Renewable and Sustainable Energy Reviews 66 (2016) 617–630

Contents lists available at ScienceDirect

Renewable and Sustainable Energy Reviews

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

Application of Artificial Intelligence Methods for Hybrid Energy System

Optimization
S.M. Zahraee a, M. Khalaji Assadi a,n, R. Saidur b
a
Department of Mechanical Engineering, Universiti Teknologi PETRONAS (UTP), Bandar Seri, Iskandar, 31750 Tronoh, Perak, Malaysia
b
Centre of Research Excellence in Renewable Energy (CoRE-RE), King Fahd University of Petroleum and Minerals (KFUPM), Dhahran 31261, Kingdom of Saudi
Arabia

art ic l e i nf o a b s t r a c t

Article history: Consciousness of the need to decrease our unnatural weather changes and of the critical increase in the
Received 11 September 2015 costs of traditional sources of energy have motivated many nations to provide innovative energy stra-
Received in revised form tegies that promulgate renewable energy systems. For example, solar, wind and hydro related energies
9 March 2016
are renewable energy sources, and they are environmentally friendly with the potential for broad use. All
Accepted 13 August 2016
of the load requirement conditions in comparison with single usage can provide more economical and
dependable electricity, as well as environmentally friendly sources, by compounding such renewable
Keywords: energy sources using backup units to shape a hybrid scheme. Sizing the hybrid system elements opti-
Artificial Intelligence mally is one of the most important matters in this type of hybrid system, which could sufficiently meet
Hybrid Energy System
all of the load demands with a minor financial investment. Although a number of studies have been
Optimization
performed on the optimization and sizing of hybrid renewable energy systems, this study presents a full
analysis of Artificial Intelligence optimum plans in the literature, making the contribution of penetrating
extensively the renewable energy aspects for improving the functioning of the systems economically.
& 2016 Elsevier Ltd. All rights reserved.

Contents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 617
2. World energy scenario . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 618
3. Literature review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 619
3.1. Commercially existing software applications for the sizing of Hybrid Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 619
3.2. Artificial Intelligence Methods in support of Hybrid Energy Systems Optimization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 620
3.2.1. Genetic Algorithms (GA) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 620
3.2.2. Particle Swarm Optimization (PSO). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 620
3.2.3. Simulated Annealing (SA) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 621
3.2.4. Hybrid models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 621
3.3. Promising method in Hybrid System sizing for future use . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 621
3.3.1. Algorithm of ant colony. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 621
3.3.2. Algorithm of the Artificial Immune System (AIS). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 622
3.4. Other promising approaches . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 624
3.4.1. Wind systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 624
3.4.2. Solar systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 624
3.4.3. Other Hybrid systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 625
4. Results and discussion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 626
5. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 627
Acknowledgment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 627
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 627

n
Corresponding author.
E-mail addresses: [email protected],
[email protected] (M. Khalaji Assadi).

http://dx.doi.org/10.1016/j.rser.2016.08.028
1364-0321/& 2016 Elsevier Ltd. All rights reserved.
618 S.M. Zahraee et al. / Renewable and Sustainable Energy Reviews 66 (2016) 617–630

1. Introduction the system's effectiveness and reliability of the energy supply

[7,14,25]. As a result, renewable energy access can be improved in
One of the prominent challenges that the world faces today is upcoming sustainable areas [13]. Renewable energy systems pre-
providing its needed energy while saving for the future simulta- sent some helpful effects in several types of applications modes,
neously [1]. In recent times, a considerable amount of energy has which have been identified to be, for example, the costs of the
been required around the world. The world depends solely on systems when preventing wide-ranging consumption, assessment
conventional energy sources, for example, coal, natural gas and efforts that are focused on fast cost falls, and the ability of these
crude oil [2,3]. In addition, the demand for energy use is growing systems in development [4]. Method designs should be most fa-
every day, which, however, results in a brisk demand for the usual vorable in operations and component selection, to achieve elec-
fossil fuels [4,5]. Whereas these sources of energy are limited and trical energy from renewable energy resources, which is related to
unload quickly, that in turn pressures the stability of potential hybrid system reliability and cost efficiency [26–28]. Therefore, the
generations of energy demand [6–8]. In addition, the un- most advantageous sizing method is to proficiently and in-
predictable supply of the aforementioned sources and the negative expensively use renewable energy resources [9]. Principally, the
influence on the administrative balances between energy (petro- most favorable size systems are required for comprehensive ana-
leum) exporting/importing nations warrants vital investigations lysis to give the location and control site-dependent factors, such
on the prospects of popular means for producing energy [9]. In as solar rays, wind rates, and temperatures, and their costs [14,29].
recent years, global warming and climate change are two main Computer-based simulation and optimization has become a pre-
important issues in the global economy and environment, and eminent technique for designing power systems; this approach
they have a considerable effect on the insufficient accessibility and involves comprehensive analysis [30]. There are some constraints
rising cost of energy [10]. The high consumption of energy in the in the formulation and solution of the design and optimization
world has already increased concerns about supply difficulties and approach, such as the resource availability, technology, efficiency,
significant environmental influences, such as global warming and mathematical models and other aspects. However, the advance-
climate change [11–13]. The evidence from [7,14–17] concludes ment in computational techniques has made it easy to address
that energy created by traditional energy sources causes an in- optimization problems by using a number of optimization and
crease in greenhouse gas discharge, which could affect global simulation techniques. A number of simulation tools, such as
warming. Efforts to reduce the volume of greenhouse gas emission HOMER (Hybrid Optimization Model of Electric Renewable), HY-
have led to the Kyoto agreement on the global decline of green- BRID2 (The Hybrid Power System Simulation Model), and HOGA
house gas production. This agreement was put in place to lessen (Hybrid Optimization using Genetic Algorithm), are used for the
the issue and the dependence on traditional energy systems. The design and optimization of hybrid systems as well as for improv-
notion of greenhouse gas reduction is valid for both developing ing their performance. One of the main branches of computer
and developed nations [18,19]. Addressing the issues discussed science is Artificial Intelligence (AI), which investigates and builds
above, continual potential action for sustainable improvement is intelligent software and machines. Russell and Norvig [31] explain
required. In addition, cost-effective, consistent, and en- that AI is “the investigate and framework of intelligent factors”, in
vironmentally friendly energy systems are the attributes of a which an intelligent factor performs actions that maximize the
sustainable energy system that efficiently uses local assets and possibility of success. AI is composed of branches such as genetic
networks [20]. Therefore, renewable energy sources are positioned algorithms (GA), particle swarm optimization (PSO), simulated
as one of the proficient and useful solutions [7,15]. There are dif- annealing (SA), artificial neural networks (ANN) and hybrid mod-
ferent types of renewable energy structures, such as wind, solar, els, including two or more previous branches. The effective and
hydro-electric, water, ocean, biomass and geothermal energy, correct application of intelligent methods cause the development
which have unique benefits and are suitable for applications. of comprehensive and useful systems with better performance or
Currently, some countries have the potential for different types of different characteristics, which cannot be obtained compared to
energy resources, such as solar, wind, water, and geothermal, and using traditional approaches [32]. This paper aims at reviewing the
in addition, many companies are developing, constructing and literature that is related to the various optimization techniques of
setting up modern and high-tech renewable energy systems. Artificial Intelligence Methods for Hybrid Energy Systems Opti-
These countries attempt to lead a large network of investigators mization, such as genetic algorithms (GAs), simulated annealing
and other partners to utilize cutting-edge and advanced technol- (SA), and particle swarm optimization (PSO). In addition, the
ogies that will provide a cost for renewable electricity generation characteristics of all of the methods are compared together, to help
that is competitive with traditional sources of energy. Increasing researchers to use them effectively and in a cost-effective manner.
the proportions of renewable energy systems such as wind and
solar have joined the grid and will impact the fossil fuel generators
on the grid, which will lead to a decrease in emissions and costs 2. World energy scenario
for the consumers. These advantages include, for example, a re-
duction in the external energy confidence and a decrease in According to the Energy Information Administration, the global
communication and conversion costs. Additionally, renewable energy consumption is increasing by approximately 2.3% per an-
energy sources supply an important improvement over the usual num [33]. The wind generates approximately 20% of the electricity
energy systems, and almost none discharge gaseous or water in Denmark. However, on a global scale, the electricity that is
pollutants during their operations [20,21]. As mentioned before in produced from the wind is less than 1%. The statistics of world
this study, renewable energy systems are measured as capable energy consumption from the year 2008–2035 is expected to grow
power producing sources. On the other hand, a disadvantage of the by approximately 53% [33]. Fig. 1 exhibits the high development of
specified energy selection is their irregular character and our the universal energy requirement.
confidence in the weather conditions. Therefore, renewable power Fig. 2 shows the United State's electricity generation from fuel
production cannot completely control the power requirement of during the years 1990–2040. Additionally, it shows that the re-
the load at any specific instant in time [14,22–24]. This type of newable share would be increased by a few percent; however,
difficulty is connected to the changeable character of these re- some sources, such as fossil fuels comprising coal and natural gas,
sources, which can be solved by assimilating resources in an ap- are still increasing.
propriate hybrid blend. This approach causes an improvement in The world's economic development is surprisingly dependent
 S.M. Zahraee et al. / Renewable and Sustainable Energy Reviews 66 (2016) 617–630 619

900 and 2012, there is increased global demand for renewable sources
800 of energy supplied at 19% of the global energy usage in the year
2011.
700

600

500 3. Literature review

Non--OECD
400 OECD The design of hybrid renewable energy systems is a significant
300 area, and many researchers are interested in this topic. Therefore,
200 there is a large amount of literature on this topic, which we can
use. The design problem mentioned above is related to the energy
100
systems that are observed to have the optimal pattern and optimal
0 location, type and sizing of generation components established on
0
1990 2000 20
008 2015 2020 2025 2030 20335
individual nodes. Therefore, this type of system can load the re-
Fig. 1. World energy consumption, 1990–2035 (quadrillion Btu). quirements with a minimum of costs [39]. The proposal of hybrid
renewable energy can estimate the cost and production over a
lifetime of this technology. The first estimate, for the lifetime cost,
usually includes two components: the operational cost, such as the
principal cost and the preservation cost, which both point to a
“fixed cost”. Additionally, in a computation of the life span cost, the
financial values are modified according to the timing and should
be considered. Therefore, the optimized hybrid system patterns
combine producer types and sizes in the minimum life span cost
and production. Thus, the design by the lowest “Net Present Value”
(or NPV) is defined as the “optimal configuration” or “optimal de-
sign”, with all probable hybrid system designs being in optimal
transition [40,41].
There are many methods for providing an “optimal design”
indicator and many software tools that are accessible commer-
cially, to act as real-time system integration. Additionally, there are
different optimal techniques that are applied from many re-
searchers to use for hybrid renewable energy system sizing. Dif-
Fig. 2. US electricity generation by fuel, 1990–2040 (trillion kilowatt-h per year) ferent Optimization methods, such as graphical construction
[34].
[42,43], probabilistic techniques [44], iterative approach, dynamic
programming, artificial intelligence (AI), linear programming
on increasing energy requirements. In addition, fossil fuels are not [45,46] and multi-objective were implemented by investigators to
consistently spread around the globe, and if the world economy optimize hybrid PV/wind energy systems. Table 1 presents a
depends heavily on them, local or global disagreement could occur summary of the optimization methods that were developed by
and create an energy crisis. The use of today's predictable com- different researchers.
bustibles, the worldwide surroundings and the situations of some
countries have been unfavorably impacted. We must emphasize 3.1. Commercially existing software applications for the sizing of
that it is important to discover new actions to connect suitable Hybrid Systems
quantities of energy [35–38]. Fig. 3 demonstrates the energy pro-
duction in the US according to different sources. In the years 2011 Various existing software applications of Hybrid Energy

Fig. 3. Energy sources in the US [35].

620 S.M. Zahraee et al. / Renewable and Sustainable Energy Reviews 66 (2016) 617–630

Table 1 Systems Optimization

Different optimization methods.
3.2.1. Genetic Algorithms (GA)
Optimization methods Optimized factors Notice
One of the optimization methods operates in terms of the ge-
Artificial Intelligence Hybrid solar-wind Using Evolution approach netic process for biological mechanisms and is called GAs, which
a. Genetic Algorithm system with battery have the ability to present a problem-solving method for difficult
b. Particle Swarm bank real-world problems [47,48]. Holland first represented the concept
c. Simulated annealing
d. Artificial Neural
of GAs [49], and afterward, it was widely utilized in many appli-
Network cations, case studies, and information mining. There many pub-
e. Hybrid model lished articles that regard GA usage in hybrid systems, for research
Iterative Method Hybrid solar-wind Using LPSP to determine dif- on sizing. For example, Yang et al. [50,51] and Bilal et al. [52] make
a. Hill climbing system ferent conditions of solar-
use of GAs for a PV hybrid wind system, to perform sizing. Kou-
b. Dynamic wind combination
Programming troulis et al. [53,54] used genetic algorithms to find the energy
c. Linear Programming expenses of the whole system and confirmed that the application
d. Multiobjective of hybrid PV/wind systems can lead to lower expenses of the
Graphical Construction Battery and PV array Usually applied for two systems in comparison with systems in which the sources of ex-
factors
Probabilistic methods Efficiency of hybrid By using statistical method of
clusively PV and WG are applied. On the other hand, Lagorse et al.
system data gathering [55] used GAs in designing a multi-source hybrid source that
Deterministic methods Standalone PV with Based on equations applied combined wind and oil. Examples of PVs are wind, oil, micro-
battery bank for finding particular values turbines, and a battery, which had improved sizes according to
when applying fixed factors
Kalantar et al. [56], from their use of Genetic algorithms. Ad-
Software based Using an input file with all of
a. HOMER the essential data based on a ditionally, Lopez et al. [57,58] programmed Hybrid Optimizations
software application based on Genetic Algorithms (HOGA), as a simulation application
to plan compounds of stand-alone hybrid energy systems with
content from renewable sources and typical diesel generators.
Lagorse et al. used a hybrid GA and simplex-based method [59],
Table 2 and Zhao et al. [60] planned a GA that includes a wind source and
Summary of software tools for designing an HES.
fundamental technical specifications as the main elements. These
Number Software Input Output were used as input parameters. Additionally, a wind source for the
energy has been established to reduce the production cost and
1 HOMER ● Load command ● improve unit sizing improve the system's reliability. Additionally, Li et al. [61] im-
● Source input ● Energy and net cost. proved the rate of gearbox proportioning by using a GA; in so
● element details, such as ca- ● Small part of renew-
pital, maintenance and re- able energy
doing, the power ratings of the multi-hybrid stable wind gen-
placement expenses erators improved. On the other hand, the wind outline for max-
● System control imizing energy production includes the positions for the wind
2 HYBRID2 ● Load demand ● Unit sizing with cost turbines on the land [62]. Grady et al. [63] offered a GA to conclude
● Resources input optimization
the optimal generation of wind turbines with the highest pro-
● primary investment and ● expenses of energy
O&M components cost ● Release proportions of duction capacity while restricting the number of turbines that
Components details the different green- were fixed and the area of the land that was filled by each wind
house gases farm. Emami and Noghreh [64] used a new coding approach and a
● System payback novel objective function with GAs to solve this problem. Their
periods
3 HYBRIDS ● Solar size ● Cost
method performed better than previous methods that were pro-
● Turbine type of wind ● Percentage of green- posed for the management of the cost, power and effectiveness of
● number and type of battery house gases the wind farm. Li et al. [65] used a multi-level GA and could solve
4 RET Screen ● Information loading ● Energy release and optimal design problems as well as active control algorithms for
● Solar size store
the wind. Additionally, Kalogirou [66] solved financial goals in
● Necessary hydrology and ● Costs
invention database ● Production decreases increasing the system of solar energy by implementing ANNs and
● Climate database ● Economic capability GAs. The ANN method is suitable for making the connection of the
● Risk analysis collector region and high container size using the minor power
5 TRNSYS ● Meteorological data input ● Dynamic simulation that is desirable for the system. Then, a GA is applied to calculate
● Ingrained models of electrical energy
sources
approximately the best size of the factors to lengthen the life-cycle
6 IHOGA ● Data loading ● Multi aim improving reserves. Varun [67] used the GA method to increase the thermal
● Resources key input data ● charge of energy execution of flat laminate solar air warmers by reflecting several
● Elements and economics ● Life cycle release systems and operating factors. Last, the GA approach was used by
facts ● study for addressing
Zagrouba et al. [68] to determine the electrical factors. This study
energy
involves photovoltaic solar cells and modules for the determina-
tion of the analogous highest power point.

Systems (HES) are discussed in this paper. Among them, HOMER 3.2.2. Particle Swarm Optimization (PSO)
plays a significant role in applications of HES, which involves One of the improvement techniques, which involves shifting
quickly searching for the optimal sizing of the energy systems. and swarm intelligence based on evolutionary calculation techni-
Additionally, it is useful to analyze the sensitivity of investigating ques, is the PSO. This method was implemented by James Kennedy
the influence of uncertainty or changing factors. A list of this and Russell Eberhart in 1995. A component of the swarm is the
software for the design of an HES is presented in Table 2. system model or social constitution of a basic person to make a
group that has some intention, such as food searching [69,70]. In
3.2. Artificial Intelligence Methods in support of Hybrid Energy comparison with GA-based methods, there is some similarity
 S.M. Zahraee et al. / Renewable and Sustainable Energy Reviews 66 (2016) 617–630 621

between them. The entry data of the PSO approach includes me- accepting or rejecting new problem-solving methods. This method
teorological circumstances, the unit cost of the hybrid components was established by Kirkpatrick, Gelatt, and Vecchi in 1983 [87],
such as installation and continuation costs, limitations and ap- and it has been implemented and expanded since the 1980s′
propriateness of purpose and the principles of specific PSO factors. simple architecture, with effective contributions [88]. The early
One of the population-based optimization procedures is the PSO- heating and cooling processes are important to the SA. The basic
based sizing methodology process. PSO has a large amount of use algorithm's criteria are the setting value, deviation, cooling plan,
in hybrid sizing research, such as with GAs, and finding literature and acceptance of search execution. Simulated annealing in ARE-
examples, such as Gas, is not difficult. Researchers such as Sanchez NA 12.0 software was used by Ekren et al. [25] to find the optimum
et al. [71], Dehghan et al. [72] and Kaviani et al. [73] planned a unit point of performance of a hybrid system by considering the loss of
sizing PSO system to PV (wind and oil) unit microgrids, and a wind the load probability and autonomy analysis on an hourly basis.
and fuel (oil) cell configuration was recognized by Tafreshi and According to Sutthibun and Bhasaputra [89], the model that was
Hakimi [74,75]. On the other hand, a wind–PV hybridization and used to identify the optimal location and size of the DRG to
PSO were planned with Wang and Singh [76] and Zhao et al. [77]. minimize the real power loss (PL), production (Epg), and the
The same study shows that Wang and Singh [78] employed a PSO possible severity index (SI) that it faced with regard to apower
in network equivalent hybrid renewable energy systems. For balance and power generation limitation, used SA as an optimi-
handling hybrid systems, a multi-criteria approach was proposed zation tool. Based on Ghadimi and Ghadimi [90] in 2012, SA
to solve hybrid systems such as wind turbine generators of wind minimized the power losses for sizing the DRG and storage banks
turbines, photovoltaic panels, and battery problems. These pro- in a distributed network. According to Fungetal [91], to obtain the
blems have multiple design objectives, such as costs, consistency generator settings and battery charge or discharge schedules in
and releases [79]. There were more than 20 varieties in the pro- everyday loads, SA was used in a diesel generator and a sine wave
posals that used PSO algorithms. Zhao et al. [80] used Hybrid inverter with a controller element. Katsigiannis et al. compared
Particle Swarm Optimization and Wavelet Mutation (HPSOWM) SAs with the Tabu algorithm for sizing in an HRES that was used to
along with GAs. HPSOWM performs efficiently and fast. The Binary minimize the COE. As a result, SA was faster to converge, although
Particle Swarm optimization (BPSO) algorithm was also introduced it was less efficient than the other method [92].
by Kennedy and Eberhart. This approach enables the use of PSO for
binary problems [80]. Pourmousavi et al. [81] claimed that particle 3.2.4. Hybrid models
swarm optimization (PSO) has a very rapid convergence time in Hybrid approaches are a useful collaboration of two or more
comparison with sequential quadratic programming optimization different methods that use the beneficial effects of the methods in
in terms of the energy management. Additionally, forecasting the achieving an optimum result for a particular design problem. Be-
wind speed is indispensable in wind-related engineering studies cause most of the difficulties that we tackle are multi-objective,
and is important in the management of wind farms. As a technique conducting a hybrid method is an excellent objective in nature,
that is essential for the future of clean energy systems, reducing and utilizing a hybrid approach is a suitable alternative method to
the forecasting errors that are related to the wind speed has al- address problems that require considerable comprehension of all
ways been an important research subject. In this paper, an opti- of the methods. Meza et al. [93] developed a multi-objective
mized hybrid method based on the Autoregressive Integrated model to generate expansion planning (MGEP) and an analytical
Moving Average (ARIMA) and Kalman filter is proposed to forecast hierarchy process (AHP) model to address a multi-objective pro-
the daily mean wind speed in western China. This approach em- blem that included costs, environmental effects, fuel price risks
ploys Particle Swarm Optimization (PSO) to find the optimum and imported fuel. One solution was achieved by Nasiraghdam and
factors of the ARIMA model, which develops a hybrid model that is Jadid [94], by using a multi-objective artificial bee colony (ABC)
best adapted to the data set, increasing the fitting accuracy and algorithm that had considerable quality and good diversity of the
avoiding over-fitting. Afterward, the suggested method is tested on pareto front in comparison with multi-objective PSO (MOPSO)
the wind farms of western China, where the proposed hybrid methods and non-dominated shorting GA-II (NSGA-II). Alsayed
model is shown to perform effectively and steadily [82]. A model et al. [95] used different multi-criteria decision analysis (MCDA)
for a wind-CAES system was developed by Kahrobaee and Asgar- optimization techniques to determine the optimum sizing of PV–
poor [83] to optimize the daily activities and long-term plans of WT. Sensitivity analysis of MCDA algorithms was performed by
systems that use the PSO algorithm to maximize their profits. In accounting for different weighting criteria approaches with dif-
addition, particle swarm optimization (PSO) was applied to find ferent types of vacillation scenarios of solar radiation and wind
the maximum placement and sizing of PV distributed generation speed. Although it was complex, beneficial results were provided
in radial distribution systems to decrease the power loss. The final that were helpful in the design of hybrid energy systems. An Ar-
result shows that PSOs can achieve a maximum power loss de- tificial Neural Network (ANN) strategy was used by considering
crease [84]. In another investigation, a novel method is suggested flow batteries to control the uncertainties in the wind outcomes
to find the optimum design of hybrid renewable energy systems based on the further lower energy costs. Table 3 shows a brief
that are composed of different generators and storage devices. The summary of additional hybrid methods in more detail.
ε-constraint approach has been used for minimizing simulta-
neously the total cost of the system, the fuel emissions and the 3.3. Promising method in Hybrid System sizing for future use
unmet load. A particle swarm optimization (PSO)-simulation
based method has been applied to address the multi-objective 3.3.1. Algorithm of ant colony
optimization problem. Finally, sensitivity analysis was performed Finding the direct path of the ants is a developmental routine
for investigating the sensibility of different factors to the con- that has recently been recognized. The basic Ant Colony algo-
structed model [85]. rithms were established on the recital of social creatures that have
the ability to look for the fastest paths to the food sources while
3.2.3. Simulated Annealing (SA) applying famous mater as a pheromone [100]. The pheromones
Studies about wind PV batteries and renewable hydrogen hy- are the chemical objects that are reserved by the ants to form a
brids used the Simulated Annealing (SA) method on optimization communication media among them. Ant Colony System Algorithm
problems through a simulated annealing process [86]. By the (ACSA) is the addition of Ant Colony Optimization (ACO). In most
ability to avoid local minima, it can integrate a possible role for engineering applications, it has better action than the ACO [101–
622 S.M. Zahraee et al. / Renewable and Sustainable Energy Reviews 66 (2016) 617–630

Table 3
Summary of Hybrid models.

Reference Systems studied Topics covered Highlights

[93] Oil/steam, coal/seam, , hydro, Total costs, CO2 emission, Fuel consumption, Energy ● The research indicates a multi-objective generation expansion
wind, nuclear price risk and minimization of outage cost planning (MGEP) model of a power electric system that involves
(reliability) renewable energy sources (RES)
● The mixed-integer linear programming (MILP) is implemented
for the suggested optimization. and an effective linearization
approach is suggested to change the non-linear reliability metrics
into a set of linear expressions.
● Fuzzy decision maker is used to choose the most-preferred so-
lution among the Pareto results.
[94] PV, Wind and Fuel cell Power loss minimization, voltage stability index, COE ● The results that were achieved by the multi-objective artificial
and emissions Bee colony algorithm have a suitable quality and better diversity
of the pareto front compared to the NSGA-II and MOPSO
methods.
[95] PV and wind generator Emissions decrease, estimated expenses and social ● This paper includes optimum sizing of PV–WT by using different
acceptance multi-criteria decision analysis (MCDA) optimization techniques.
● Sensitivity analysis of the MCDA algorithms has been performed,
in terms of weighting criteria approaches with different fluc-
tuation in the scenarios of wind speed and solar radiation
profiles.
● The proposed approach gives the decision maker the ability to
comprise any type of criteria, enabling a confirmation of the in-
fluence of these criteria on the optimal solutions, by considering
different input data sensitivity scenarios.
[96] Costs, environmental effects, Conventional steam units, coal units, combined cycle ● In this study, a multi-objective model for a generation expansion
imported fuel and fuel price risks modules (CC), nuclear, gas turbines (TG), wind farms, planning (MGEP) model is proposed. A model is developed to
geothermal and hydro units suggest the nondominated solutions and using the Analytical
Hierarchy Process (AHP) to choose the “best solution” among the
representative (clustered) solutions.
● A large problem is time: access to the main decision makers is
restricted, and it could be expensive. Another problem is that
there is comparatively limited experience with such approaches
in a group setting, in which group members have a variety of
priorities.
[97] PV, Wind, Diesel, Biodiesel and COE and total greenhouse gas emissions (GHG) ● The large sizes of diesel-fueled generators cause COEs that are
Battery bank too small and CO2-eq. emissions that are large because enormous
sizes of biodiesel-fueled generators cause a reversal in the out-
comes. Moreover, the use of FC with natural gas as a fuel is not
suggested because of the enormous expense and the high CO2-
eq. emissions that are released.
[98] Wind generator and battery Lowest cost of battery connection with large wind ● The paper shows sizing and control approaches for a zinc–bro-
farm mine flow battery-based energy storage system.
● The results present that the power flow control strategy does
have an important effect on the proper sizing of the rated power
and energy of the system. Specifically, ANN control strategies
lead to less expended energy in the storage systems than sim-
plified controllers.
[99] PV, Wind and Battery Installation expenses and productivity ● This paper assesses the productivity of a hybrid system that in-
cludes combinations of renewable energy generation and energy
storage to satisfy a controllable HVAC load

104]. Wang et al. improved an optimization process based on ACSA total maintenance cost; this sum should be minimized. The opti-
in 2008 by reducing a compound consistency index to determine mization is separately performed for three renewable energy
the best rec-loser and DRG locations. The authors recommended systems, including hybrid systems, solar stand alone and wind
the idea to extend the simultaneous locations of both re-closers stand alone. A complete data set, a regular optimization formula-
and DRGs, which are together dependent on the reliability en- tion and ACOR-based integer programming are the main features
hancement [105]. Additionally, Sookananta et al. [106] planned of this paper. The optimization results showed that this method
ACSA to find the best location and sizing of the DRG in radial gives the best results in only a few seconds. Additionally, the re-
distribution systems, to decrease the general line losses of the net. sults are compared with other artificial intelligence (AI) ap-
Other research, such as in Fetanat and Khorasaninejad [107] used proaches and a conventional optimization method. Moreover, the
ant colony optimization applied to constant domains (ACOR) based results are very promising and prove that the authors’ proposed
digital programming for size optimization in a hybrid photovoltaic approach outperforms them in terms of reaching an optimal so-
(PV)–wind energy system. ACOR is an easy extension of ant colony lution and speed [108].
optimization (ACO). Additionally, it is the notable ant-based algo-
rithm for perpetual optimization. In this setting, the variables are 3.3.2. Algorithm of the Artificial Immune System (AIS)
first considered to be real and are then rounded in each step of the AISs are motivated by immunology, the role of the immune
iteration. The number of solar panels, wind turbines and batteries system and the values viewed in the natural world [109], which
are selected as decision variables of the integer programming was presented in the 1990s. In addition, the immune system is a
problem. The main goal of the design of a PV–wind system is the vital defense against self-approach that protects human health
total design cost, which is the sum of the total capital cost and the from enemies or pathogens such as microbes and viruses. This
 S.M. Zahraee et al. / Renewable and Sustainable Energy Reviews 66 (2016) 617–630 623

Table 4
Brief summary of Artificial Intelligence Methods for Hybrid Energy Systems.

Reference Systems studied Topics covered Highlights

Solar and Wind Systems

[113] Only standalone PV–wind hybrid systems ● Criteria for optimizations and simulation modeling of ● Artificial intelligence techniques are identified to be
photovoltaic systems, wind energy systems, battery promising. which require further exploration
storage systems
● Software tools for hybrid solar–wind system reviewed
are
● HOMER
● HYBRID2
● HOGA
[114] Study covers PV–battery, PV–wind–battery ● Optimization Techniques reviewed: ● GA and PSO as the most useful and promising multi-
and PV–wind–diesel–battery hybrid ● Genetic Algorithms objective optimization methods in hybrid system
systems ● Honey Bee mating Optimization design
● Particle Swarm Optimization
● Evolutionary Algorithm
● Artificial Intelligence
● Pareto-based multi-objective optimization and par-
allel processing
[115] Study includes standalone and grid-con- ● Standalone PV systems size optimization methods ● Artificial intelligence techniques have the potential
nected PV systems, PV–diesel generator reviewed are: to improve the process of optimization
systems, PV–wind systems, PV–wind–die- ● Intuitive methods
sel generator systems ● Numerical methods
● Analytical methods
● Other methods(Artificial Intelligence)
● Reviewed grid connected hybrid systems sizing opti-
mization methods:
● Intuitive methods
● Numerical methods
● Artificial intelligence methods
All Types of Hybrid Energy Systems
[116] Covers all types of renewable energy- ● Software tools discussed are: HOMER, HYBRID2, ● Promising techniques identified are: Ant colony al-
based hybrid systems (ORIENTE). (GAMS). The OptQuest, LINDO, WDILOG2, gorithm, Artificial immune system algorithm, Tabu
Dividing Rectangles (DIRECT). (SimPhoSys), (DOIRES). Search, Honey Bee Algorithm, Bacterial Algorithm,
(GRHYSO), H2RES and The Geo- Spatial Planner for Game Theory
Energy Investment Strategies.
● Optimization techniques discussed are:
● Genetic algorithm
● Particle swarm optimization
● Simulated annealing
● Linear programming
● Simplex algorithm
● Neural Networks
● Evolutionary algorithm
● Random, repeatable, probabilistic, parametric and
numerical way.
[117] Study includes all types of renewable en- ● Reviewed hybrid system performance indicators (Loss ● Study suggests that hybrid optimization methodol-
ergy-based hybrid systems of power supply probability (LPSP), Levelized cost of ogies are superior to other methods.
energy (LCE))
● Hybrid energy system sizing methodologies reviewed
are:
● Probabilistic methods,
● Iterative methods,
● Hybrid methods (Genetic Algorithm, Artificial
Intelligence)
● Analytical methods including software or numerical
approximations of component
[118] Covers all types of renewable energy- ● Discussed design parameters, evaluation criteria and ● Hybrid optimization methodologies are re-
based hybrid systems control and energy management of hybrid energy commended for hybrid systems research to avoid
systems _ Software tools reviewed are the limitations of one methodology
● HOMER
● HOGA
● RETScreen
● HYBRIDS
● TRNSYS
● Sizing methodologies reviewed are:
● Graphic construction methods
● Probabilistic methods
● Analytical methods
● Iterative methods
● Artificial intelligence methods
● Hybrid methods
624 S.M. Zahraee et al. / Renewable and Sustainable Energy Reviews 66 (2016) 617–630

procedure will be able to differentiate among self-cells and non- multi-step wind speed forecasting in terms of the Adaboost
self-cells. Afterward, the immune system executes an immune (Adaptive Boosting) algorithm and the MLP (Multilayer Percep-
answer to remove the non-self-cells [110–112]. Based on the AIS tron) neural networks. In the hybrid Adaboost–MLP forecasting
optimization process for hybrid system sizing, the explanations in architecture, four important algorithms are adopted for the
the search area could be used as encryption in an antigen popu- training and modeling of MLP neural networks, including the GD-
lation in the AIS algorithm. During each iteration, the structure of ALR-BP algorithm, GDM-ALR-BP algorithm, CG-BP-FR algorithm
the antigen populace that was adopted and the peoples’ perfor- and BFGS algorithm. This research aims at investigating the pro-
mance through the eradication of impossible solutions were moted prediction percentages of MLP neural networks by Ada-
evaluated. Previous antigens, including affinities, were supplanted boost algorithm optimization under various training algorithms.
by new antigens that were beneficial to the collection affinities. A The hybrid models in the performance comparison include Ada-
summary of an artificial intelligence application that considers the boost– GD-ALR-BP–MLP, Adaboost–GDM-ALR-BP–MLP, Adaboost–
energy produced based on solar, wind and all types of hybrid CG-BP-FR–MLP, Adaboost–BFGS–MLP, GD-ALR-BP–MLP, GDM-ALR-
energy systems is shown in Table 4. BP–MLP, CG-BP-FR–MLP and BFGS–MLP. Two experimental results
showed the following: (1) the proposed hybrid Adaboost–MLP
3.4. Other promising approaches forecasting architecture is effective for wind speed forecasting;
(2) the Adaboost algorithm has promoted the prediction perfor-
Many methods can be used as conceptual methods in the fu- mance of the MLP neural networks considerably; (3) among the
ture for improving the proficiency and cost-effectiveness in sizing proposed Adaboost–MLP forecasting models, the Adaboost–CG-BP-
hybrid energy systems. The most utilized methods in applications FR–MLP model has the highest performance; and (4) the improved
are the Tabu Search [119,120], honey bee mating algorithm percentages of the MLP neural networks by the Adaboost algo-
[121,122], bacterial food algorithm [123,124], game playing theory rithm decreased step by step with the following sequence of
[125,126] and combination of metaheuristic algorithms. In this training algorithms: GD-ALR-BP, GDM-ALR-BP, CG-BP-FR and BFGS
section, the combination of different methods with artificial in- [130]. Zhang et al. [131] recommends a hybrid computational de-
telligence approaches are reviewed based on their energy pro- sign in terms of Sequential Quadratic Programming (SQP) and
duction, which is produced from, e.g., wind, solar and other hybrid PSOs, to approach the Combined Unit Commitment and Emission
and renewable systems. (CUCE) intricacy. Developing a model that comprises both thermal
dynamos and wind farms, the suggested hybrid computational
3.4.1. Wind systems design manages the scheduling and greenhouse gasses emissions
Detailed predictions of the wind speed and power are essential expenses, and it used a set of mathematical models. The final re-
for improving the safety of renewable energy utilization. Com- sults cited that the proposed hybrid method is better based on the
paring this goal with physical techniques showed that statistical speed and accuracy. The main contribution of this method is the
procedures are usually simpler and produce better fits for small construction of an emissions unit commitment model that is
amounts of land. Determined by the designs of wavelets and es- combined with wind energy and connects the SQP and PSO
tablished time series analysis, a new short-term forecasting ap- methods to obtain rapid and higher performance optimization
proach is proposed. Simulation of real-time data shows that [131].
(1) the relative mean deviation in many-step predictions is related
to the proposed method by a minute. However, this finding is 3.4.2. Solar systems
better than standard time series methodology as well as BP Various optimization methods have been recommended to
methodology. (2) The suggested method is strong in skipping data. identify the parameters of solar cells. Nevertheless, most of them
(3) The proposed method applies to both the wind speed and obtain sub-optimal solutions due to their precipitate convergence
power prediction [127]. Pousinho et al. [128] proposed a stochastic and their difficulty in overcoming local minima in multi-modal
programming method to trade wind energy in a market context problems. Oliva et al. [132] proposed the use of the ABC (Artificial
under ambiguity. The main reason for the profits obtained by the Bee Colony) algorithm to identify the solar cells' parameters ac-
power producers is the energy market values. The unpredictable curately. The ABC algorithm is an evolutionary technique that is
nature of wind power shows an additional source of uncertainty. inspired by the intelligent foraging habits of honey bees. Com-
The wind speed series show uncertain and nonlinear events. paring this technique with other evolutionary algorithms, ABC
The essential and safe use of renewable energy utilization depends manifests a better search capacity to face the multi-modal objec-
on the accurate forecasting of the wind speed. Hybrid models al- tive functions. To illustrate the proficiency of the intended ap-
ways have tremendous accuracy. According to the theories of proach, it is compared to other well-known optimization methods.
Wavelets, particle swarm optimization, artificial neural networks, The experimental results demonstrate the high performance of the
and genetic algorithms, two hybrid predicting frameworks “(GA) proposed method in terms of the robustness and accuracy [132].
and (PSO)” are suggested to predict the non-fixed wind rates. The measurement of the PVT properties of natural gas in gas pi-
Comparisons of the forecasting performance using several algo- pelines, gas storage systems, and gas reservoirs requires accurate
rithm combinations showed that the various components in these values of the compressibility factor. Although the equation of state
two hybrid frameworks involved the following: and empirical correlations were utilized to estimate compressi-
bility factor, the demands for novel, more reliable, and easy-to-use
i. Two proposed hybrid predicting structures involve a proper models encouraged the researchers to introduce modern tools
variety of precision supplied in wind speed calculations, which such as artificial intelligence systems. This paper introduces PSO
can also apply to wind power sources. and GA as population-based stochastic search algorithms to opti-
ii. The augmentation of the GA, as well as the PSO components, is mize the weights and biases of the networks and to prevent
not notable, whereas the Wavelet component is important trapping in local minima. Hence, in this paper, GA and PSO were
[129]. used to minimize the neural network error function [133].
An absolute mathematical paradigm is a useful tool for simu-
The approach of wind speed forecasting plays a leading role in lation, estimation, and management as well as the optimization of
providing the safety of wind power performance. In this paper, solar cell operations. This approach is useful because of the non-
four different hybrid methods are proposed for high-precision linearity of the solar cell models and the limitations of traditional
 S.M. Zahraee et al. / Renewable and Sustainable Energy Reviews 66 (2016) 617–630 625

optimization methods in the identification of unknown factors. hybrid system behavior. It is associated with wind speed and solar
Simulated Bee Swarm Optimization (ABSO) is a newly proposed radiation, battery storage life span, and fuel costs. The Hybrid In-
algorithm that is encouraged by the intelligent characteristics of telligent Algorithm proposed a combination of analysis of a Monte
honey bees. In this paper, the proposed ABSO-based parameter Carlo simulation method and ANN training in a GA optimization
identification procedure is described in terms of diode models that model. It was applied to define input and output data sets that
are composed of a single and double models for a 57-mm diameter were chosen from 519 tasters, which were later utilized to train
commercial (R.T.C.) silicon solar cell. The final results showed that the ANNs and decrease the effort that was needed. The general-
they are quite favorable and outperform those that are conducted ization power of the ANNs was calculated based on the RMSE
by the other research methods [134]. (Root Mean Square Error), MBE (Mean Bias Error), MAE (Mean
Absolute Error) and R-squared estimators while applying another
3.4.3. Other Hybrid systems 200 samples. The conclusions showed that the accessible model
Recently, it has become obvious from studies that hypotheses can symbolize the main characteristics of a hybrid power in non-
based on Knowledge-based professional systems have become reliable operating situations [139].
important instruments for scientists and engineers. Today, it is not Berrazouane and Mohammedi [140] presented the develop-
easy to arrive at solutions by previous methods because there are ment of an optimized fuzzy logic controller (FLC) for operating a
many attractive features that lead to addressing real and difficult standalone hybrid power system, which was based on the cuckoo
engineering problems. Furthermore, by the growing worldwide search algorithm. The FLC inputs are the batteries’ state of charge
demand for various types of energy, highly developed intelligent (SOC) and net power flow, and the FLC outputs are the power rates
forecasting techniques are necessary to establish the basics of of the batteries, photovoltaic and diesel generator. Data for weekly
making decisions. Thus, in this study, a new approach was pre- solar irradiation, ambient temperature and load profile are used to
sented to create professional systems to simulate various types of tune the proposed controller by using the cuckoo search algo-
energy requests with related factors and pressures. The capability rithm. The optimized FLC is able to minimize the loss of power
of this method is evaluated by implementing it in three case stu- supply probability (LPSP), excess energy (EE) and levelized energy
dies, namely, annual electricity demands, natural gas demands and cost (LEC). Moreover, the results of the CS optimization are better
oil products required in Iran. The results from this method (COR- than those of particle swarm optimization PSO for a fuzzy system
ACO-GA) provide more accurate and stable calculations than for controller [140]. The tradeoff between the expense and reliability
neuro fuzzy systems (ANFISs). Additionally, it can support decision of the method is a major bargain in devising hybrid methods. Thus,
makers in appropriate arrangements for a future (subsequent) the Optimization of a Hybrid Micro-Grid System (HMGS) is ex-
period [135]. In one investigation, a multi-agent solution (MAS) to amined. A hybrid wind/PV system with battery storage and a
energy management in a distributed hybrid renewable energy diesel generator is used for this purpose, and the Multi-Objective
system was presented. This system has constituents, character- Particle Swarm Optimization (MOPSO) technique is used to find
istics, and excitation devices. The validation of the MAS showed its the reliable configuration of the system. Data from a set of wind
feasibility in achieving all of the system requirements. Therefore, speeds from three meteorological stations was gathered on an
five varieties of agents are offered, which for each agency is built hourly basis in Iran. These selected stations (Nahavand, Rafsanjan,
in a three-layered architecture. A macro MAS is also shown in and Khash) were examined for HMGS optimizations. However, the
detail. Its framework contains an overall optimization function designs of the systems mentioned above and the results agree in
such as JADE (Java Agent Development). As a result, studies that the MOPSO optimization model produces components of
showed MAS to be a suitable solution for the energy management appropriate sizes at each location. It is also argued that the use of
of the distributed hybrid renewable energy system [136]. To en- HMGS can be considered to be a real option for promoting elec-
hance the amount of efficiency for future projections in China, this trification projects and improving energy access within isolated
study used a hybrid algorithm and PSO along with GA for a top Iranian areas or other developing nations that experience the same
Energy Demand Estimating (PSO–GA EDE) model. The parameters or similar climatic situations [141]. Leou [142] used genetic algo-
of the three types of approaches in the model (linear, exponential, rithms combined with linear programming (GALP) to find the
and quadratic) was optimized by PSO–GA, while applying de- optimum capacity and working performance of a VRB energy
terminants, such as GDP, economic structure, population, rate of storage system. He considered the operations and maintenance
urbanization, and energy consumption structure, which influence costs, installation expenses and incomes as comprising the energy
the demand. The simulation outcomes of the suggested model price for decreasing transmission access expenses and delaying
have increased correctness and reliability compared to other single facility investment. Tan et al. [143] used PSO along with a grav-
optimization approaches over 20 years. [137]. itational search algorithm (GSA) to develop a novel optimization
Another optimization approach used the ANFIS (Adaptive model for the siting and sizing of DGs based on real power losses
Neuro-Fuzzy Inference System) to illustrate the PV and wind ori- and the grid VA requirement. The final results showed that the
gins. The algorithm developed is related to “HOMER (Hybrid Op- proposed approach is effective, robust and proficient for addres-
timization Model for Electric Renewables)” with HOGA (Hybrid sing the mixed integer nonlinear optimization problem. Another
Optimization by Genetic Algorithms) software, and the final re- study proposed a new self-adaptive optimization algorithm based
sults demonstrate a precision of 96% for wind and PV. PSCAD/ on the θ-Particle Swarm Optimization (θ-PSO) algorithm to find
EMTDC was used to simulate the optimized operation, and the the whole search space globally. In this paper, a novel probabilistic
results claimed that a small amount of excess energy is released design based on a 2 m Point Estimate Method (2 m PEM) was
[108]. In another study, ABSO was used by modeling all of the suggested for considering the uncertainties in the optimum energy
segments and describing an objective function that was based on management of MicroGrids (MGs) that involve a variety of re-
the total annual cost. Thus, the maximum acceptable loss in the newable power sources such as Micro Turbine (MT), Photovoltaics
power supply probability (LPSPmax) is determined for a reliable (PVs), storage devices and Wind Turbines (WTs) [144]. Battery
system. This approach used conditional rules to find the global storage and Standalone microgrids with renewable sources play a
solution. As a result, LPSPmax was set to 0%, 0.3% and 1%, and the significant role in handling power supply problems in remote
PV/WT/FC is a highly cost-effective energy system, and thus, at areas such as islands. The lifetime parameter of a battery energy
LPSPmax 2%, the WT/FC is the most cost-effective hybrid system storage system must be fully studied to obtain economic and re-
[138]. A model using an ANN has been recommended to assess liable performance of a standalone microgrid as well as to consider
626 S.M. Zahraee et al. / Renewable and Sustainable Energy Reviews 66 (2016) 617–630

Table 5
Summary of combined different artificial methods.

Reference Systems studied Topics covered Highlights

Wind Systems
[127] wind speed and wind power forecasting ● statistical methods ● error is small, which is better than in classical time series
methodology;
● robust in addressing jumping data;
● applicable to both wind speed and wind power forecasting
[128] trading wind energy in a market en- ● stochastic programming approach ● comparison of optima in the market at different risk levels.
vironment under uncertainty ● proposed method on a realistic case study
[129] predict non-stationary wind speeds ● traditional time series examination, ● both of them are suitable for different accurateness re-
● GA quirements in wind speed forecasting.
● (PSO) and artificial neural networks, ● the GA and the PSO components in improving the MLP are
● two hybrid forecasting frameworks [(GA)-(PSO)] not significant, whereas the Wavelet component is
significant
[130] wind speed prediction ● MLP neural networks, including ● Adaboost–the MLP forecasting architecture is effective for
● GD-ALR-BP algorithm wind speed predictions
● GDM-ALR-BP algorithm ● the Adaboost algorithm has promoted the forecasting
● CG-BP-FR algorithm performance of the MLP neural networks considerably;
● BFGS algorithm ● the Adaboost–CG-BP-FR–MLP model has the best
performance
● Adaboost algorithm decreases step by step with the se-
quence of training algorithms
[131] minimize the scheduling cost and green- ● Sequential Quadratic Programming (SQP) and ● Implementation of a release unit model with wind energy
house gases emissions cost Particle Swarm Optimization (PSO) when combining the SQP and PSO
Solar Systems
[132] accurately identify the solar cells' ● ABC (artificial bee colony) algorithm ● high performance of the proposed method in terms of
parameters robustness and accuracy
[133] solar cell systems ● Particle swarm optimization (PSO) ● Helpful tool for simulation, estimation, management, and
● Genetic algorithm (GA) optimization of solar cells.
● meta heuristic algorithms have attracted significant
attention
[134] diameter commercial (R.T.C. France) sili- ● Artificial bee swarm optimization (ABSO) ● The results found by the other studied methods
con solar cell
Other Types of Hybrid Systems
[135] construct expert systems by ability in ● supportive (COR-ACO-GA) ● to signify that COR ACO GA
modeling and imitation of energy ● adaptive (ANFISs)
commands ● (ANNs)
[136] energy management in a distributed hy- ● multi-agent (MAS) solution ● suitable solution for energy management of a distributed
brid renewable energy hybrid renewable energy generation system
[137] optimal Energy Demand Estimating (PSO– ● Particle Swarm Optimization ● proposed model has greater accuracy and reliability than
GA EDE) model, for China ● and Genetic Algorithm other single optimization methods
[138] Optimally size a hybrid energy system ● Artificial bee swarm optimization (ABSO) ● stochastic rules to escape local optima and find a global
solution
[139] wind and solar, battery life span, and fuel ● ANN (artificial neural network) ● it can represent the main uniqueness of a typical hybrid
prices ● Monte Carlo simulation approach power system under doubtful operating conditions
● genetic algorithm optimization model
[140] operating a standalone hybrid power ● optimized fuzzy logic controller (FLC) ● results of CS optimization are better than those of particle
system ● cuckoo search algorithm swarm optimization (PSO) for a fuzzy system controller
[141] Hybrid Micro-Grid System (HMGS) ● Multi-Objective Particle Swarm Optimization ● MOPSO optimization model produces suitable sizing of the
(MOPSO) elements for each place
[144] renewable power sources ● 2 m Point Estimate Method (2 mPEM) with a ● A novel self-adaptive modification approach based on the
self-adaptive modification PSO method (SAM-θ- θ-PSO algorithm was proposed.
PSO) ● Several renewable sources such as PV, WT, FC and MT as
well as storage devices are considered.
● θ-PSO algorithm is used for the first time to solve MG
operations management.
[145] standalone microgrid ● a multi-objective optimization and non-domi- ● Minimize the power generation cost and maximize the life
nated sorting genetic algorithm (NSGA-II) of the lead–acid batteries

the rate of renewable resource utilization. In this research, a intelligence methods to optimize the hybrid energy systems.
standalone microgrid on Dongfushan Island in China was selected
as a case study to investigate its economic operation. To achieve
this goal, an optimization model composed of operations and 4. Results and discussion
maintenance expenses, battery life losses and environmental costs
was proposed to find a set of optimum factors of operations The increase in global energy requirements and the environ-
strategies. Therefore, a multi-objective optimization and non- mental issues that are based on fossil energy performance have
dominated sorting genetic algorithm (NSGA-II) was combined to encouraged the broad, extensive study of the use of renewable
minimize the power generation expenses as well as to maximize energy techniques instead of traditional fossil fuels. Specifically,
the applicable life of the batteries. The final results showed that hybrid systems, which are described as a combination of renew-
the suggested approach can optimize the system operation by able and back-up parts or traditional energy sources, play a sig-
considering a variety of scenarios, and it can aid users in achieving nificant role in finding suitable solutions for handling the chal-
the optimum operation design of the actual microgrid system lenges that the world confronts today with regard to the sustain-
[145]. Table 5 indicates a summary of other combined artificial ability concerns of energy demands and environmental safety.
 S.M. Zahraee et al. / Renewable and Sustainable Energy Reviews 66 (2016) 617–630 627

Hybrid energy systems can be designed and optimized to satisfy AIS has a higher potential to be used in sizing studies based on
the essential requirements of an area based on different factors, its likeness to the GA and a possible effective route to finding the
such as topography, energy availability of potential resources and comprehensive optimum in difficult problems. However, the GA
types of energy demands. Therefore, the best sizing of the re- applicability is greater than the AIS in its ability to address a large
newable energy sources, which relates to hybrid systems, con- number of parameters. Most of the time, hybrid optimization
siderably enhances the economic and industrial aspects of the methods have been proposed to combine two or more meth-
supplied power performance, such as encouraging the use of such odologies to improve them, to increase their convergence time in
environmentally friendly sources. Several sizing methods are used the optimization process. These methods can be characterized
to find the best hybrid renewable energy system in terms of because of their flexibility and dynamics during the sizing process.
technology and economy. Finding the optimum design of hybrid Therefore, they are the most dominant sizing methodologies.
renewable energy systems can be important in increasing the
economic and technical efficiency of the power supply and in
encouraging the extensive use of environmental resources. 5. Conclusions
Different sizing methodologies in existing software with dis-
similar optimization methods are tested here. Thus, every sizing This study presents a summary of prior research concerning the
procedure has its attributes, and many new approaches are pos- use of optimization in artificially intelligent algorithms for de-
sible for upcoming usage. Additionally, by the selection of the signing, planning and controlling problems in the field of hybrid
appropriate method, the types of tools and the users’ needs could energy systems. There were over one hundred papers reviewed
change. As a result, every developed sizing method has the po- from the major and popular referenced journals in the areas of
tential to substantially improve renewable energy systems and, renewable energy and computational optimization that offer im-
thus, has an enormous significance in the renewable energy area. portant and useful conclusions for renewable energy research.
The ease of leaving out a local minimum and the well-organized Optimization studies during the past 2.5 decades by researchers
capability of finding the universal optimal is the most significant using traditional and new generation methods are analyzed, and
benefit of GAs when using hybrid system sizing. This advantage of optimization methods, including hybrid algorithms, are presented.
code-ability makes it appropriate in sizing studies because this Artificial intelligence algorithms are mostly used during the past
advantage is not accessible in other methods such as PSO, which decade because they utilize less computational time and have
will be described below. For example, the application of GA and better accuracy, with good convergence in comparison to tradi-
PSO can be encoded into three parameters at most (calculated in tional methods. In conclusion, this study shows, in the beginning,
the next section), which includes more than three elements, such the number of investigations that use optimization methods in
as in the PV wind fuel cell. Moreover, the GA method does not solving renewable energy problems, mainly for wind and solar
require non-original data. Nevertheless, the GA is difficult to code energy systems. There are many research papers that use heuristic
because of its complex structure. Furthermore, if the number of optimization methods, especially GAs and PSOs, to address these
parameters is larger, the GA would be more difficult, and there problems. However, there are some optimization techniques that
would be an increase in the response time of the GA. The PSO has involve traditional methods, such as mixed-integer and interval
some benefits over GA, even though both the GA and PSO method linear-programming, Lagrangian relaxation, quadratic program-
have brilliant effectiveness in using the same repeatable searching ming, and Nelder–Mead Simplex search. Future research could
approach. This method is very easy to implement in any optimi- pave the way for hybridization and multi-objective implementa-
zation software and is very fast. However, if there are more than tions of bio-inspired solutions. Attempts to exploit the advantages
three elements, it would be more capable when utilizing the GA and disadvantages of different algorithms have been made by
approach. implementing hybrid algorithms. These approaches have been
This method is a simple perception that involves an easy im- proven to be faster, more accurate and more powerful than in-
plementation in a software environment. Consequently, the cal- dividual systems. The choice of algorithm to be implemented de-
culation time is short, and the requirement for recollection is low. pends solely on the application, and hence, a thorough under-
However, the reliability of finding the global optimum in a search standing is needed to justify the merits and demerits. For example,
area is lower than with a GA-based method. Additionally, the PSO high dimensional problems such as sizing can be addressed better
approach is less appropriate than the GA in problems that include by using PSO rather than GAs. Artificial intelligence techniques are
a coordinate description of particle bases, which have more than also applied with optimization algorithms in some power system
three PSO parameters, and it can only be identified on the x, y, z applications. This review will be useful for researchers, to address
plane. For example, when considering a PV wind fuel cell hybrid the complexity and challenges in renewable energy-based hybrid
system in which the size is optimized, the x-axis is used to re- systems research.
present the PV panels’ numbers. The y-axis is the number of wind
turbines, and the z-axis is based on the fuel cell system in the KW.
As a result, the three components in using the PSO, which is more Acknowledgment
capable than the GA, are explained below. However, as mentioned
before, if there are more than three available components, then it This work was supported by University Research Internal Fund
is more suitable to use the GA method as a replacement for the (URIF, Cost center: 0153AA-B84) from the Universiti Teknologi
PSO. In addition, using SA in a hybrid system sizing is not as PETRONAS, Malaysia.
prevalent as methods such as the GA or PSO, but currently, re-
search interest in SA is increasing, and the approved area of use is
growing. The ACS algorithm has been proposed to reduce the References
distribution of system losses and to balance the factors of radial
distribution. Additionally, it has proven to be better than the GA by [1] Sadeghifam AN, Zahraee SM, Meynagh MM, Kiani I. Combined use of design
achieving 44.626% as an average loss reduction. Similar to the GA, of experiment and dynamic building simulation in assessment of energy
the AIS's fundamental optimization has “collection” and “transfor- efficiency in tropical residential buildings. Energy Build 2015;86:525–33.
[2] Goedeckeb M, Therdthianwong S, Gheewala SH. Life cycle cost analysis of
mation” operatives, which can considerably improve the chance of alternative vehicles and fuels in Thailand. Energy Policy 2007;35(6):3236–46.
the algorithm finding the most globally advantageous point. [3] Straatman PJT, van Sark WGJHM. A new hybrid ocean thermal energy
628 S.M. Zahraee et al. / Renewable and Sustainable Energy Reviews 66 (2016) 617–630

conversion-offshore solar pond (OTEC-OSP) design: a cost optimization ap- Release. 〈http://www.eia.gov/forecasts/aeo/er/pdf/0383er(2012).pdf〉. (Ac-
proach. Sol Energy 2008;82(6):520–7. cessed 12 December 2014).
[4] Rehman S, El-Amin IM, Ahmad F, Shaahid SM, Al-Shehri AM, Bakhashwain [35] Wang C. Modeling and control of hybrid wind/photovoltaic/fuel cell dis-
JM, et al. Feasibility study of hybrid retrofits to an isolated off-grid diesel tributed generation systems (Ph.D. Thesis). Department of Electrical and
power plant. Renew Sustain Energy Rev 2007;11(4):635–53. Computer Engineering, Montant State University; 2006.
[5] Hosseini SE, Wahid MA. Utilization of palm solid residue as a source of re- [36] U.S. Energy Information Administration. Electric Power Monthly. 〈http://
newable and sustainable energy in Malaysia. Renew Sustain Energy Rev www.eia.gov/energy_in_brief/article/renewable_electricity.cfm〉. (Accessed
2014;40:621–32. 12 December 2014).
[6] Ball M, Wietschel M, Rentz O. Integration of a hydrogen economy into the [37] European Photovoltaic Industry Association. Global Market Outlook for
German energy system: an optimising modelling approach. Int J Hydrog Photovoltaics 2013–2017. 〈http://www.epia.org/ fileadmin/user_upload/Pub
Energy 2007;32(10–11):1355–68. lications/GMO_2013_-_Final_PDF.pdf〉. (Accessed 12 December 2014).
[7] Shaahid SM, Elhadidy MA. Technical and economic assessment of grid-in- [38] GWEC. Global Wind Report – Annual Market Update 2012. 2013.
dependent hybrid photovoltaic–diesel–battery power systems for commer- [39] Ter-Gazarian AG, Kagan N. Design model for electrical distribution systems
cial loads in desert environments. Renew Sustain Energy Rev 2007;11 considering renewable, conventional and energy storage units. IEE Proc C:
(8):1794–810. Gener Transm Distrib 1992;139(6):499–504.
[8] Hosseini SE, Abdul Wahid M. The role of renewable and sustainable energy [40] Anglani N, Muliere G. Analyzing the impact of renewable energy technolo-
in the energy mix of Malaysia: a review. Int J Energy Res 2014;38(14):1769– gies by means of optimal energy planning. In: Proceedings of the 9th in-
92. ternational conference on environment and electrical engineering (EEEIC).
[9] Yilmaz P, Hocaoglu MH, Konukman AES. A pre-feasibility case study on in- 2010. p. 1–5.
tegrated resource planning including renewables. Energy Policy 2008;36 [41] Mizani S, Yazdani A. Design and operation of a remote microgrid. In: Pro-
(3):1223–32. ceedings of the 35th annual conference on industrial electronics (IECON’09).
[10] Bargshady G, Ahmadi M, Abdulrazzaq AW, Zahraee SM. Evaluation of firm's 2009. p. 4299–304.
potential in adoption of green IT. Am Eurasia J Sustain Agric 2014:8–14. [42] Borowy BS, Salameh ZM. Methodology for optimally sizing the combination
[11] Zahraee SM, Kafuku JM. An empirical survey of supplier participation in of a battery bank and PV array in a Wind/PV hybrid system. IEEE Trans
sustainable green supply chain: a case study of Malaysian automotive Energy Convers 1996;11(2):367–75.
manufacturers. Am Eurasia J Sustain Agric 2014:1–8. [43] Markvart T. Sizing of hybrid photovoltaic-wind energy systems. Sol Energy
[12] Zahraee SM, Khademi A, Khademi S, Abdullah A, Ganjbakhsh H. Application 1996;57(4):277–81.
of design experiments to evaluate the effectiveness of climate factors on [44] Karaki S, Chedid R, Ramadan R. Probabilistic performance assessment of
energy saving in green residential buildings. J Teknol 2014;69:5. autonomous solar-wind energy conversion systems. IEEE Trans Energy
[13] Zahraee SM, Hatami M, Bavafa AA, Ghafourian K, Rohani JM. Application of Convers 1999;14(3):766–72.
statistical taguchi method to optimize main elements in the residential [45] Kellogg W, Nehrir M, Venkataramanan G, Gerez V. Optimal unit sizing for a
buildings in Malaysia based energy consumption. Appl Mech Mater hybrid wind/photovoltaic generating system. Electr Power Syst Res 1996;39
2014;606:265–9 [Trans Tech Publications]. (1):35–8.
[14] Zhou W, Lou C, Li Z, Lu L, Yang H. Current status of research on optimum [46] Chedid R, Saliba Y. Optimization and control of autonomous renewable en-
sizing of stand-alone hybrid solar–wind power generation systems. Appl ergy systems. Int J Energy Res 1996;20(7):609–24.
Energy 2010;87(2):380–9. [47] Goncalves JF, Mendes JJM, Resende MGC. A genetic algorithm for the re-
[15] Kornelakis A. Multiobjective particle swarm optimization for the optimal source constrained multi-project scheduling problem. Eur J Oper Res
design of photovoltaic grid-connected systems. Sol Energy 2010;84 2008;189(3):1171–90.
(12):2022–33. [48] Chen SM, Huang CM. A new approach to generate weighted fuzzy rules using
[16] Hosseini SE, Wahid MA, Aghili N. The scenario of greenhouse gases reduction genetic algorithms for estimating null values. Expert Syst Appl 2008;35
in Malaysia. Renew Sustain Energy Rev 2013;28:400–9. (3):905–17.
[17] Hosseini SE, Abdul Wahid M. Pollutant in palm oil production process. J Air [49] Holland JH. Adaptation in natural and artificial systems. Massachusetts,
Waste Manag Assoc 2015;65(7):773–81. Cambridge: MIT Press; 1975.
[18] Zhou W, Yang H, Fang Z. Battery behavior prediction and battery working [50] Yang H, Zhou W, Lu L, Fang Z. Optimal sizing method for stand-alone hybrid
states analysis of a hybrid solar–wind power generation system. Renew solar–wind system with LPSP technology by using genetic algorithm. Sol
Energy 2008;33(6):1413–23. Energy 2008;82(4):354–67.
[19] Golroudbary SR, Zahraee SM. System dynamics model for optimizing the [51] Yang H, Zhou W, Lou C. Optimal design and techno-economic analysis of a
recycling and collection of waste material in a closed-loop supply chain. hybrid solar–wind power generation system. Appl Energy 2009;86(2):163–9.
Simul Model Pract Theory 2015;53:88–102. [52] Bilal BO, Sambou V, Ndiaye PA, Kébé CMF, Ndongo M. Optimal design of a
[20] Hepbasli A. A key review on exergetic analysis and assessment of renewable hybrid solar–wind–battery system using the minimization of the annualized
energy resources for a sustainable future. Renew Sustain Energy Rev 2008;12 cost system and the minimization of the loss of power supply probability
(3):593–661. (LPSP). Renew Energy 2010;35(10):2388–90.
[21] Nayar CV, Islam SM, Dehbonei H, Tan K, Sharma H. Power electronics for [53] Koutroulis E, Kolokotsa D, Potirakis A, Kalaitzakis K. Methodology for optimal
renewable energy sources. In: Rashid MH, editor. Power electronics hand- sizing of stand-alone photovoltaic/wind-generator systems using genetic
book. 2nd ed.. United Kingdom: Academic Press – Imprint of Elsevier Inc; algorithms. Sol Energy 2006;80(9):1072–88.
2007. p. 673–716. [54] Koutroulis E, Kolokotsa D. Design optimization of desalination systems
[22] Yang H, Lu L, Zhou W. A novel optimization sizing model for hybrid solar– power-supplied by PV and W/G energy sources. Desalination 2010;258(1–
wind power generation system. Sol Energy 2007;81(1):76–84. 3):171–81.
[23] Chen F, Duic N, Alves LM, Carvalho MG. Islands – renewable energy solutions [55] Lagorse J, Paire D, Miraoui A. Hybrid stand-alone power supply using PEMFC,
for islands. Renew Sustain Energy Rev 2007;11(8):1888–902. PV and battery –modelling and optimization. In: Proceedings of the Inter-
[24] Alawi AA, Alawi SMA, Islam SM. Predictive control of an integrated PV–diesel national conference on clean electrical power (ICCEP). 2009. p. 135–40.
water and power supply system using an artificial neural network. Renew [56] Kalantar M, Mousavi SMG. Dynamic behavior of a stand-alone hybrid power
Energy 2007;32(8):1426–39. generation system of wind turbine, microturbine, solar array and battery
[25] Ekren O, Ekren BY. Size optimization of a PV/wind hybrid energy conversion storage. Appl Energy 2010;87(10):3051–64.
system with battery storage using simulated annealing. Appl Energy 2010;87 [57] Lopez RD, Agustin JLB. Design and control strategies of PV–diesel systems
(2):592–8. using genetic algorithms. Sol Energy 2005;79(1):33–46.
[26] Capizzi G, Tina G. Long-term operation optimization of integrated generation [58] Lopez RD, Agustin JLB. Influence of mathematical models in design of PV–
systems by fuzzy logic-based management. Energy 2007;32(7):1047–54. diesel systems. Energy Convers Manag 2008;49(4):820–31.
[27] Mellit A, Benghanem M, Kalogirou SA. Modeling and simulation of a stan- [59] Lagorse J, Paire D, Miraoui A. Sizing optimization of a stand-alone street
dalone photovoltaic system using an adaptive artificial neural network: lighting system powered by a hybrid system using fuel cell, PV and battery.
proposition for a new sizing procedure. Renew Energy 2007;32(2):285–313. Renew Energy 2009;34(3):683–91.
[28] Anagnostopoulos JS, Papantonis DE. Pumping station design for a pumped [60] Zhao M, Chen Z, Blaabjerg F. Optimization of electrical system for offshore
storage wind–hydro power plant. Energy Convers Manag 2007;48(11):3009– wind farms via genetic algorithm. IE Trans Renew Power Gener 2009;3
17. (2):205–16.
[29] Phuangpornpitak N, Kumar S. PV hybrid systems for rural electrification in [61] Li H, Chen Z, Polinder H. Optimization of multibrid permanent-magnet wind
Thailand. Renew Sustain Energy Rev 2007;11(7):1530–43. generator systems. IEEE Trans Energy Convers 2009;24(1):82–92.
[30] Hwang JJ, Chang WR, Su A. Dynamic modeling of a solar hydrogen system [62] Rǎ suo BP, Bengin AC. Optimization of wind farm layout. FME Trans
under leakage conditions. Int J Hydrog Energy 2008;33(14):3615–24. 2010;38:107–14.
[31] Russell SJ, Norvig P. Artificial intelligence: a modern approach. 2nd ed.. Up- [63] Grady SA, Hussaini MY, Abdullah MM. Placement of wind turbines using
per Saddle River; 2003. genetic algorithms. Renew Energy 2005;30(12):259–70.
[32] Bhandari B, Lee KT, Lee GY, Cho YM, Ahn SH. Optimization of hybrid re- [64] Emami A, Noghreh P. New approach on optimization in placement of wind
newable energy power systems: a review. Int J Precis Eng Manuf Green turbines within wind farm by genetic algorithms. Renew Energy 2010;35
Technol 2015;2(1):99–112. (7):1559–64.
[33] U.S. Energy Information Administration. International Energy Outlook 2011. [65] Li QS, Liu DK, Fang JQ, Tam CM. Multi-level optimal design of buildings with
〈http://www.eia.gov/forecasts/archive/ieo11〉/(Accessed 12 December 2014). active control under winds using genetic algorithms. J Wind Eng Ind Aero-
[34] U.S. Energy Information Administration. Annual Energy Outlook 2012 Early dyn 2000;86(1):65–86.
 S.M. Zahraee et al. / Renewable and Sustainable Energy Reviews 66 (2016) 617–630 629

[66] Kalogirou SA. Optimization of solar systems using artificial neural-networks [94] Nasiraghdam H, Jadid S. Optimal hybrid PV/WT/FC sizing and distribution
and genetic algorithms. Appl Energy 2004;77(4):383–405. system reconfiguration using multi-objective artificial bee colony (MOABC)
[67] Varun S. Thermal performance optimization of a flat plate solar air heater algorithm. Sol Energy 2012;86(10):3057–71.
using genetic algorithm. Appl Energy 2010;87(5):1793–9. [95] Alsayed M, Cacciato M, Scarcella G, Scelba G. Multicriteria optimal sizing of
[68] Zagrouba M, Sellami A, Bouaicha M, Ksouri M. Identification of PV solar cells photovoltaic-wind turbine grid connected systems. Energy Convers IEEE
and modules parameters using the genetic algorithms: application to max- Trans 2013;28(2):370–9.
imum power extraction. Sol Energy 2010;84(5):860–6. [96] Aghaei J, Akbari MA, Roosta A, Gitizadeh M, Niknam T. Integrated renewable-
[69] Miyatake M, Toriumi F, Endo T, Fujii N. A novel maximum power point conventional generation expansion planning using multiobjective frame-
tracker controlling several converters connected to photovoltaic arrays with work. Gener Transm Distrib IET 2012;6(8):773–84.
particle swarm optimization technique. In: Proceedings of the European [97] Katsigiannis YA, Georgilakis PS, Karapidakis ES. Multiobjective genetic al-
conference on power electronics and applications. 2007. p. 1–10. gorithm solution to the optimum economic and environmental performance
[70] Hakimi SM, Tafreshi SMM, Kashefi A. Unit sizing of a stand-alone hybrid problem of small autonomous hybrid power systems with renewables. Re-
power system using particle swarm optimization (PSO). In: Proceedings of new Power Gener IET 2010;4(5):404–19.
the IEEE international conference on automation and logistics. 2007. p. 3107– [98] Brekken TK, Yokochi A, Von Jouanne A, Yen ZZ, Hapke HM, Halamay DA.
12. Optimal energy storage sizing and control for wind power applications.
[71] Sanchez V, Ramirez JM, Arriaga G. Optimal sizing of a hybrid renewable Sustain Energy IEEE Trans 2011;2(1):69–77.
system. In: Proceedings of the IEEE international conference on industrial [99] Arabali A, Ghofrani M, Etezadi-Amoli M, Fadali MS, Baghzouz Y. Genetic-
technology (ICIT). 2010. p. 949–54. algorithm-based optimization approach for energy management. Power
[72] Dehghan S, Saboori H, Parizad A, Kiani B. Optimal sizing of a hydrogen-based Deliv IEEE Trans 2013;28(1):162–70.
wind/PV plant considering reliability indices. In: Proceedings of the inter- [100] Santarelli M, Pellegrino D. Mathematical optimization of a RES-H2 plant
national conference on electric power and energy conversion systems using a black box algorithm. Renew Energy 2005;30(4):493–510.
(EPECS’09). 2009. p. 1–9. [101] Margeta J, Glasnovic Z. Feasibility of the green energy production by hybrid
[73] Kaviani AK, Riahy GH, Kouhsari SHM. Optimal design of a reliable hydrogen solar þhydro power system in Europe and similar climate areas. Renew
based stand-alone wind/PV generating system, considering component Sustain Energy Rev 2010;14(6):1580–90.
outages. Renew Energy 2009;34(11):2380–90. [102] Cabral CVT, Filho DO, Diniz ASAC, Martins JH, Toledo OM, Neto LVBM. A
[74] Tafreshi SMM, Hakimi SM. Optimal sizing of a stand-alone hybrid power stochastic method for stand-alone photovoltaic system sizing. Sol Energy
system via particle swarm optimization (PSO). In: Proceedings of the Inter- 2010;84(9):1628–36.
national power engineering conference (IPEC). 2007. p. 960–5. [103] Tan CW, Green TC, Aramburo CAH. A stochastic method for battery sizing
[75] Hakimi SM, Tafreshi SMM. Optimal sizing of a stand-alone hybrid power with uninterruptible-power and demand shift capabilities in PV (photo-
system via particle swarm optimization for Kahnouj area in south-east of voltaic) systems. Energy 2010;35(12):5082–92.
Iran. Renew Energy 2009;34(7):1855–62. [104] Katti PK, Khedkar MK. Alternative energy facilities based on site matching
[76] Wang L, Singh C. PSO-based multidisciplinary design of a hybrid power and generation unit sizing for remote area power supply. Renew Energy
generation system with statistical models of wind speed and solar isolation. 2007;32(8):1346–62.
In: Proceedings of the International conference on power electronics, drives [105] Wang L, Singh C. Reliability-constrained optimum placement of reclosers and
and energy systems (PEDES). 2006. p. 1–6. distributed generators in distribution networks using an Ant Colony System
[77] Zhao YS, Zhan J, Zhang Y, Wang DP, Zou BG. The optimal capacity config- Algorithm. IEEE Trans Syst Man Cybern Part C: Appl Rev 2008;38(6):757–64.
uration of an independent wind/PV hybrid power supply system based on [106] Sookananta B, Utaton P, Khongsila R. Determination of the optimal location
improved PSO algorithm. In: Proceedings of the 8th international conference and sizing of Distributed Generation using Ant Colony Search. In: Proceed-
on advances in power system control, operation and management (APS- ings of the international conference on electrical engineering/electronics
COM). 2009. p. 1–7. computer telecom- munications and information technology (ECTI-CON).
[78] Wang L, Singh C. PSO-based multi-criteria optimum design of a grid- 2010. p. 814–7.
connected hybrid power system with multiple renewable sources of energy. [107] Fetanat A, Khorasaninejad E. Size optimization for hybrid photovoltaic–wind
In: Proceedings of the IEEE swarm intelligence symposium (SIS). 2007. p. energy system using ant colony optimization for continuous domains based
250–7. integer programming. Appl Soft Comput 2015;31:196–209.
[79] Wang L, Singh C. Multi criteria design of hybrid power generation systems [108] Rajkumar RK, Ramachandaramurthy VK, Yong BL, Chia DB. Techno-eco-
based on a modified particle swarm optimization algorithm. IEEE Trans nomical optimization of hybrid pv/wind/battery system using Neuro-Fuzzy.
Energy Convers 2009;24(1):163–72. Energy 2011;36(8):5148–53.
[80] Zhao J, Li X, Hao J, Lua J. Reactive power control of a wind farm made up with [109] Prakash A, Deshmukh SG. A multi-criteria customer allocation problem in
doubly fed induction generators in distribution system. Electr Power Syst Res supply chain environment: an artificial immune system with fuzzy logic
2010;80(6):698–706. controller based approach. Expert Syst Appl 2011;38(4):3199–208.
[81] Pourmousavi SA, Nehrir MH, Colson CM, Wang C. Real-time energy man- [110] Aydin I, Karakose M, Akin E. A multi-objective artificial immune algorithm for
agement of a stand-alone hybrid wind-microturbine energy system using parameter optimization in support vector machine. Appl Soft Comput
particle swarm optimization. Sustain Energy IEEE Trans 2010;1(3):193–201. 2011;11(1):120–9.
[82] Su Z, Wang J, Lu H, Zhao G. A new hybrid model optimized by an intelligent [111] Gong M, Jiao L, Zhang X. A population-based artificial immune system for
optimization algorithm for wind speed forecasting. Energy Convers Manag numerical optimization. Neurocomputing 2008;72(1–3):149–61.
2014;85:443–52. [112] Basu M. Artificial immune system for dynamic economic dispatch. Electr
[83] Kahrobaee S, Asgarpoor S. Optimum planning and operation of compressed Power Energy Syst 2011;33(1):131–6.
air energy storage with wind energy integration. In: Proceedings of the [113] Zhou W, Lou C, Li Z, Lu L, Yang H. Current status of research on optimum
North American power symposium. Vol. 1. 2013. p. 1–6. sizing of stand-alone hybrids solar–wind power generation systems. Appl
[84] Bhumkittipich K, Phuangpornpitak W. Optimal placement and sizing of Energy 2010;87:380–9.
distributed generation for power loss reduction using particle swarm opti- [114] Fadaee M, Radzi MAM. Multi-objective optimization of a stand-alone hybrid
mization. Energy Procedia 2013;34:307–17. renewable energy system by using evolutionary algorithms: a review. Renew
[85] Sharafi M, ELMekkawy TY. Multi-objective optimal design of hybrid re- Sustain Energy Rev 2012;16:3364–9.
newable energy systems using PSO-simulation based approach. Renew En- [115] Khatib T, Mohamed A, Sopian K. A review of photovoltaic systems size op-
ergy 2014;68:67–79. timization techniques. Renew Sustain Energy 2013;22:454–65.
[86] Giannakoudis G, Papadopoulos AI, Seferlis P, Voutetakis S. Optimum design [116] Erdinc O, Uzunoglu M. Optimum design of hybrid renewable energy systems:
and operation under uncertainty of power systems using renewable energy over view of different approaches. Renew Sustain Energy Rev 2012;16:1412–
sources and hydrogen storage. Int J Hydrog Energy 2010;35(3):872–91. 25.
[87] Vidal RVV. Applied simulated annealing. Berlin: Springer-Verlag; 1993. [117] Luna-Rubio R, Trejo-Peres M, Vargas-Vazquez D, Rı´os-Moreno GJ. Optimal
[88] Roa-Sepulveda CA, Pavez-Lazo BJ. A solution to the optimal power flow using sizing of renewable hybrids energy systems: a review of methodologies. Sol
simulated annealing. Int J Electr Power Energy Syst 2003;25(1):47–57. Energy 2012;86:1077–88.
[89] Sutthibun T, Bhasaputra P. Multi-objective optimal distributed generation [118] Upadhyay S, Sharma MP. A review on configurations, control and sizing
placement using simulated annealing. In: Proceedings of the international methodologies of hybrid energy systems. Renew Sustain Energy 2014;38:47–
conference on electrical engineering/electronics computer telecommunica- 63.
tions and information technology (ECTI-CON). 2010. p. 810–3. [119] Kalinli A, Karaboga D. Training recurrent neural networks by using parallel
[90] Ghadimi N, Ghadimi R. Optimal allocation of distributed generation and tabu search algorithm based on crossover operation. Eng Appl Artif Intell
capacitor banks in order to loss reduction in reconfigured system. Res J Appl 2004;17(5):529–42.
Sci Eng Technol 2012;4(9):1099–104. [120] Valls V, Perez MA, Quintanilla MS. A tabu search approach to machine
[91] Fung CC, Hoand SCY, Nayar CV. Optimisation of a hybrid energy system using scheduling. Eur J Oper Res 1998;106(2–3):277–300.
Simulated Annealing technique. IEEETENCON 1993;5:235–8. [121] Marinaki M, Marinakis Y, Zopounidis C. Honey bees mating optimization
[92] Katsigiannis YA, Georgilakis PS, Karapidakis ES. Hybrid simulated annealing– algorithm for financial classification problems. Appl Soft Comput 2010;10
tabu search method for optimal sizing of autonomous power systems with (3):806–12.
renewables. IEEE Trans Sustain Energy 2012;3(3):330–8. [122] Marinakis Y, Marinaki M, Dounias G. Honey bees mating optimization al-
[93] Meza JLC, Yildirim MB, Masud AS. A model for the multiperiod multi- gorithm for the Euclidean traveling salesman problem. Inf Sci 2011;181
objective power generation expansion problem. Power Syst IEEE Trans (20):4684–98.
2007;22(2):871–8. [123] Panigrahi BK, Pandi VR, Sharma R, Das S, Das S. Multiobjective bacteria
630 S.M. Zahraee et al. / Renewable and Sustainable Energy Reviews 66 (2016) 617–630

foraging algorithm for electrical load dispatch problem. Energy Convers colony optimization-genetic algorithm approach for construction of energy
Manag 2011;52(2):1334–42. demand forecasting knowledge-based expert systems. Knowl Based Syst
[124] Hota PK, Barisal AK, Chakrabarti R. Economic emission load dispatch through 2013;39:194–206.
fuzzy based bacterial foraging algorithm. Electr Power Energy Syst 2010;32 [136] Jun Z, Junfeng L, Jie W, Ngan HW. A multi-agent solution to energy man-
(7):794–803. agement in hybrid renewable energy generation system. Renew Energy
[125] Hennet JC, Arda Y. Supply chain coordination: a game-theory approach. Eng 2011;36(5):1352–63.
Appl Artif Intell 2008;21(3):399–405. [137] Yu S, Wei YM, Wang K. A PSO–GA optimal model to estimate primary energy
[126] Sharma R, Gopal M. Synergizing reinforcement learning and game theory – a demand of China. Energy Policy 2012;42:329–40.
new direction for control. Appl Soft Comput 2010;10(3):675–88. [138] Maleki A, Askarzadeh A. Artificial bee swarm optimization for optimum
[127] Liu H, Tian HQ, Chen C, Li YF. A hybrid statistical method to predict wind sizing of a stand-alone PV/WT/FC hybrid system considering LPSP concept.
speed and wind power. Renew Energy 2010;35(8):1857–61. Sol Energy 2014;107:227–35.
[128] Pousinho HMI, Mendes VMF, Catalão JPS. A risk-averse optimization model [139] Lujano-Rojas JM, Dufo-López R, Bernal-Agustín JL. Probabilistic modelling
for trading wind energy in a market environment under uncertainty. Energy and analysis of stand-alone hybrid power systems. Energy 2013;63:19–27.
2011;36(8):4935–42. [140] Berrazouane S, Mohammedi K. Parameter optimization via cuckoo optimi-
[129] Liu H, Tian HQ, Chen C, Li YF. An experimental investigation of two wavelet- zation algorithm of fuzzy controller for energy management of a hybrid
MLP hybrid frameworks for wind speed prediction using GA and PSO opti- power system. Energy Convers Manag 2014;78:652–60.
mization. Int J Electr Power Energy Syst 2013;52:161–73. [141] Maleki A, Pourfayaz F. Optimal sizing of autonomous hybrid photovoltaic/
[130] Liu H, Tian HQ, Li YF, Zhang L. Comparison of four adaboost algorithm based wind/battery power system with LPSP technology by using evolutionary al-
artificial neural networks in wind speed predictions. Energy Convers Manag gorithms. Sol Energy 2015;115:471–83.
2015;92:67–81. [142] Leou RC. An economic analysis model for the energy storage system applied
[131] Zhang Y, Yao F, Iu HH, Fernando T, Trinh H. Wind–thermal systems operation to a distribution substation. Int J Electr Power Energy Syst 2012;34(1):132–7.
optimization considering emission problem. Int J Electr Power Energy Syst [143] Tan WS, Hassan MY, Rahman HA, Abdullah MP, Hussin F. Multi-distributed
2015;65:238–45. generation planning using hybrid particle swarm optimisation-gravitational
[132] Oliva D, Cuevas E, Pajares G. Parameter identification of solar cells using ar- search algorithm including voltage rise issue. Gener Transm Distrib IET
tificial bee colony optimization. Energy 2014;72:93–102. 2013;7(9):929–42.
[133] Chamkalani A, Mae'soumi A, Sameni A. An intelligent approach for optimal [144] Baziar A, Kavousi-Fard A. Considering uncertainty in the optimal energy
prediction of gas deviation factor using particle swarm optimization and management of renewable micro-grids including storage devices. Renew
genetic algorithm. J Nat Gas Sci Eng 2013;14:132–43. Energy 2013;59:158–66.
[134] Askarzadeh A, Rezazadeh A. Artificial bee swarm optimization algorithm for [145] Zhao B, Zhang X, Chen J, Wang C, Guo L. Operation optimization of standa-
parameters identification of solar cell models. Appl Energy 2013;102:943–9. lone microgrids considering lifetime characteristics of battery energy storage
[135] Ghanbari A, Kazemi SMR, Mehmanpazir F, Nakhostin MM. A cooperative ant system. Sustain Energy IEEE Trans 2013;4(4):934–43.