Energy 123 (2017) 108e118

Contents lists available at ScienceDirect

Energy
journal homepage: www.elsevier.com/locate/energy

Multi objective particle swarm optimization of hybrid micro-grid

system: A case study in Sweden
Maher Azaza*, Fredrik Wallin
€lardalens University, Va
Future Energy Center, Department of Energy, Building and Environment, Ma €sterås, Sweden

a r t i c l e i n f o a b s t r a c t

Article history: Distributed energy resources DERs are small scale energy system which could provide local supply when
Received 3 November 2016 placed at customers' premises. They aggregate multiple local and diffuse production installations, con-
Received in revised form sumer facilities, storage facilities and monitoring tools and demand management. The main challenges
12 January 2017
when assessing the performance of an off-grid hybrid micro-grid system HMGS are the reliability of the
Accepted 29 January 2017
system, the cost of electricity production and the operation environmental impact. Hence the tradeoff
Available online 3 February 2017
between three conflicting objectives makes the design of an optimal HMGS seen as a multi-objective
optimization task. In this paper, we consider the optimization and the assessment of a HMGS in
Keywords:
Hybrid micro-grid
different Swedish cities to point out the potential of each location for HMGS investment. The HMGS
DERs consists of photovoltaic panels, wind turbines, diesel generator and battery storage. The HMGS model
Renewable energy was simulated under one-year weather conditions data. A multi objective particle swarm optimization is
Multi-objective optimization used to find the optimal system configuration and the optimal component size for each location. An
Particle swarm optimization energy management system is applied to manage the operation of the different component of the system
Sensitivity analysis when feeding the load. The techno economics analysis shows the potential of HMGS in the Swedish rural
development.
© 2017 Elsevier Ltd. All rights reserved.

1. Introduction order to enable a broad liberalization of the electricity market and

In recent decades, the climate change and gas emissions are the
greatest challenges to the sustainable development. Therefore,
most states have adopted actions to reduce greenhouse gas emis-
sions resulting in larger increase in distributed generation DG.
Electrical systems in most developed countries are undergoing
significant changes. These changes are the result of electricity
market liberalization and the increase of renewables in the energy
mix [1]. The topology of the electrical system has not changed since
its creation in the early twentieth century. At that time, a central-
ized architecture was implemented: electricity is produced in po-
wer plants of high power, transported between regions through
networks of high voltage and distributed to consumers through low
voltage networks. This centralized architecture was the one that
best met the constraints this time, it also ensured a fast network
development and quality of supply. Thus, most of the country has a
rather centralized electrical system and “vertically integrated”.
However, in recent years, this system begins to be questioned in
* Corresponding author.
E-mail address: maher.azaza@mdh.se (M. Azaza).
increasing the share of electricity generators using renewable en-
ergies while maintaining constant quality of the energy supplied to
consumers.
Also called mini smart grids or intelligent micro-networks, mi-
cro-grids are small power systems, designed to provide a reliable
electricity supply and better quality for a small number of con-
sumers. They can be connected directly to the grid or operated in
island mode. The micro-grids are sometimes called sustainable
communities, because of strong emphasis on energy management
and optimal use of renewable energy resources [2,3]. In the concept
of sustainable city, the environment is not divorced from urban
development projects or economic, cultural or social orientations of
the city. This desire for integration takes into account a develop-
ment in the long term and a global perspective. “Smart Grid Cities”
are the cities designed primarily to manage more effectively their
customer energy consumption. This requires, inter alia, local opti-
mization of energy supplies and energy consumption [4], intelli-
gent network management and a balance between the production
and the consumption to facilitate the integration of renewable re-
sources of energy and new electricity uses (electric vehicles). The
role of new information and communications technology (ICT) is
key here to meet these challenges [5].

http://dx.doi.org/10.1016/j.energy.2017.01.149
0360-5442/© 2017 Elsevier Ltd. All rights reserved.
 M. Azaza, F. Wallin / Energy 123 (2017) 108e118
Nomenclature
a, b coefficients of fuel consumption
COE Cost of electricity
CB Capacity of the battery
CRF Capital recovery factor
DERs Distributed energy resources
DG Diesel generator
DOD Depth of discharge
EG Available power from renewable sources
EL Loads demand
EMS Energy management system
ESS Energy storage system
GA Genetic algorithm
Gref Solar radiation at reference conditions
Due the increasing energy demand, up to 53% by 2035 [6], the
penetration of renewable energy resources has been steadily
increasing over the past decades to remedy the acute issues of high
energy demand and global warming [7,8]. Defined as group of
interconnected loads and DERs operating with clear electrical
boundaries, the HMGS offers a cost-effective solution for remote
areas promoting localized renewable resources [9]. For instance,
HMGS provides electricity supply to remote areas where required
transmission and distribution facilities are not available or expen-
sive to deploy. DERs could be seen as a micro-grid when three
characteristics are met; the electrical boundaries are predefined, it
includes energy management system and the power generation
capacity must exceed the peak critical load [10,11].
The design of a HMGS requires a thorough analysis on decision
making of the optimal power mix and component size as per load
requirement. Various key factors to be considered on the decision
making. First the cost of electricity COE which refers to the cost of
operation to satisfy the load demand. Here a lower COE makes
HMGS investment more profitable. Second the system reliability
need to be at maximum level so that the power delivery in-
terruptions are avoided. Different metrics can be used to describe
the HMGS reliability such the loss of power supply probability LPSP,
described in the next sessions. The third factors to be considered on
HMGS design refers to the environmental impact of system oper-
ations when the conventional sources (e.g. diesel generator) are
used in combination with renewable sources. A conventional en-
ergy source is used as complementary system to the renewable
resources to avoid power interruption when weather conditions are
not favorable for power generation. Then the task of a HMGS here is
to minimize the utilization of non-renewable resources so that is
environmental impact is minimized. The tradeoff between the
aforementioned conflicting factors needs an optimal solution that
balances between the three objectives.
2. Literature review
In the literature several studies show a significant development
in HMGS design, configuration and optimization over the last
decade [12e16] [17]. studied a hybrid photovoltaic PV/wind turbine
WT with hydrogen-based energy storage system ESS system
considering the system components outages and using particle
swarm optimization PSO. The study reveals the impact of compo-
nents outages on the cost of the system and its reliability [18]. used
ant colony algorithm to achieve the minimum power loss and
better load balancing to solve the problem of optimal switching
operation of distributed generation. In Ref. [19] proposed an
109
HMGS Hybrid micro-grid system
HPS Hybrid power systems
LPSP Loss of power supply probability
Pout Output power
Pr Rated power
PRG Production gross of renewable fraction
PSO Particle swarm optimization
PV Photovoltaic
RF Renewable factor
SOC State of charge
Tamb Ambient temperature
Vcut-in Cut-in wind speed
Vcut-out Cut-out wind speed
Vr Rated wind speed
WT Wind turbine
approach to maximize the reliability of PV/WT with battery energy
storage system by minimizing the loss of power supply probability
LPSP using non-dominated storing genetic algorithm [20]. intro-
duced a structural methodology to transform existing radial dis-
tribution network into autonomous micro-grid. The authors used
genetic algorithm GA and PSO to find the optimal sites for the
autonomous operation [21]. proposed a hybrid simulated
annealing-Tabu search method which add-up the advantages of
each method to solve the problem of optimal sizing of an WT/PV/
Diesel generator DG/Biodiesel. The objective of the optimization
considers only the COE [22]. investigates an operating policy to
achieve high renewable energy resources penetration levels
involving WT/PV and battery energy storage system in small scale
island system. the GA is used to optimally size the studied config-
uration based on two objectives the levelized energy cost and the
renewable energy sources penetration levels. In Ref. [23] the au-
thors reported an optimization framework based on the simulated
annealing algorithm to find out the optimal size of DG/battery
system and the optimal charge/discharge battery schedules for a
daily load demand. Besides to the aforementioned methodologies
developed for HMGS design an optimization, software tools are
widely used for HMGS performance study. The commonly used
software tools are HOMER [24] and HYBRID2 [25]. However, the
main drawbacks of this software tools are the black box utilization
[26]. A comprehensive review on the advantages and the limita-
tions of the available software that are used for the analysis of
HMGS is stated in Ref. [27].
Although there are many research works about HMGS design
and optimization [28e35], but there are few works that consider
more than one or two objectives. In this work, a multi objective PSO
based approach is adopted to the design of the optimal HMGS and
to find the tradeoff between three conflicting objectives; the reli-
ability, the cost of operation and the environmental impact. It could
be stated as well that an assessment of the applicability of HMGS in
some specifics regions such as Nordic countries, is still lacking,
where initial believes that these areas are not suited for HMGS
operation due to its weather conditions.
Since the oil crisis of the early 1970s, Sweden has invested
heavily in the search for alternative energy sources. Its ultimate
goal is to completely abandon the use of oil. It is interesting to note
that the fulfillment of his plan is on track since 1979. Subsequently,
in 1997, the current energy policy has been implemented and the
Swedish Energy Agency was created to achieve the intended ob-
jectives. The objectives are quite clear: by 2020, Sweden wants to
increase its use of renewable resources to at least 50% and
completely eliminate the use of oil [36]. Therefore, even though the
110 M. Azaza, F. Wallin / Energy 123 (2017) 108e118

Swedish population is a major energy consumer per capita, about deep techno economical knowledge where HMGS could be a
16 000 kWh per person per year, carbon emissions are compara- potential alternative for energy production in Sweden.
tively low compared to other countries [37]. To do this, it uses (ii) The questions related to the technical configuration of the
mainly hydropower, solar and wind. A second important factor that HMGS and the cost of the energy production are answered in
makes HMGS a crucial solution to achieve the Swedish energy this paper so that it highlights the regions where HMGS based
policy is the sparsely populated areas across the inland regions. solution could be an alternative. (iii) The optimal HMGS
These areas are mainly concentrated in the north and central part of design was selected after multi objective optimization to find
the country [38]. Further, 30 thousand live in deprived areas (iso- the tradeoff between three objectives: the reliability of the
lated islands) where connection to the grid utility is costly opera- system, the cost of operation and the environmental impact.
tion [38,39]. (iii) A sensitivity analysis was carried out to in each studied
This work, will investigate the potential of HMGS application in location to reveal the limitations of the HMGS and the cor-
the Nordic countries, Sweden as case of study, the main contribu- relations between the optimal objectives and different load
tions addressed in this paper are as follow: curves.
(iv) The seasonality effect was studied to have a deep insight on
(i) In this study 12 regions were selected to assess their po- the performance of the system from reliability and produc-
tential for HMGS deployment. The studied areas, as marked tion overview. That will help energy provider on more effi-
in Fig. 1, are located in the inland regions and the cost regions cient energy operations planning depending on the location
across the country from the north part up the south part of across Sweden.
the country. This evaluation will investigate the renewable
energy potential of different Swedish regions and gives a This paper is organized as follow: Section 2 provides a literature
review, section 3 describe the simulation model of the HMGS used
for this study as well the energy management system. Section 4
provides the multi objective particle swarm optimization applica-
tion to the design and the optimal configuration of HMGS. Section 5
gives a sensitivity analysis and point out the impact of seasonality
on the system performance, conclusion is drawn in section 6.

3. Modeling hybrid power system

Hybrid power systems (HPS) involving at least two technologies

complementary: one or more conventional energy resources,
typically diesel generators, and at least one source of renewable
energy [4]. Due to their dependency to the weather conditions, the
power generation from renewable energy resources is stochastic.
Therefore, combination with conventional sources provides more
stable and continuous power output. HPS incorporate basically 3
layers; the load, the power and storage system and an energy
management system EMS, Fig. 2.

3.1. Power system

3.1.1. PV system
Fig. 1. Swedish solar radiation profile. It is commonly believed that the Nordic countries are not

Fig. 2. Hybrid micro-grid system architecture.

 M. Azaza, F. Wallin / Energy 123 (2017) 108e118 111

favorable areas for investment in PV installations. However, this Uppsala

may seem to an odd idea: long and dark daytime during winter 30

Temperature (°C)
making solar energy practically useless. However, the other side of 20
the medal may become remarkable during summer when these 10
areas benefits from long shining daytime much longer hours than
0
the southern part of Europe. As it could be seen in Fig. 1, that a big
part of the country receives on average more than 1600 h of sun- -10
shine per year. Furthermore, a substantial part of Sweden gets over -20
than 925 kWh/m2 of global solar irradiation.

ai

t
ry

r
y

in

ly
il

us

be
be

be
ar
ar

pr

ob
Ju

Ju
ua

ug
M

em
nu

em

em
A
br

ct
The power supplied by the PV generator can be expressed as a

A
Ja

ov

ec
pt
Fe

Se

D
N
Solar Radiation (W/m )
function of the solar radiation and the ambient temperature [40] as

2
800
given:
h  i. 600
Pout ¼ Pr $G$ 1 þ Kt ðTamb þ ð0:0256  GÞÞ  Tref Gref
400
(1) 200

Pout and Pr are the output power of the PV and the rated power 0
under the standard conditions respectively. Solar radiation at

ai

t
ry

er

r
Fe y

in

ly
il

O r
us

be

be
be
ar
ar

pr

Ju

Ju

ob
ua

ug
M

em
nu

em
standard condition Gref ¼ 1000 (W/m2), the ambient temperature

em
A

ct
br

A
Ja

ov

ec
pt
Se
Tamb ¼ 25 ( C). The solar power and the measured hourly solar

D
N
40
radiation and temperature at different locations in Sweden is
graphically shown in Figs. 3e5.

PV Out (kW)
30

20
3.1.2. Wind turbine 10
Wind power is the energy source currently the fastest growing
in the world. In the case of Sweden, this energy production has 0
increased fivefold since 2000 and the early 2010, Sweden had

ry

ai

r
in

ly
Fe y

il

r
us

be

be
be
ar
ar

pr

Ju

Ju

ob
ua

ug
M

em

em
nu

em
A
br

ct
approximately 1400 wind turbines [41]. The wind power genera-

A
Ja

ov
O
pt

ec
Se

D
N
tion is produced from the conversion of the wind kinetic energy Month

Fig. 4. Hourly ambient temperature, solar radiation and solar power in Uppsala, 2015.
Kiruna
40
Malmö
Temperature (°C)

40
Temperature (°C)

20

0 20

-20 0
-40
-20
t
ai
ry

er

D ber

r
Fe y

in

ly
il

O r
us

be

be
ar
ar

pr

Ju

Ju

ob
ua

ug
M

em

pt st

ec er
nu

s
em

em

il

ai
Fe ry
ry

in

ly

r
A

be
be
ar
br

ct

pr

u
M
A

ob

b
Ju

Ju
a
ua
Ja

ov

ec

ug
pt

em
em

em
nu

ct
br
Se

A
Ja

ov
O
800
Solar Radiation (W/m )

Se

D
Solar Radiation (W/m2)

N
2

600 800
400 600
400
200
200
0
0
ai

er

r
ry

in

ly
s

il

r
Fe y

O r
us

be
be

be
ar
r

pr

Ju

Ju

ob
ua
ua

ug
M

p t st

ec er
s
em

il

ai
Fe ry
ry

r
em

em

in

ly
A

be

be
ar
br

ct
n

pr

u
M

ob

b
A

Ju

Ju
a
ua
Ja

M
ov

ec

ug
pt

em
nu

em

em
A

ct
br
Se

D
N

A
Ja

ov
O

40
Se

D
N

40
PV Out (kW)

30
PV Out (kW)

30
20
20
10
10
0
0
ry

ai

r
er
in

ly
s

il

r
Fe y

us

be
be

be
ar
ar

pr

Ju

Ju

ob
ua

ug
M

ec er
s

il

ai

t
Fe ry
ry

r
em

in

ly
nu

em

em
A

Se us

be

be
ar
br

ct

pr

ob

b
Ju

Ju
A

a
ua
Ja

ug
M
O

ov

ec
pt

em
nu

em

em
A

ct
br
Se

D
N

A
Ja

ov
pt

D
N

Month
Month
Fig. 3. Hourly ambient temperature, solar radiation and solar power in Kiruna, 2015.
€ , 2015.
Fig. 5. Hourly ambient temperature, solar radiation and solar power in Malmo
112 M. Azaza, F. Wallin / Energy 123 (2017) 108e118

into electrical energy. 3.1.4. DC/AC inverter

The output power of wind turbine can be expressed as function In most cases, conventional hybrid systems contain two bus
of the rated wind speed Vr, the cut-in Vcut-in and the cut-out wind [44]: a DC bus for the sources, DC loads and the batteries and an AC
speed Vcut-out by the following equations [42]: bus for AC generators and the distribution system. Renewable en-

8
>
< 0;
 .   .  V < Vcutin ; V > Vcutout
Pr V3r  V3cutin  Pr $ V3cutin V3r  V3cutin ; Vcutin  V  Vr (2)
>
:
Pr ; Vr  V  Vcutout

V denotes the wind speed on each time period, Pr is the rated power
of the wind turbine. It is to mention that for small-scale wind
turbine deployment, models with lower cut-in speed is selected.
That allow more efficient operation even at lower wind speed.
In order to represent the wind profile within specific area, the
wind velocity is commonly treated as random variables with the
Weibull distribution [43]. Fig. 6, shows the hourly per year wind
speed distribution in different location in Sweden.
3.1.3. Battery bank
The storage system uses a battery bank to be able to cover the
average charge for several days. Apart from the main charge, a HPS
may also include auxiliary charges (deferred load, shedding load) to
achieve energy balance. If the main load is powered without
interruption, the auxiliary loads are powered by priority, only when
there is a surplus of energy. Thus, in HPS with storage batteries and
auxiliary loads, if there is an excess of energy (from renewable
energy sources and diesel), it will go first in the batteries and then it
will be used to power other auxiliary loads according to their pri-
ority. In such a system, storage batteries play a dual role: source and
load. The sizing of the batteries depends on the days of autonomy N
and (EL e EG) where EL denotes the loads demand and EG the
available power from renewable sources. The battery capacity CB is
expressed as follow:
ðEL  EG Þ
CB ¼ N$
hB $hinv $DOD
where hB and hinv are the battery and the inverter efficiencies, DOD
is the depth of discharge. It is noteworthy that the preceding
expression is valid only when the energy from the PV system wind
turbine is deficient.
ergy sources can be connected to the bus AC or DC depending on
size and system configuration. The interconnection between the
two buses can be achieved through power electronics: Inverter/
rectifier or bidirectional converters.
3.1.5. Diesel generator
In the case of a renewable energy system, the production of
electric energy depends on resources (wind, sun …) and not of the
demand side. For standalone installations, it is therefore necessary
to resort to storage or to add one or more diesel generators. The fuel
consumption of the DG expressed as function of the output power
is given as [45e47]:
FðhÞ ¼ a$Pr þ b$Pout ðhÞ (4)
F(h) is the hourly fuel consumption (L/h), Pr denote the rated power
and Pout(h) is the electrical produced power (kW), a and b are co-
efficients of fuel consumption curve (L/kW) describing the con-
version ratio of the fuel to electrical power and approximated to
0.24 and 0,084 respectively [48].
3.2. Load curve
The load profile is critical component in the design and the
sizing of reliable and efficient micro grid system. c, see Fig. 7 [49],
usually shows a base load, a significant peak in the evening, which
(3)
corresponds mainly to lighting uses, and significant demand in the
morning and the afternoon. The base load generally is related to the
early hours, but can sometimes include an evening consumption.
The maximum peak power seen at night is two to five times higher
than the highest power demand in the base period [49] with a peak
value of 2.8 kW.

Kiruna Göteborg
0.3
0.8

0.25
0.6
Probability Density

Probability Density

0.2

0.4 0.15

0.1
0.2
0.05

0 0
0 2 4 6 8 10 0 2 4 6 8 10
Wind speed(m/s) Wind speed(m/s)

€ terberg.
Fig. 6. Annual wind speed distribution in Kiruna and Go
 M. Azaza, F. Wallin / Energy 123 (2017) 108e118
Fig. 7. Typical Swedish household electrical load [40].
3.3. HMGS energy management
The high cost of the renewable energy technologies and the
unpredictable aspect are the main limitations to more renewable
energy integration. To overcome this challenges an energy man-
agement system is an important component to match the load
demand under variable weather conditions and different time
distribution. The operations scenarios considered to simulate the
EMS are as follow, see Fig. 8:
- RE resources match the load demand and the surplus energy is
stored during times when the production exceeds consumption
via bi-directional inverters.
- The RE resources are insufficient to match the load demand and
the battery storage is also depleted. The diesel generator is
operated to supply the loads and to charge the battery bank.
- RE resources fail to satisfy the load demand. In this case, the
priority is to use the batteries to cover the shortage of power.
- The surplus of produced RE energy exceeds the battery bank
capacity. In this case the extra energy is dissipated by a dump
load if the batteries state of charge SOC reaches the maximum
level.
4. Multi objective optimization
The optimization of the HMGS is a challenging task to come-up
with the optimal system configuration during the design of an
Fig. 8. Operation flowchart of the HMGS.
113
HMGS. In fact, the objectives functions related to the HMGS opti-
mization are nonlinear and conflicting. At the same time the goal is
to increase the system reliability at a lower price and a lower
environmental impact. The optimization problems are classified
according to the mathematical characteristics of the objective
functions, constraints associated with the problem and design
variables. The multi objective optimization problem can be
described as follow:
23
f1 ðxÞ 
6 f2 ðxÞ 7
Minimize FðxÞ ¼ 6 7; subject to GðxÞ  0 ; where xmin
4 « 5 HðxÞ ¼ 0
fm ðxÞ
 x  xmax
(5)
where x ¼ ½x1 ; x2 ; …; xN  is the variables vector and
f1 ðxÞ; f2 ðxÞ; …; fm ðxÞ are the individual values of the objective
functions.
Meta-heuristic methods are commonly used to resolve multi-
objective optimization problem. Among these methods, there
may be mentioned the Genetic algorithm GA, Monte-Carlo method
known as particles swarm optimization PSO, adaptive simulated
annealing ASA and differential evolution DE methods [50]. In this
study, the PSO method was selected as optimization algorithm.
Through the cooperation and competition among the potential
solutions, heuristic “particle swarm optimization” [51] is motivated
by the simulation of social behavior. In this technique, for the
iteration t ¼ 0, a set P0 of m solutions xi (0) is randomly selected from
the domain of the function to be minimized, each particle is
described by its position xi(t) and its velocity vi(t). In each iteration
t, the “fitness” of each position xi(t) is calculated and compared to
best individual position gi(t) and the global best position giglobal ðtÞ so
that the update rules considering D ¼ 1 are given as follow [52]:
xi ðt þ 1Þ ¼ xi ðtÞ þ Dt $vi ðt þ 1Þ (6)

vi ðt þ 1Þ ¼ u$vi ðtÞ þ c1 $r1 ðgi ðtÞ  xi ðtÞÞ þ c1 $r1 giglobal ðtÞ

 xi ðtÞ (7)
where i indicates the number of particles of swarm; t denotes the
iterations number; u is the inertia weight; c1 and c2 acceleration
parameters represent the cognitive and social factors respectively.
c1 and c2 are responsible for varying the particle velocity towards gi
and giglobal ; r1 and r2 are two random functions with uniform dis-
tribution in the range [0,1].
4.1. Objectives
4.1.1. Reliability
In stand-alone mode, the objective of the HMGS sizing is to
maximize the reliability of the HMGS in case of a blackout, avoiding
power supply interruptions and their related costs. The loss of
power supply probability (LPSP) is a statistical assessment indicator
that describe the power supply probability failure due to unmet
load demand. That indicates how reliable the HMGS to satisfy the
load demand. The LPSP can be calculated using probabilistic tech-
niques based on the energy storage accumulative effect as
expressed in the following equation [53,54]:
114 M. Azaza, F. Wallin / Energy 123 (2017) 108e118

Table 1
Component cost estimation of HMGS. ð1 þ iÞn
CRF ¼ (10)
Components List price (SEK) ð1 þ iÞn  1
Inverter (3 phase) 21.300/unit
PV system (7 kWh) 12.750 unit
Wind turbine (2 kWh) 17.000/unit
Wind regulator 8.500/unit 4.1.3. Renewable factor
Battery (40 kWh) 2.300/unit The performance of HMGS can be characterized from environ-
Diesel generator (5 kW) 8.500/unit mental impact point of view using indictor of the renewable energy
sources fraction. The objective is to minimize the diesel generator
usage, lowering the CO2 emissions and reducing the cost of oper-
Table 2 ation. The renewable factor RF is expressed as follow:
Range of decision space variables.
 P 
PD
Variable Minimum Maximum RF ð%Þ ¼ 1 P  100 (11)
PPV þ PWT
PV panels 0 100
Wind Turbine 0 10 It could be noted that a RF of 100% indicates that the system is
Autonomy days 0 6
Diesel Generator 0 4
operating basically with renewable resources which is an ideal
operational condition with the lowest cost operation and lowest
CO2 emissions.

P 
4.2. Decision space variables
PL  PPV  PWT þ PSOC;min þ PD
LPSP ¼ P (8)
PL
The HMGS configuration variables that are related to the hard-
ware component were optimized to achieve the best system
configuration. The lower and the upper bounds are listed in Table 2.
4.1.2. Cost of electricity
The cost of electricity (COE) is a profitability economic indicator 5. Results and discussion
that shows the cost of electricity production of the HMGS. The COE
includes the hybrid system hardware initial cost listed in Table 1, 5.1. Simulation results
the operation and maintenance cost. It is defined as the price per
unit produced energy (SEK/kWh) and expressed as in the following Different locations in Sweden are investigated in this study, see
equations [55,56]: Fig. 1. The locations are selected so that the HMGS is evaluated
under different solar radiation and wind speed profile. That allows
Total Net Present cost to reveal the potential of each area to the HMGS deployment and to
COE ¼ Ph¼8760  CRF (9)
h¼1 PL ðhÞ study the dependency of the HMGS investment to weather condi-
tions. The simulation conditions of the HMGS optimization process
the total net present cost includes hybrid system hardware initial over one year consider a community of 10 households and a typical
cost, the Operation and maintenance cost (O&M), and the daily household load with a peak of 2 kWh. Weather conditions
replacement cost. The capital recovery factor (CRF) is the ratio of data used for the simulation is based on 1- hour resolution data.
hybrid system hardware to the present value of the system for a The results of the optimization of HMGS configuration in 12
given length of time. Taking into account an interest rate i, and the Swedish cities are shown in Figs. 9e11.
amortization period of the system n, the CRF is expressed by: It could be seen in Figs. 9 and 11 that the energy mix in different

PV Autonomy Days Wind Turbine Diesel Generator

100
Normlized Size PV, AD, WT, DG (%)

80

60

40

20

0
d
an

o
ea

nd

le
lea
na

g
la
k

al

sta

m
or
ok

av

sa
m
um
iru

rsu

sv
Lu

al
eb
m

pp

la
U

G
nd

M
K

ste
or

ar

ot
kk

U
Su
St

G
Jo

Fig. 9. HMGS optimal sizing.

 M. Azaza, F. Wallin / Energy 123 (2017) 108e118 115

LPSP RF COE
1 4

0.8

COE [SEK/kWh]
0.6

LPSP & RF
2

0.4

0.2

0 0

d
ea
an

le

o
l
na

la

g
k

al
le

un

sta

m
or
ok

av

sa
m
um
iru

sv
Lu

al
eb
rs
m

pp

la
U

G
nd

M
K

or

ste

ot
ar
kk

U
Su
St

G
Jo

O
Location

Fig. 10. Optimal objectives values.

PV WIND BATTERY DIESEL

Kiruna Jokkmokk Luleå Storuman

14% 16% 17%
20%

40%
43% 45% 47%

33%
34% 25% 32%

13% 7% 10% 4%

Umeå Östersund Sundsvall Gävle

14% 11% 13% 16%

41% 42%
44% 45%

31% 34%
35% 28%

14% 13% 8% 11%

Uppsala Karlastad Göteborg Malmö

14% 14% 18%
21%
27%

39%
42%
45%
24% 16%
29%
22%

13% 16% 24% 39%

Fig. 11. Energy mix in 12 Swedish locations.

116 M. Azaza, F. Wallin / Energy 123 (2017) 108e118

locations is based basically on renewable resources PV and WT.
Further, the HMGS configuration depends on the location where
installed. That could be explained by the differences on weather
conditions. For instance, it could be remarked that the studied lo-
cations show 3 different groups, Fig. 9; First group with high
number of WT (e.g. Malmo € , Go
€ teborg, Karlastad). Second group
shows high number of PV panels (e.g. Luleå, Storuman, Sundsvall).
The third group shows un-particular number of PV panels or WT
(e.g. Ostersund, Umeå, Gavle, Uppsala). Fig. 10 shows the results of
optimization (LPSP, COE, RF) in the studied locations which reflects
the characteristics of each location for HMGS deployment. It could
be seen that the LPSP has a highest value 50 (%) in Jokkmokk and
achieve the lowest value of 12 (%) in Malmo €. The COE exhibit the
same profile with highest values 3.2 (SEK/kWh) in the north part of
Sweden and lowest values 2 (SEK/kWh) in the south part. Whereas
the LPSP and the COE shows clear difference depending on the
location, the RF is almost kept unchanged upon the different
locations.
Fig. 11 shows the yearly energy mix of the optimal HMGS
configuration over one year in 12 locations. Almost, all the Swedish
regions shows important PV contribution more than 40%. The WT
contributions is remarkably important in the top south and the cost
regions of Sweden. The batteries contribution is ranging from 16%
to 35% with maximum values observed in the north regions. The
diesel generator contribution is almost the same in different re-
gions ranging from 11% to 21%.
The result of simulation shows clearly that the different Swedish
regions have an important potential for HMGS with particular low
values of LPSP and COE in the south part. Moreover, even in the
north part, where it is commonly believed that the weather con-
ditions are not favorable for renewable energy investment, these
regions show a renewable resources based energy mix with more
than 50%. That could be confirmed further with a renewable factor
exceeding almost 90% and low diesel generator contribution overall
the studied locations. these results will be further confirmed
through a sensitivity analysis to reveal the robustness of the opti-
mization results.
5.2. Sensitivity analysis
The sensitivity analysis has been carried out to investigate the
effect of increasing the load linearly towards qualify the perfor-
mance as well the financial attractiveness of the HMGS with respect
to the boundaries values in Table 2. The sensitivity curves are

LPSP RF COE

Kiruna Jokkmokk Luleå

1 10 1 10 1 10

COE [SEK/kWh]
COE [SEK/kWh]

COE [SEK/kWh]
0.8 0.8 0.8
LPSP & RF

0.6 0.6 0.6

5 5 5
0.4 0.4 0.4
0.2 0.2 0.2
0 0 0 0 0 0
5 10 15 20 25 5 10 15 20 25 5 10 15 20 25

Storuman Umeå Östersund

1 10 1 10 1 10
COE [SEK/kWh]

COE [SEK/kWh]

COE [SEK/kWh]
0.8 0.8 0.8
LPSP & RF

0.6 0.6 0.6

5 5 5
0.4 0.4 0.4
0.2 0.2 0.2
0 0 0 0 0 0
5 10 15 20 25 5 10 15 20 25 5 10 15 20 25

Sundsvall Gävle Uppsala

1 10 1 10 1 4
COE [SEK/kWh]

COE [SEK/kWh]
COE [SEK/kWh]

0.8 8 0.8 8 0.8

LPSP & RF

0.6 6 0.6 6 0.6

2
0.4 4 0.4 4 0.4

0.2 2 0.2 2 0.2

0 0 0 0 0 0
5 10 15 20 25 5 10 15 20 25 5 10 15 20 25

Karlastad Göteborg Malmö

1 4 1 10 1 4
COE [SEK/kWh]

COE [SEK/kWh]

COE [SEK/kWh]

0.8 0.8 0.8

LPSP & RF

0.6 0.6 0.6

2 5 2
0.4 0.4 0.4

0.2 0.2 0.2

0 0 0 0 0 0
5 10 15 20 25 5 10 15 20 25 5 10 15 20 25
Number of houses Number of houses Number of houses

Fig. 12. Sensitivity analysis.

 M. Azaza, F. Wallin / Energy 123 (2017) 108e118 117

0.7
LPSP Yearly LPSP Summer LPSP Winter
of HMGS were studied; the LPSP, the COE, and the RF. Then, the
0.6 optimal model configuration was simulated based on one-year
0.5 hourly weather conditions data. An assessment of different areas
0.4 in Sweden for HMGS operation was performed to underline the
LPSP

0.3 potential of each location for HMGS investment. Finally, a sensi-

0.2 tivity analysis was conducted to point out the performance of
0.1
HMGS under load variation. Further a seasonal performance of the
0
HMGS was pointed out.

d
ea

le
an

o
a

la
ll

g
k

ta
le

un
n

m
or
ok

va

av

sa
m
um
iru

It was found through the results analysis that the studied lo-

s
Lu

al
eb
rs

s
m

pp

la
U

G
nd

M
K

or

ste

ar

ot
kk

U
Su
St

G
O
Jo

cations are favorable for HMGS operation with lower COE of 2 (SEK/
4

3.5
PRG Yearly PRG Summer PRG Winter kWh) in the south areas with high renewable contribution and
3
lowest LPSP. Second, the energy mix among the studied regions
2.5 remains based on renewable resources. That prove the potential of
renewable resources across the north and the south part of the
PRG

1.5 country. For instance, the long solar radiation hours during the
1
warm season makes the HMGS profitable in the long term albeit the
0.5
cold unfavorable weather conditions. In fact, a PRG more than 50%
0
during cold season in the north part still could be considerable and
d
an

nd

o
na

ea

le

a
l

rg
k

al
le

sta
al

m
ok

av

o
m
m
iru

sv

s
Lu

al
eb
rs
m

la
pp
U

G
u

nd

M
K

ste
or

valuable renewable. During the warm season the HMGS achieve

ot
ar
kk

U
Su
St

G
Jo

optimal performance with lowest LPSP and highest PRG. the LPSP is
Fig. 13. Seasonality effect.
ranging between 0.01 as minimal value and 0.22 as maximal value.
According to this results, it could be concluded the important po-
depicted in Fig. 12. It could be seen that the RF is kept unchanged, tential of HMGS as alternative for rural and islanded areas in
the LPSP is linearly correlated to the number of houses whereas the Sweden from technical and economic point of view. That is in
COE is varying inversely to the load. It could be remarked that from accordance with the rural areas development program and the
the northern regions to the southern regions the line slope of LPSP short and long term Swedish energy policy.
variations is decreasing. That means the north regions are more
sensitive to the loads variations due to the weather conditions Acknowledgment
characteristics (less solar radiation, less wind speed profile).
The authors would like to thank the Knowledge Foundation (KK-
€ m for the
stiftelsen) for the financial support and Lucas Lundstro
5.3. Seasonality effect
weather data tool used in this work.
The HMGS has been further analyzed to point out the season-
ality effect on the system performance. In fact, a seasonal variation References
analysis of LPSP value is carried out including the production gross
[1] Hanimann R, Vinterb€ ack J, Mark-Herbert C. Consumer behavior in renewable
PRG originating from renewable in units of load demand. For electricity: can branding in accordance with identity signaling increase de-
simplicity reasons, we consider two season; a cold season for 6 mand for renewable electricity and strengthen supplier brands? Energy Policy
months stating from October and warm season for 6 months as well 2015;78:11e21.
[2] Hafez O, Bhattacharya K. Optimal planning and design of a renewable energy
stating from April. The results of the analysis are depicted in Fig. 13. based supply system for microgrids. Renew Energy 2012;45:7e15.
It could be seen the clear difference between the yearly and [3] Osmani A, Zhang J. Optimal grid design and logistic planning for wind and
seasonal values of LPSP and PRG. The LPSP value during the warm biomass based renewable electricity supply chains under uncertainties. En-
ergy 2014;70:514e28.
season is much lower than its value during the cold season. That is [4] Hostick D, Belzer DB, Hadley SW, Markel T, Marnay C, Kintner-Meyer M.
basically explained by the high PV production during the warm Renewable electricity futures study. In: Volume 3. End-use electricity de-
season which is confirmed by the PRG in Fig. 13. Albeit the weak PV mand; 2012.
[5] Lund PD, Lindgren J, Mikkola J, Salpakari J. Review of energy system flexibility
production during winter which affect the yearly LPSP, still, it could measures to enable high levels of variable renewable electricity. Renew Sus-
be noted that, in many locations the PRG is relatively important tain Energy Rev 2015;45:785e807.
(more than 50%). Some locations such as Malmo € and Go€ teborg are [6] Dusonchet L, Telaretti E. Comparative economic analysis of support policies
for solar PV in the most representative EU countries. Renew Sustain Energy
still favorable regions for HMGS operation even during winter cold
Rev 2015;42:986e98.
season with low LPSP, 0.12 and 0.05 respectively, which is due their [7] Bauer N, Mouratiadou I, Luderer G, Baumstark L, Brecha RJ, Edenhofer O, et al.
favorable wind speed profile for WT generation. Global fossil energy markets and climate change mitigation e an analysis with
REMIND. Clim Change 2013;136:69e82.
Regarding the reliability of the HMGS during the summer sea-
[8] Zhang S, Bauer N, Luderer G, Kriegler E. Role of technologies in energy-related
son, the results shows clearly that the studied locations are favor- CO2 mitigation in China within a climate-protection world: a scenarios
able for HMGS deployment technically with low LPSP values and analysis using REMIND. Appl Energy 2014;115:445e55.
high PRG. In fact, the observed LPSP during warm season is ranging [9] Ismail M, Moghavvemi M, Mahlia T, Muttaqi K, Moghavvemi S. Effective uti-
lization of excess energy in standalone hybrid renewable energy systems for
between 0.01 observed in Goteborg and 0.22 observed in Storuman. improving comfort ability and reducing cost of energy: a review and analysis.
Renew Sustain Energy Rev 2015;42:726e34.
[10] Davari M, Mohamed YA-RI. Robust multi-objective control of VSC-based DC-
6. Conclusion voltage power port in hybrid AC/DC multi-terminal micro-grids. IEEE Trans
Smart Grid 2013;4:1597e612.
[11] Parhizi S, Lotfi H, Khodaei A, Bahramirad S. State of the art in research on
This work has given some insight and techno economic analysis
microgrids: a review. IEEE Access 2015;3:890e925.
of HMGS deployment in different Swedish regions. The HMGS [12] Wang L, Singh C. Multicriteria design of hybrid power generation systems
model consist of PV/WT/DG a battery bank and an energy man- based on a modified particle swarm optimization algorithm. IEEE Trans En-
agement system. The optimal system size and configuration is ergy Convers 2009;24:163e72.
[13] Bilal BO, Sambou V, Ndiaye P, Ke  be
 C, Ndongo M. Optimal design of a hybrid
performed through PSO multi-objective optimization algorithm. solarewind-battery system using the minimization of the annualized cost
Three objectives that affect the technical and economic investment system and the minimization of the loss of power supply probability (LPSP).
118 M. Azaza, F. Wallin / Energy 123 (2017) 108e118
Renew Energy 2010;35:2388e90.
[14] Maheri A. Multi-objective design optimisation of standalone hybrid wind-PV-
diesel systems under uncertainties. Renew Energy 2014;66:650e61.
[15] Sharafi M, Elmekkawy TY. Multi-objective optimal design of hybrid renewable
energy systems using PSO-simulation based approach. Renew Energy
2014;68:67e79.
[16] Hassan MA, Abido MA. Optimal design of microgrids in autonomous and grid-
connected modes using particle swarm optimization. IEEE Trans Power
Electron 2011;26:755e69.
[17] Kaviani AK, Riahy G, Kouhsari S. Optimal design of a reliable hydrogen-based
stand-alone wind/PV generating system, considering component outages.
Renew Energy 2009;34:2380e90.
[18] Wu Y-K, Lee C-Y, Liu L-C, Tsai S-H. Study of reconfiguration for the distribu-
tion system with distributed generators. IEEE Trans Power Deliv 2010;25:
1678e85.
[19] Abbes D, Martinez A, Champenois G. Eco-design optimisation of an autono-
mous hybrid windephotovoltaic system with battery storage. IET Renew
Power Gener 2012;6:358e71.
[20] Kirthiga MV, Daniel SA, Gurunathan S. A methodology for transforming an
existing distribution network into a sustainable autonomous micro-grid. IEEE
Trans Sustain Energy 2013;4:31e41.
[21] Katsigiannis YA, Georgilakis PS, Karapidakis ES. Hybrid simulated annea-
lingetabu search method for optimal sizing of autonomous power systems
with renewables. IEEE Trans Sustain Energy 2012;3:330e8.
[22] Vrettos EI, Papathanassiou SA. Operating policy and optimal sizing of a high
penetration RES-BESS system for small isolated grids. IEEE Trans Energy
Convers 2011;26:744e56.
[23] Fung C, Ho S, Nayar C. Optimisation of a hybrid energy system using simulated
annealing technique. Proceedings of TENCON '93 IEEE region 10 international
conference on computers, communications and automation.
[24] Prodromidis GN, Coutelieris FA. Simulations of economical and technical
feasibility of battery and flywheel hybrid energy storage systems in autono-
mous projects. Renew Energy 2012;39:149e53.
[25] Mills A. Simulation of hydrogen-based hybrid systems using Hybrid2. Int J
Hydrogen Energy 2004;29:991e9.
[26] Erdinc O, Uzunoglu M. Optimum design of hybrid renewable energy systems:
overview of different approaches. Renew Sustain Energy Rev 2012;16:
1412e25.
[27] Sunanda S, Chandel SS. Review of software tools for hybrid renewable energy
systems. Renew Sustain Energy Rev 2014;32:192e205.
[28] Zhou W, Lou C, Li Z, Lu L, Yang H. Current status of research on optimum
sizing of stand-alone hybrid solarewind power generation systems. Appl
Energy 2010;87:380e9.
[29] Dufo-Lo pez R, Bernal-Agustín JL, Yusta-Loyo JM, Domínguez-Navarro JA,
Ramírez-Rosado IJ, Juan Lujano J, et al. Multi-objective optimization mini-
mizing cost and life cycle emissions of stand-alone PVewindediesel systems
with batteries storage. Appl Energy 2011;88:4033e41.
[30] Belfkira R, Zhang L, Barakat G. Optimal sizing study of hybrid wind/PV/diesel
power generation unit. Sol Energy 2011;85:100e10.
[31] Ould Bilal B, Sambou V, Ndiaye PA, Ke  be
 CMF, Ndongo M. Optimal design of a
hybrid solarewindebattery system using the minimization of the annualized
cost system and the minimization of the loss of power supply probability
(LPSP). Renew Energy 2010;35:2388e90.
[32] Koutroulis E, Kolokotsa D, Potirakis A, Kalaitzakis K. Methodology for optimal
sizing of stand-alone photovoltaic/wind-generator systems using genetic al-
gorithms. Sol Energy 2006;80:1072e88.
[33] Moghaddas-Tafreshi SM, Hakimi SM. Optimal sizing of a stand-alone hybrid
power system via particle swarm optimization for Kahnouj area in south-east
of Iran. Renew Energy 2009;2009(34):1855e62.
[34] Bernal-Agustín JL, Dufo-Lo  pez R. Efficient design of hybrid renewable energy
systems using evolutionary algorithms. Energy Convers Manag 2009;50:
479e89.
[35] Lagorse J, Paire D, Miraoui A. Sizing optimization of a stand-alone street-
lighting system powered by a hybrid system using fuel cell, PV and battery.
Renew Energy 2009;34:683e91.
[36] Blindheim B. A missing link? The case of Norway and Sweden: does increased
renewable energy production impact domestic greenhouse gas emissions?
Energy Policy 2015;77:207e15.
[37] Energy policies of IEA countries: Sweden: 1996 review. Paris: OECD; 1996.
[38] An Overview of Spatial Policy in the Sweden. http://www.mlit.go.jp/
kokudokeikaku/international/spw/general/sweden/index_e.html.
[39] Sweden - rural development. http://www.un.org/esa/agenda21/natlinfo/
countr/sweden/ruraldevelopment.pdf.
[40] Daud A-K, Ismail MS. Design of isolated hybrid systems minimizing costs and
pollutant emissions. Renew Energy 2012;44:215e24.
[41] Hof AR, Jansson R, Nilsson C. Future of biodiversity in the Barents region. Kbh.:
Nordisk Ministerrad; 2015.
[42] Wang L, Singh C. PSO-Based multi-criteria optimum Design of A Grid-
connected hybrid power system with multiple renewable sources of energy.
In: 2007 IEEE swarm intelligence symposium; 2007.
[43] Kefayat M, Ara AL, Niaki SN. A hybrid of ant colony optimization and artificial
bee colony algorithm for probabilistic optimal placement and sizing of
distributed energy resources. Energy Convers Manag 2015;92:149e61.
[44] Arul P, Ramachandaramurthy VK, Rajkumar R. Control strategies for a hybrid
renewable energy system: a review. Renew Sustain Energy Rev 2015;42:
597e608.
[45] Ashari M, Nayar C. An optimum dispatch strategy using set points for a
photovoltaic (PV)edieselebattery hybrid power system. Sol Energy 1999;66:
1e9.
[46] Tazvinga H, Zhu B, Xia X. Energy dispatch strategy for a photo-
voltaicewindedieselebattery hybrid power system. Sol Energy 2014;108:
412e20.
[47] Tazvinga H, Zhu B, Xia X. Energy dispatch strategy for a photo-
voltaicewindedieselebattery hybrid power system. Sol Energy 2014;108:
412e20.
[48] Azoumah Y, Yamegueu D, Ginies P, Coulibaly Y, Girard P. Sustainable elec-
tricity generation for rural and peri-urban populations of sub-Saharan Africa:
the “flexy-energy” concept. Energy Policy 2011;39:131e41.
[49] Jean PZ. End-use metering campaign in 400 households in Sweden. Assess-
ment of the potential electricity savings. Report. 2009.
[50] Tian Z, Fong S. Survey of meta-heuristic algorithms for deep learning training.
In: Optimization algorithms - methods and applications; 2016.
[51] Kennedy J, Eberhart̊ R. Particle swarm optimization. Proceedings of ICNN'95-
international conference on neural networks.
[52] Eberhart R, Kennedy J. A new optimizer using particle swarm theory. MHS'95
proceedings of the sixth international symposium on micro machine and
human science.
[53] Deshmukh M, Deshmukh S. Modeling of hybrid renewable energy systems.
Renew Sustain Energy Rev 2008;12:235e49.
[54] Yang H, Zhou W, Lu L, Fang Z. Optimal sizing method for stand-alone hybrid
solarewind system with LPSP technology by using genetic algorithm. Sol
Energy 2008;82:354e67.
[55] Kolhe M. Techno-economic optimum sizing of a stand-alone solar photovol-
taic system. IEEE Trans Energy Convers 2009;24:511e9.
[56] Sharma VK, Colangelo A, Spagna G. Photovoltaic technology: basic concepts,
sizing of a stand alone photovoltaic system for domestic applications and
preliminary economic analysis. Energy Convers Manag 1995;36:161e74.