cost and emissions
for some selected cases is also presented and it is proved that the
optimal solution is superior to the other solutions. The significance
of the proposed method lies in the fact that it eschews under-sizing
and over-sizing because it searches every possible solution. Also,
to make the methodology more practical, forced outage rates of
Fig. 9 Variation in generation and demand throughout the year
WT and PV and utilisation factor of BESS are considered. The

Table 3 Comparison between different possible solutions

Case index PV, WT, BES, BES, MG energy, Utility energy, MG cost, Total cost, CO2,
MW MW MWh MW GWh GWh c/kWh c/kWh kt
case I — — — — — 478.8 — 10 527.77
case II 82 100 53 17 268.9 210 15.91 13.32 231.65
case III 57 187 63 20 339.3 139 18.73 16.18 153.33
case IV 40 250 94 30 369.8 109 21.11 18.58 120.24
case V 10 650 133 43 422.8 56 41.67 37.96 61.73

IET Renew. Power Gener., 2018, Vol. 12 Iss. 1, pp. 72-80 79

© The Institution of Engineering and Technology 2017
proposed methodology is quite general and it can be extended to
other types of generation and storage technologies and also equally
valid for any geographical location.
7 Acknowledgments
The authors acknowledge the support provided by the Deanship of
Research (DSR) at King Fahd University of Petroleum & Minerals
(KFUPM) for funding this work through Project Number
IN161038. Also, they thank Research Institute (RI) at KFUPM and
Saudi Electricity Company (SEC) for providing renewables and
demand data to carry out this research. In particular, they
appreciate the support of Mr. Hussain A. Al-Harthi for arranging
time series data.
8 References
[1] http://unfccc.int/paris_agreement/items/ 9485.php, accessed August 2017
[2] Kuznetsova, E., Ruiz, C., Li, Y.F., et al.: ‘Analysis of robust optimization for
decentralized microgrid energy management under uncertainty’, Int. J. Electr.
Power Energy Syst., 2015, 64, pp. 815–832
[3] Carrasco, J.M., Franquelo, L.G., Bialasiewicz, J.T., et al.: ‘Power-electronic
systems for the grid integration of renewable energy sources: a survey’, IEEE
Trans. Ind. Electron., 2006, 53, (4), pp. 1002–1016
[4] Ram, J.P., Babu, T.S., Rajasekar, N.: ‘A comprehensive review on solar PV
maximum power point tracking techniques’, Renew. Sustain. Energy Rev.,
2017, 67, pp. 826–847
[5] Chen, Z., Guerrero, J.M., Blaabjerg, F.: ‘A review of the state of the art of
power electronics for wind turbines’, IEEE Trans. Power Electron., 2009, 24,
(8), pp. 1859–1875
[6] Blaabjerg, F., Teodorescu, R., Chen, Z., et al.: ‘Power converters and control
of renewable energy systems’. Proc. 6th Int. Conf. Power Electron, 2004, pp.
1–20
[7] Kyritsis, A., Voglitsis, D., Papanikolaou, N., et al.: ‘Evolution of PV systems
in Greece and review of applicable solutions for higher penetration levels’,
Renew. Energy, 2017, 109, pp. 487–499
[8] Khalid, M., Savkin, A.V.: ‘A model predictive control approach to the
problem of wind power smoothing with controlled battery storage’, Renew.
Energy, 2010, 35, (7), pp. 1520–1526
[9] Thapa, S., Karki, R.: ‘Reliability benefit of energy storage in wind integrated
power system operation’, IET. Gener. Transm. Distrib., 2016, 10, (3), pp.
807–814
[10] Denholm, P., Margolis, R.M.: ‘Evaluating the limits of solar photovoltaics
(PV) in electric power systems utilizing energy storage and other enabling
technologies’, Energy Policy, 2007, 35, (9), pp. 4424–4433
[11] Khatamianfar, A., Khalid, M., Savkin, A.V., et al.: ‘Improving wind farm
dispatch in the Australian electricity market with battery energy storage using
model predictive control’, IEEE Trans. Sustain. Energy, 2013, 4, (3), pp. 745–
755
[12] O'Dwyer, C., Ryan, L., Flynn, D.: ‘Efficient large-scale energy storage
dispatch: challenges in future high renewable systems’, IEEE Trans. Power
Syst., 2017, 32, (5), pp. 3439–3450
[13] Poonpun, P., Jewell, W.T.: ‘Analysis of the cost per kilowatt hour to store
electricity’, IEEE Trans. Energy Convers., 2008, 23, (2), pp. 529–534
[14] Khalid, M., Ahmadi, A., Savkin, A.V., et al.: ‘Minimizing the energy cost for
microgrids integrated with renewable energy resources and conventional
80
generation using controlled battery energy storage’, Renew. Energy, 2016, 97,
pp. 646–655
[15] Xiaokang, X., Bishop, M., Oikarinen, D.G., et al.: ‘Application and modeling
of battery energy storage in power systems’, CSEE J. Power Energy Syst.,
2016, 2, (3), pp. 82–90
[16] He, G., Chen, Q., Kang, C., et al.: ‘Cooperation of wind power and battery
storage to provide frequency regulation in power markets’, IEEE Trans.
Power Syst., 2017, 32, (5), pp. 3559–3568
[17] Deeba, S.R., Sharma, R., Saha, T.K., et al.: ‘Evaluation of technical and
financial benefits of battery-based energy storage systems in distribution
networks’, IET Renew. Power Gener., 2016, 10, (8), pp. 1149–1160
[18] Savkin, A.V., Khalid, M., Agelidis, V.G.: ‘A constrained monotonic charging/
discharging strategy for optimal capacity of battery energy storage supporting
wind farms’, IEEE Trans. Sustain. Energy, 2016, 7, (3), pp. 1224–1231
[19] Kellogg, W., Nehrir, M., Venkataramanan, G., et al.: ‘Generation unit sizing
and cost analysis for stand-alone wind, photovoltaic, and hybrid wind/PV
systems’, IEEE Trans. Energy Convers., 1998, 13, (1), pp. 70–75
[20] Xu, L., Ruan, X., Mao, C., et al.: ‘An improved optimal sizing method for
wind-solar battery hybrid power system’, IEEE Trans. Sustain. Energy, 2013,
4, (3), pp. 774–785
[21] Yang, H., Lu, L., Zhou, W.: ‘A novel optimization sizing model for hybrid
solar-wind power generation system’, Sol. Energy, 2007, 81, (1), pp. 76–84
[22] Wang, L., Singh, C.: ‘PSO-based multi-criteria optimum design of a grid-
connected hybrid power system with multiple renewable sources of energy’.
IEEE Swarm Intelligence Symp., 2007, pp. 250–257
[23] Clerc, M.: ‘Particle swarm optimization’, vol. 93 (John Wiley & Sons, 2010)
[24] Moradi, M.H., Eskandari, M.H., Mahdi, S.: ‘Operational strategy optimization
in an optimal sized smart microgrid’, IEEE Trans. Smart Grid, 2015, 6, (3),
pp. 1087–1095
[25] Kellogg, W., Nehrir, M., Venkataramanan, G., et al.: ‘Optimal unit sizing for a
hybrid wind/photovoltaic generating system’, Electr. Power Syst. Res., 1996,
39, (1), pp. 35–38
[26] Ahmadi, S., Abdi, S.: ‘Application of hybrid big bang-big crunch algorithm
for optimal sizing of a stand-alone hybrid PV/wind/battery system’, Sol.
Energy, 2016, 134, pp. 366–374
[27] Ogunjuyigbe, A.S.O., Ayodele, T.R., Akinola, O.A.: ‘Optimal allocation and
sizing of PV/wind/split-diesel/battery hybrid energy system for minimizing
life cycle cost, emission and dump energy of remote residential building’,
Appl. Energy, 2016, 171, pp. 153–171
[28] Yang, P., Nehorai, A.: ‘Joint optimization of hybrid energy storage and
generation capacity with renewable energy’, IEEE Trans. Smart Grid, 2014,
5, (4), pp. 1566–1574
[29] Atia, R., Yamada, N.: ‘Sizing and analysis of renewable energy and battery
system in residential microgrids’, IEEE Trans. Smart Grid, 2016, 7, (3), pp.
1204–1213
[30] Lai, C.S., McCulloch, M.D.: ‘Sizing of stand-alone solar PV and storage
system with anaerobic digestion biogas power plants’, IEEE Trans. Ind.
Electron., 2017, 64, (3), pp. 2112–2121
[31] Abeywardana, D.B.W., Hredzak, B., Agelidis, V., et al.: ‘Supercapacitor
sizing method for energy controlled filter based hybrid energy storage
systems’, IEEE Trans. Power Electron., 2017, 32, (2), pp. 1626–1637
[32] Boyle, G.: ‘Renewable energy’ (Oxford University Press, UK, 2004)
[33] Roy, S.: ‘Market constrained optimal planning for wind energy conversion
systems over multiple installation site’, IEEE Trans. Energy Convers., 2002,
17, (1), pp. 124–129
[34] Chen, S.X., Gooi, H.B., Wang, M.Q.: ‘Sizing of energy storage for
microgrids’, IEEE Trans. Smart Grid, 2012, 3, (1), pp. 142–151
IET Renew. Power Gener., 2018, Vol. 12 Iss. 1, pp. 72-80
© The Institution of Engineering and Technology 2017