Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
PII: S0960-1481(18)30064-8
DOI: 10.1016/j.renene.2018.01.058
Please cite this article as: Makbul A.M. Ramli, H.R.E.H. Bouchekara, Abdulsalam S. Alghamdi,
Optimal Sizing of PV/wind/diesel hybrid microgrid system using multi-objective self-adaptive
differential evolution algorithm, Renewable Energy (2018), doi: 10.1016/j.renene.2018.01.058
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to
our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form.
Please note that during the production process errors may be discovered which could affect the
content, and all legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT
9 Abstract
13 Microgrid systems, such as solar photovoltaic power (PV) and wind energy, integrated
14 with diesel generators are promising energy supplies and are economically feasible for
15 current and future use in relation to increased demands for energy and depletion of
16 conventional sources. It is thus important to optimize the size of hybrid microgrid system
17 (HMS) components, including storage, to determine system cost and reliability. In this
18 paper, optimal sizing of a PV/wind/diesel HMS with battery storage is conducted using
20 city of Yanbu, Saudi Arabia. Using the multi-objective optimization approach, the
22 computational time. One of the main criteria to consider when designing and optimizing
23 the HMS is the energy management strategy, which is required to coordinate the different
24 units comprising the HMS. The multi-objective optimization approach is then used to
25 analyze the Loss of Power Supply Probability (LPSP), the Cost of Electricity (COE), and
26 the Renewable Factor (RF) in relation to HMS cost and reliability and is tested using
27 three case studies involving differing house numbers. Results verify its application in
10 * Corresponding author. Department of Electrical and Computer Engineering, King Abdulaziz University, Jeddah
11 21589, Saudi Arabia. 1
12 E-mail addresses:mramli@kau.edu.sa (M.A.M. Ramli), bouchekara.houssem@gmail.com (H.R.E.H. Bouchekara)
ACCEPTED MANUSCRIPT
28 optimizing the HMS and in its practical implementation. In addition, optimization results
29 using the proposed approach provided a set of design solutions for the HMS, which will
31 Moreover, it is important to select optimally sized HMS components to ensure that all
32 load demands are met at the minimum energy cost and high reliability.
33 Keywords: hybrid system; renewable energy; wind energy; PV; optimization; differential
34 evolution algorithm.
35 1. Introduction
37 renewable energy technology, shortages of electricity remain a problem in rural areas and
38 on islands with abundant renewable energy sources. Microgrid systems (using solar PV
39 cells (PV) and wind energy) integrated with diesel generators are considered to be
40 promising in providing an energy supply, and are also economically feasible for current
41 and future usage. However, it is crucial to determine the system cost and reliability when
42 optimizing the size and storage capabilities of the components of a hybrid microgrid
43 system (HMS).
44 Wind and solar energy are popular renewable sources that are being utilized because
47 solar energy development. However, despite the tangible benefits, the main hindrance to
48 the public accepting wind and solar energy systems is the related high capital investment
49 [1]. Nevertheless, wind turbines and PV have considerable potential in meeting the load
2
ACCEPTED MANUSCRIPT
50 demand of power generation [2-5], and electricity generation from these sources has
51 gained global attention due to its various advantages, such as enabling conservation of
53 Wind turbines and PV are the top two most promising energy resources used in the
54 kingdom of Saudi Arabia and are prominently used in coastal areas. A certain amount of
55 research and feasibility studies have been conducted in relation to wind energy system
56 implementation in locations around the Kingdom [6-10], and research exploring wind and
57 solar energy potential in the Kingdom dates back to 1985 [11]. However, the use of solar
58 energy has been more popular in terms of actual system implementation, which is
59 possibly attributed to the highly abundant solar resources present throughout the
60 Kingdom, but is also related to the vast amount of research and development conducted
62 With the aim of utilizing diverse energy sources, a large number of technical studies
63 have been conducted on supplementing energy obtained from fossil fuel combustion with
64 solar and wind energy. For example, Dalton et al. conducted a feasibility study of a
66 feasible to use this for a grid-connected hotel and small–medium tourist facilities [12,13].
68 analyzed its use in future energy supplies, while Malik presented an analysis of the
69 potential and development of renewable energy resources in Brunei [15]. In other studies,
70 the performance of a PV/diesel hybrid system has been analyzed in Thailand [16,17].
71 Research results indicate that integrating renewable energy systems, such as wind and
72 PV, with diesel generators can reduce capital investment and energy costs and improve
3
ACCEPTED MANUSCRIPT
73 system reliability, particularly in developing countries. The authors’ previous work [18]
74 analyzed the economic feasibility of a PV/diesel system with flywheel energy storage and
75 concluded that by reducing the operation of diesel generators the hybrid system could
76 minimize fuel consumption, total Net Present Cost (NPC), Cost of Electricity (COE), and
77 CO2 emissions.
78 Solar and wind energies are discontinuous, highly variable resources that are site-
81 demand [19–21]. However, the combined operation of solar and wind energy systems
82 helps to smooth this fluctuation and provides a more reliable energy supply compared to
83 the separate use of solar and energy systems. [22–25]. In addition, storage and diesel
84 generator systems can also be used to cope with the intermittency of solar and wind
85 energy supply [26–28], where the storage system relieves the mismatch between
87 In general, diesel generators with a reliable power supply capacity are operated as the
88 main source of electricity and maintain a generation and consumption power balance. For
89 the HMS, it is imperative that energy sources are optimally operated in combination with
91 Over the years, several software applications have been used to optimize microgrid
92 systems, with or without the use of storage and diesel generator systems such as HOMER
94 associated with these products have called for the development of optimized
4
ACCEPTED MANUSCRIPT
96 have been intensively developed to improve microgrid systems from a technical and
97 economical point of view, and an increase in hybrid system efficiency has been obtained
99 The genetic algorithm (GA) has also been proposed to optimize hybrid systems that
100 have a large number of parameters, and its performance has been evaluated [32,33]. The
101 GA is effective in optimizing hybrid systems that have various components with differing
102 sizes so that they meet load demands, and is also effective in evaluating hybrid systems
103 based on the defined fitness function. However, the drawback of this method is that it has
105 Particle swarm optimization (PSO) methods have also been employed to optimize
106 various problems. PSO is easy in coding, is robust to control parameters, and gives high-
107 quality solutions with stable convergence. This method makes faster computations than
108 the GA, and is simple to use for a broad variety of problems [34,35]. In a recent study, a
109 multi-objective particle swarm optimization (MOPSO) was applied to optimize an HMS
110 consisting of solar and wind energies and a diesel generator storage system to meet load
111 demand in Nahavand, Rafsanjan, and Khash [36]. The results of the study concluded the
112 need for appropriate sizing of the components to optimize the model, and both COE and
113 LPSP were considered when determining optimal sizing of the hybrid system.
114 The main objective of this study is to exploit the capabilities of the MOSaDE
115 algorithm for optimizing a PV/wind/diesel HMS with battery storage. For this purpose,
116 the mutation parameter is adaptively adjusted to improve the Differential Evolution (DE)
5
ACCEPTED MANUSCRIPT
119 computational time. Optimal sizing of the HMS components are obtained by considering
120 the COE and LPSP when analyzing cost and reliability of the HMS, and the proposed
121 approach is used to achieve the goal of optimal economic system operation. It is
122 considered that the results of this study will be useful in a number of ways, such as when
123 investigating optimal HMS component scheduling, as a power reference for the economic
124 operation of PV and wind turbine generators, and when ensuring power supply reliability.
125 It will also be useful for regulating diesel generation in a normal range, and to determine
126 the operational set points of PV and wind turbine generators and the energy storage
127 system.
128 The following section describes the HMS used in this study. Section 3 presents the
129 optimization case study, Section 4 discusses the developed MOSaDE algorithm and
130 optimization process, and Section 5 analyzes the optimization results and the effect of
131 design variables on COE. Finally, Section 6 concludes results of research results and
132 findings.
133
135 The HMS evaluated in this study is composed of five major system components: a PV
136 system, wind turbines, diesel generators, an inverter, and a battery bank. The
6
ACCEPTED MANUSCRIPT
Diesel Generator
Wind PV
Loads Converter
Battery bank
AC Bus DC Bus
139
141 Power output of PV panels is calculated using the following equation [37],
𝐺
142 𝑃𝑃𝑉 ‒ 𝑜𝑢𝑡 = 𝑃𝑁 ‒ 𝑃𝑉 × 𝐺
𝑟𝑒𝑓
[1 + 𝐾𝑡((𝑇𝑎𝑚𝑏 + (0.0256 × 𝐺)) ‒ 𝑇𝑟𝑒𝑓)] (1)
143 where PN-PV is the PV rated power at standard test conditions (STC); G is solar radiation
144 (W/m2); Gref is 1 kW/m2; Kt is a constant, 3.7 103 (1/˚C); Tamb is ambient temperature;
146 Hourly solar radiation data for the city used in this study and for this calculation is
7
ACCEPTED MANUSCRIPT
149 As wind speed changes considerably with height, conversion from wind speed
150 measured at anemometer height to a desired hub height can be made according to the
𝑣2 ℎ2 𝛼
152 𝑣1
= () ℎ1
(2)
153 where v1 is the speed at a reference height, h1; v2 is the speed at a hub height, h2; and is
154 the coefficient of friction. is defined by certain parameters, i.e. roughness of terrain,
155 wind speed, temperature, height above ground, hour of the day, and time of year [40–
156 42]. In technical literature, it is commonly defined by different types of terrain [43– 45];
157 however, the recommendations of IEC standards [46,47] give a friction coefficient value
158 of 0.11 for extreme wind conditions, and 0.20 for normal wind conditions.
{
0 𝑉 < 𝑉𝑐𝑢𝑡 ‒ 𝑖𝑛, 𝑉 > 𝑉𝑐𝑢𝑡 ‒ 𝑜𝑢𝑡
(
3
160 𝑉
(
3
3
𝑃𝑟𝑎𝑡𝑒𝑑
3
𝑉𝑟𝑎𝑡𝑒𝑑 ‒ 𝑉𝑐𝑢𝑡 ‒ 𝑖𝑛 ) ‒ 𝑃𝑟𝑎𝑡𝑒𝑑
𝑉𝑐𝑢𝑡 ‒ 𝑖𝑛
3 3
𝑉𝑟𝑎𝑡𝑒𝑑 ‒ 𝑉𝑐𝑢𝑡 ‒ 𝑖𝑛 ) 𝑉𝑐𝑢𝑡 ‒ 𝑖𝑛 ≤ 𝑉 < 𝑉𝑟𝑎𝑡𝑒𝑑
161 (3)
162 where V is current time step wind speed, Vrated is the nominal wind speed, Vcut-out is the
163 cut out wind speed, Vcut-in is the cut in wind speed, and Prated is the rated power.
164 Details of the wind turbine used in the HMS optimization process are shown in Table
165 1.
8
ACCEPTED MANUSCRIPT
167 The diesel generator is the secondary power source, and thus plays a significant role
168 in maintaining stable system operation. Therefore, maintaining diesel generation within a
169 safe range is of considerable importance. The generator works at a low efficiency with a
170 low loading rate [49]. Therefore, to proficiently utilize energy and realize an adequate
171 safety margin for power fluctuations, such as a sudden increase in load consumption,
172 diesel generation needs to be operated within a normal operating range in order to avoid
174 In the design of a hybrid system, the fuel consumption of a diesel generator, q(t), can
177 where P(t) is the generated power, a and b are coefficients of the fuel consumption
178 parameters, and Prated is the rated power. In this study, a and b are approximated as 0.246
180 The overall diesel generator efficiency can be calculated using the following equation
181 [54],
183 where overall is the overall efficiency and brake thermal is the brake thermal efficiency.
185
187 The efficiency of the inverter is calculated using the following equation,
9
ACCEPTED MANUSCRIPT
𝑃
188 𝜂𝑖𝑛𝑣 = 2 (6)
𝑃 + 𝑃0 + 𝑘𝑃
2
191 𝑃0 = 1 ‒ 99 ( 10
‒
1
𝜂10 𝜂100
‒ )
9 (8)
1
192 𝑘=𝜂 ‒ 𝑃0 (9)
100
193 in which 10 and 100 represent the inverter efficiency at 10% and 100% of nominal
196
198 The battery capacity is calculated based on demand and autonomy days as follows,
𝐸𝐿 𝐴𝐷
199 𝐶𝐵 = 𝐷𝑂𝐷 𝜂 (10)
𝑖𝑛𝑣 𝜂𝑏
200 where EL is load, AD is autonomy days, DOD is depth of discharge (80%), inv is inverter
203 As previously mentioned, the HMS optimized in this work consists of a photovoltaic
204 power plant, wind energy system, battery energy storage system, and a diesel generator,
10
ACCEPTED MANUSCRIPT
205 all of which are local generation units located near the area of electrical consumption;
206 therefore, there are no electrical losses due to distribution considered here.
207 One of the main criteria to consider when designing and optimizing the HMS is the
208 energy management strategy, which is required to coordinate the different units
209 comprising the HMS. The energy management strategy used in this study is given in Fig.
210 2 [36]. Renewable resources provide the main source of energy, while the diesel
211 generator plays the role of a dispatchable source and the battery bank is charged when
212 generated power is greater than the load. In addition, if renewable resources generate
213 more power than the load and battery bank can use, the excess power is sent to a dump
214 load, whereas if renewable resources generate less power than required by the load then
215 the power deficit is provided by the battery bank. However, if both renewable resources
216 and battery bank are unable to supply the load then the diesel generator operates to
11
ACCEPTED MANUSCRIPT
Start
Read data
PI(t)
If Pw(t) + PPV ‒ out(t) ≥ ηinv
No Yes
PI(t)
Pdch(t) = ∙ (Pw(t) + PPV ‒ out(t)) If Pw(t) + PPV ‒ out(t) > PI(t)
ηinv
Edch(t) = Pdch(t) ∙ 1hr (iteration time) No
Eb(t) = Eb(t ‒ 1)
No
If Ech(t) ≤ Ebmax ‒ Eb(t)
( ( ))
PI(t)
Eb(t) = Eb(t ‒ 1) + Pg ∙ ηinv + Pw(t) ‒ ∙1
ηinv
Yes No
diesel(t) = Pg ∙ ηinv
Eb(t) = Eb(t ‒ 1) + Ech Eb(t) = Ebmax
No
return
Yes
Eb(t) = Ebmax
Edump(t) = Ech(t) ∙ (Ebmax ‒ Eb(t))
return
218
219 Fig. 2. Flowchart of HMS energy management strategy.
12
ACCEPTED MANUSCRIPT
220
222 Saudi Arabia has a hot dry climate and abundant renewable energy resources,
223 particularly those of solar and wind. In this study, optimization of a hybrid
225 industrial city in the western region of the Kingdom (latitude of 24o0520″ N and
226 longitude 38o0349″ E) that is situated on the Red Sea coastal area and has both good
227 levels of solar irradiation and great wind potential [7,59]; an average annual wind speed
228 of 3.53 m/s and solar irradiation of 5.95 kWh/m2/day are considered in this study [60]
229 (Fig. 3 shows the hourly wind speed, and Fig. 4 shows hourly solar irradiation in the
230 city). The temperature of the city is in the range of 15–40 o C, while the average annual
231 temperature is about 29 o C. Real weather data obtained in Yanbu is used in optimization
232 process.
16
12
Wind Speed (m/s)
233 0
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
13
ACCEPTED MANUSCRIPT
1.2
1.0
0.6
0.4
0.2
235 0.0
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
237
238 4. Optimization
240 In this paper, the HMS is optimized based on a multi-objective problem that can
𝑇
Minimize F(𝐱) = [f1(𝐱),f2(𝐱),…,f𝑘(𝐱)]
242 Subject to 𝑔(𝐱) = 0 (11)
and h(𝐱) ≤ 0
243 where x is the vector of design variables, fi(𝐱) is the ith objective function, F(x) is the
244 vector of objective functions (k is the number of objective functions), g(𝐱) is the set of
247 In this work, COE and the LPSP parameters are selected as objective functions,
248 where the aim is to optimize an HMS that can guarantee a reliable energy supply at a low
249 cost.
14
ACCEPTED MANUSCRIPT
251 In general energy development and in analysis of the COE of hybrid systems in
252 particular, the total net present cost (NPC) is a critical factor and consists of capital,
253 operation and maintenance (O&M), and replacement costs. Hybrid renewable energy
254 systems have no fuel expense and therefore the O&M costs are very low. However, the
255 investment or capital costs are very high. In this study, the costs involved in operating the
256 HMS are costs for PV array, wind turbine, diesel generator, inverter, and battery
257 components.
258 The COE is the average cost of useful electrical energy generated by the system, in
261 where Pload is the hourly amount of consumed power and CRF is a ratio given by
𝑛
𝑖(1 + 1)
262 𝐶𝑅𝐹 = 𝑛 × 𝐶𝑅𝐹 (13)
(1 + 1) ‒ 1
265 As a statistical parameter, the loss of power supply probability (LPSP) is defined as
266 the probability of power supply failure to meet load demand, and occurs because of a
267 technical failure or a low renewable resource input. LPSP can be calculated using the
270 In addition, in the analysis of system reliability, the following condition is applied,
15
ACCEPTED MANUSCRIPT
274 The RF is defined to identify the quantity of power generated from the diesel
275 generator compared to the amount generated from renewable resources, and is expressed
276 as follows,
(
∑𝑃
100 × 1 ‒ ∑𝑃
diesel
) = (%)RF
renewable resources
(16)
279 The design variables considered in this work are the nominal power of PV (PV), the
280 number of autonomy days (AD), and the number of wind turbines (WT).
282 The differential evolution (DE) algorithm is one of the most commonly used and
283 promising evolutionary algorithms for solving real world optimization problems. It was
284 initially developed by Price and Storn in 1995 while trying to solve the Chebyshev
285 polynomial fitting problem [65]. The DE starts with an initial population randomly
286 generated, which then evolves until termination conditions are fulfilled. While the
289 In a standard version of a DE, it is necessary to pre-specify one certain trial vector
290 generation strategy that has its parameters tuned using a time-consuming trial and error
291 scheme. However, a new version of DE called the Self-adaptive Differential Evolution
16
ACCEPTED MANUSCRIPT
292 (SaDE) algorithm has been developed [68], in which both trial vector generation
293 strategies and their associated parameters can be automatically adapted according to their
295 The SaDE is widely used in different research fields and for many applications, and
296 this motivated Huang et al. [69] to extend this algorithm to solve multi-objective
297 optimization problems. The developed algorithm is known as the Multi-Objective Self-
298 Adaptive Differential Evolution algorithm (MOSaDE), which can be described using the
300 Step 1: Randomly generate a population of NP individuals in the search space. Initialize
301 certain parameters such as the probability of strategy pk where k=1K, (K is the number
302 of available strategies), and the median value of the Crossover Probability (CR ∈ [0,1])
303 that is (CRmk) for each strategy and the learning period (LP=50). It is of note here that
304 there are two implemented strategies in the MOSaDE algorithm, which are as follows:
𝑥𝑖,𝑗 + 𝐹 ∙ (𝑥𝑟 ,𝑗 ‒ 𝑥𝑟 )
𝑢𝑖,𝑗 = { 1
𝑥𝑖,𝑗
2,𝑗
𝑖𝑓 rand[0,1] < 𝐶𝑅 𝑜𝑟 𝑗 = 𝑗𝑟𝑎𝑛𝑑
otherwise (17)
307 where F is the differential weight (usually ∈ [0,2]) and r1, r2, r3, r4, r5, and r6 are mutually
308 exclusive integers randomly generated within the range [1, NP].
309 Step 2: Evaluate the objective functions of each individual, and then fill the external
17
ACCEPTED MANUSCRIPT
312 (1) Calculate the strategy probability, pk, which is the percentage of the success rate
313 of trial vectors generated by each strategy during the learning period.
314 (2) Assign trial vector generation strategy and parameters to each target vector, 𝐱𝑖:
315 a. Use Stochastic Universal Sampling (SUS) for selecting one strategy, k, for
317 b. Assign control parameters, F and CR, where F is generated using a normal
318 distribution with a linearly reducing mean value from 1.0 to 0.05 and a
FOR i=1:NP
𝑘
Evaluate the trial vector, 𝐮 𝑖 , and compare it with the target
𝑘
vector, 𝐱𝑛𝑖, nearest to 𝐮 𝑖 in the solution space.
𝑘
IF 𝐱𝑛𝑖 dominates 𝐮 𝑖
𝑘
discard 𝐮 𝑖 ,
ELSE
𝑘
IF 𝐮 𝑖 dominates 𝐱𝑛𝑖
𝑘
replace 𝐱𝑛𝑖 with 𝐮 𝑖 .
However,
𝑘
IF neither 𝐮 𝑖 or 𝐱𝑛𝑖 dominate each
18
ACCEPTED MANUSCRIPT
other then
select the less crowded vector
as the new target vector.
𝑘
𝐮 𝑖 will enter the external archive if:
𝑘
I. 𝐮 𝑖 dominates an individual(s) in the
individuals
END IN THE
FOLLOWING
CIRCUMSTANCES:
k
If there is a case where 𝐮 i is better than 𝐱ni. Record the
326
328 As previously mentioned, the city of Yanbu is used as a case study for optimization of
329 the HMS. The hourly load profile of a single house over one year in Yanbu city is shown
330 in Fig. 5; this profile was generated by conducting a building energy simulation in eQuest
331 [70]. The economical parameters of the HMS are given in Table 1.
19
ACCEPTED MANUSCRIPT
334
335 Table 1
336 Economical parameters of HMS [36].
337
Component Parameter Value Unit
PV PV regulator cost 1500 $
Rated power 7.3 kWh
Initial cost 3400 $/kW
Life time 24 year
PV regulator efficiency 95 %
20
ACCEPTED MANUSCRIPT
339 Three cases are considered: CASE 1, CASE 2, and CASE 3 correspond to 5, 10, and
340 15 houses, respectively. The MOSaDE is run for each case using a population of 100
341 individuals and 200 iterations, and the lower and upper bounds for PV, AD and WT are
343 The Pareto fronts obtained for CASE 1, CASE 2, and CASE 3 are shown in Fig. 6,
344 Fig. 7, and Fig. 8, respectively. In addition, 30 selected solutions for every case along
345 with the detailed results are tabulated in Table 2, Table 3, and Table 4.
346 From these results, it is evident that very interesting sets of solutions are determined
347 for each case when the MOSaDE is applied to HMS optimization. A system designer
348 would therefore be able to select a solution based on his/her experience and the
349 associated specifications. For example, solution # 1 of CASE 1 has a power of 18.2 kW
350 for PV panels, 5 autonomy days, and 10 wind turbines. This solution corresponds to a
21
ACCEPTED MANUSCRIPT
351 COE of 0.083 $/kWh, an LPSP of 0.219, and a RF of 95.51%. It is of note that the COE
352 has a higher value and the LPSP a lower value than the other available solutions. The
353 energy contribution from each part (i.e. PV panels, wind turbines, diesel generator, and
356 compared to solution # 15 and solution # 30. The diesel generator and battery supplied 4
357 % and 27 % of energy, which are lower than energy contribution of diesel generator and
358 battery in solution # 15 and solution # 30. Wind turbines contribute 10 % of energy, the
359 same percentage for all solutions. For solution # 15 of CASE 1, we can see that the COE
360 is 0.081 $/kWh, LPSP is 0.228, and RF is 95.12 %. The energy contribution for each part
362 Fig. 9b. PV panels supplied 58 % of energy while amount of energy generated by
363 diesel generator and battery are 5% and 28%, respectively. If we select solution # 30 of
365 Fig. 9c, we can see that the COE is 0.081 $/kWh, LPSP is 0.280 %, and RF is 93.56
366 %, which corresponds to a lower value of COE and a higher LPSP than the other
367 solutions. In solution # 30, PV panels contribute 56% energy, the lowest percentage of
368 energy contribution in all available solutions. Contribution of energy provided by diesel
369 generator and battery are 6% and 28%, respectively, which are higher as compared to
371 The same analysis can also be made for the results of CASE 2 and CASE 3, when the
22
ACCEPTED MANUSCRIPT
373 In a previous study [36], the multi-objective optimization of HMS was only treated as
374 a single objective optimization problem. In other words, the original problem was
375 converted to a single objective optimization problem where different objective functions
376 were aggregated into one objective function. This approach had the disadvantage of
377 determining only one optimal solution that did not express the trade-off between different
378 objective functions. In order to offer the designer multiple choices, several runs using
379 different weights and constraints were required, which involved a greater amount of time.
380 However, the proposed approach in this paper is a true multi-objective one, in which the
381 objectives are treated simultaneously and independently. Moreover, this approach offers
382 the designer a set of optimal solutions known as the Pareto front.
0.0835
0.083
0.0825
COE ($/kWh)
0.082
0.0815
Solution # 15
0.081
0.0805
0.2 0.22 0.24 0.26 0.28 0.3
383 LPSP
385
386 Table 2
387 Selected solutions from Pareto front for CASE 1.
388
PV AD WT COE LPSP RF PV WIND BATTERY DIESEL
23
ACCEPTED MANUSCRIPT
Solution # 1 18.2 5.0 10 0.083 0.219 95.51 82356 13534 36771 5951
Solution # 2 18.0 5.0 10 0.083 0.220 95.49 81698 13534 36771 5954
Solution # 3 17.9 5.0 10 0.083 0.220 95.47 81307 13534 36771 5958
Solution # 4 17.9 5.0 10 0.083 0.220 95.47 81257 13534 36771 5958
Solution # 5 17.8 5.0 10 0.082 0.220 95.45 80849 13534 36771 5962
Solution # 6 17.6 5.0 10 0.082 0.221 95.39 79608 13534 36771 5995
Solution # 7 17.5 5.0 10 0.082 0.221 95.38 79396 13534 36771 5998
Solution # 8 17.4 5.0 10 0.082 0.222 95.34 78828 13534 36771 6017
Solution # 9 17.4 5.0 10 0.082 0.222 95.34 78761 13534 36771 6017
Solution # 10 17.3 5.0 10 0.082 0.223 95.31 78390 13534 36771 6039
Solution # 11 17.3 5.0 10 0.082 0.223 95.29 78177 13534 36771 6046
Solution # 12 17.2 5.0 10 0.082 0.223 95.29 78127 13534 36771 6054
Solution # 13 17.2 4.9 10 0.081 0.227 95.15 77750 13534 36771 6204
Solution # 14 17.0 5.0 10 0.081 0.227 95.14 76890 13534 36771 6186
Solution # 15 16.9 5.0 10 0.081 0.228 95.12 76723 13534 36771 6204
Solution # 16 16.9 5.0 10 0.081 0.229 95.07 76390 13534 36771 6249
Solution # 17 16.8 5.0 10 0.081 0.230 95.04 76199 13534 36771 6271
Solution # 18 16.8 5.0 10 0.081 0.230 95.04 76127 13534 36771 6271
Solution # 19 16.8 5.0 10 0.081 0.230 95.03 76070 13534 36771 6278
Solution # 20 16.7 5.0 10 0.081 0.231 95.01 75808 13534 36771 6289
Solution # 21 16.7 5.0 10 0.081 0.231 94.99 75632 13534 36771 6315
Solution # 22 16.7 4.9 10 0.081 0.233 94.93 75808 13534 36771 6392
Solution # 23 16.7 4.9 10 0.081 0.233 94.93 75808 13534 36771 6392
Solution # 24 16.7 4.9 10 0.081 0.234 94.93 75808 13534 36771 6392
Solution # 25 16.7 4.9 10 0.081 0.235 94.91 75808 13534 36771 6425
Solution # 26 16.7 4.8 10 0.081 0.238 94.82 75808 13534 36771 6528
Solution # 27 16.2 4.6 10 0.081 0.260 94.11 73535 13534 36772 7297
Solution # 28 16.1 4.2 10 0.081 0.277 93.61 73117 13534 36772 7883
Solution # 29 16.1 4.2 10 0.081 0.278 93.59 73151 13534 36772 7916
Solution # 30 16.1 4.2 10 0.081 0.280 93.56 73151 13534 36772 7949
389
24
ACCEPTED MANUSCRIPT
0.075
0.07
COE ($/kWh)
0.065
0.06
0.055
0.25 0.3 0.35 0.4 0.45 0.5
390 LPSP
392
393 Table 3
394 Selected solutions from Pareto front for CASE 2.
395
PV AD WT COE LPSP RF PV WIND BATTERY DIESEL
Solution # 1 36.7 5.0 10 0.072 0.262 97.30 167630 13534 79511 7040
Solution # 2 36.1 5.0 10 0.072 0.266 97.20 163099 13534 79511 7183
Solution # 3 36.0 5.0 10 0.072 0.266 97.20 163099 13534 79511 7183
Solution # 4 36.0 5.0 10 0.072 0.266 97.20 163099 13534 79511 7183
Solution # 5 36.0 5.0 10 0.072 0.266 97.20 163099 13534 79511 7183
Solution # 6 35.6 5.0 10 0.072 0.269 97.20 163099 13534 79511 7183
Solution # 7 35.6 5.0 10 0.072 0.269 97.20 163099 13534 79511 7183
Solution # 8 32.8 4.5 10 0.070 0.317 96.12 149508 13534 79512 9413
Solution # 9 32.9 4.3 10 0.069 0.321 96.12 149508 13534 79512 9413
Solution # 10 31.2 3.5 10 0.066 0.372 95.71 140446 13534 79521 10024
Solution # 11 30.9 3.5 10 0.066 0.375 95.18 140446 13534 79521 11253
Solution # 12 30.9 3.3 10 0.065 0.381 95.18 140446 13534 79521 11253
Solution # 13 31.1 3.1 10 0.064 0.388 95.18 140446 13534 79521 11253
Solution # 14 30.9 3.1 10 0.064 0.391 95.18 140446 13534 79521 11253
Solution # 15 30.9 2.9 10 0.064 0.397 95.18 140446 13534 79521 11253
Solution # 16 30.9 2.7 9 0.063 0.408 95.13 140446 12181 80120 11338
Solution # 17 29.7 2.6 10 0.062 0.415 94.94 135916 13534 79640 11603
Solution # 18 28.2 2.5 10 0.062 0.424 94.43 126855 13534 80394 12302
Solution # 19 28.2 2.4 10 0.062 0.430 94.02 126855 13534 80394 13204
Solution # 20 28.4 2.4 9 0.062 0.431 93.92 126855 12181 81195 13384
25
ACCEPTED MANUSCRIPT
Solution # 21 27.9 2.3 10 0.061 0.436 94.02 126855 13534 80394 13204
Solution # 22 28.7 2.1 9 0.061 0.441 94.22 131385 12181 80573 12965
Solution # 23 28.2 2.0 10 0.060 0.443 94.02 126855 13534 80394 13204
Solution # 24 28.1 1.9 10 0.060 0.446 94.02 126855 13534 80394 13204
Solution # 25 27.5 1.7 10 0.059 0.453 93.64 122324 13534 81260 13818
Solution # 26 27.2 1.7 10 0.059 0.457 93.64 122324 13534 81260 13818
Solution # 27 26.9 1.7 10 0.059 0.457 93.64 122324 13534 81260 13818
Solution # 28 27.1 1.2 10 0.057 0.475 93.22 122324 13534 81260 14716
Solution # 29 26.8 1.2 10 0.057 0.475 93.22 122324 13534 81260 14716
Solution # 30 28.2 1.0 9 0.057 0.475 93.46 126855 12181 81195 14392
396
0.07
0.065
COE ($/kWh)
0.06
0.055
0.05
0.045
0.36 0.38 0.4 0.42 0.44 0.46 0.48
397 LPSP
399
400 Table 4
401 Selected solutions from Pareto front for CASE 3.
402
PV AD WT COE LPSP RF PV WIND BATTERY DIESEL
Solution # 1 44.9 4.7 10 0.064 0.372 96.90 203874 13534 122741 10547
Solution # 2 44.9 4.6 10 0.064 0.373 96.90 203874 13534 122741 10547
Solution # 3 44.9 4.2 10 0.062 0.381 96.77 203874 13534 122741 10992
Solution # 4 44.9 3.9 10 0.061 0.386 96.77 203874 13534 122741 10992
Solution # 5 44.8 3.5 10 0.059 0.395 96.51 203874 13534 122741 11883
Solution # 6 44.9 3.4 10 0.059 0.396 96.51 203874 13534 122741 11883
Solution # 7 44.9 3.3 10 0.058 0.397 96.51 203874 13534 122741 11883
Solution # 8 44.9 3.3 10 0.058 0.398 96.51 203874 13534 122741 11883
26
ACCEPTED MANUSCRIPT
Solution # 9 44.4 3.3 9 0.058 0.401 96.38 199343 12181 123889 12133
Solution # 10 44.6 3.1 10 0.057 0.410 96.51 203874 13534 122741 11883
Solution # 11 44.1 2.7 10 0.056 0.419 96.40 199343 13534 123137 12081
Solution # 12 44.4 2.6 10 0.056 0.420 96.40 199343 13534 123137 12081
Solution # 13 43.9 2.6 9 0.055 0.422 96.38 199343 12181 123889 12133
Solution # 14 43.6 2.6 10 0.055 0.423 96.40 199343 13534 123137 12081
Solution # 15 43.5 2.6 10 0.055 0.424 96.40 199343 13534 123137 12081
Solution # 16 43.2 2.5 10 0.055 0.429 96.00 194813 13534 123823 13285
Solution # 17 43.4 2.2 9 0.053 0.440 95.96 194813 12181 124637 13388
Solution # 18 43.2 2.2 10 0.053 0.440 96.00 194813 13534 123823 13285
Solution # 19 43.5 2.1 10 0.053 0.441 96.12 199343 13534 123137 13024
Solution # 20 42.3 2.1 10 0.053 0.444 95.85 190282 13534 124675 13645
Solution # 21 43.2 2.0 10 0.053 0.446 96.00 194813 13534 123823 13285
Solution # 22 42.4 1.9 9 0.052 0.449 95.80 190282 12181 125509 13771
Solution # 23 41.3 1.8 9 0.051 0.454 95.62 185752 12181 126474 14205
Solution # 24 40.4 1.7 10 0.051 0.458 95.41 181221 13534 126660 14768
Solution # 25 40.4 1.7 10 0.051 0.459 95.41 181221 13534 126660 14768
Solution # 26 40.3 1.7 10 0.051 0.459 95.41 181221 13534 126660 14768
Solution # 27 40.4 1.6 10 0.051 0.463 95.41 181221 13534 126660 14768
Solution # 28 40.6 1.6 9 0.050 0.465 95.62 185752 12181 126474 14205
Solution # 29 40.4 1.6 10 0.050 0.466 95.41 181221 13534 126660 14768
Solution # 30 40.4 1.4 10 0.050 0.473 95.18 181221 13534 126660 15482
403
BATTERY BATTERY
27%
28% BATTERY
28%
PV
PV
PV 56%
59% 58%
10% WIND
WIND
WIND 10%
10%
405 Fig. 9. Contribution of energy supplied by PV panels, wind turbines, diesel generator,
406 and battery for certain selected solutions for CASE 1.
408 A sensitivity analysis was conducted to quantify the effect of design variables on the
409 COE. In this study, each variable was varied between 50% and 150% of each of the
27
ACCEPTED MANUSCRIPT
410 optimal values; the curves obtained for solution 15 of CASE 1 are shown in Fig. 10. It
411 can be seen from these curves that the COE is relatively stable when the number of wind
412 turbines is varied. However, if the value of PV varies around the optimal point, the COE
413 increases. In addition, the COE decreases in line with a decrease in the number of
414 autonomy days, and when the number of days increases so does the COE.
0.13
PV
AD
0.12 WT
0.11
COE ($/kWh)
0.1
0.09
0.08
0.07
0.5 1 1.5
415 Parameter variation
417
418 6. Conclusion
419 In this paper, the MOSaDE algorithm is used to optimize an HMS composed of solar
420 PV, wind turbines, battery bank, and diesel generator. Use of a multi-objective
421 optimization approach ensures the objectives are treated simultaneously and
423 functions treated in this paper are the COE and LPSP, and three case studies are
424 performed, each involving a particular number of houses in the city of Yanbu. The results
28
ACCEPTED MANUSCRIPT
425 obtained include not only one optimal solution, but a set of optimal solutions (Pareto
426 front), thereby offering a designer several options. Optimization of the HMS is performed
427 so that the optimal capacities of a wind turbine, a number of PV modules, and the
428 capacity of a diesel generator and battery can be selected. In addition, a sensitivity
429 analysis is conducted to test the effect of variation of optimal design variables. It is
430 important to select optimally sized HMS components to ensure that all load demands are
432 It is considered that the results of this study are useful in investigating optimal
433 scheduling of HMS components, and can be used as a power reference for the economic
434 operation of PV and wind turbine generators. In addition, the results are also useful for
435 ensuring a reliable power supply, regulating diesel generation within a normal range,
436 operating the set points of PV and wind turbine generators, and providing an optimal
437 energy storage system, thereby contributing to the realization of microgrid projects in the
438 country. The aim of all these efforts is to reduce the price of renewable energy resources,
439 especially solar PV and wind energy, and to make products more competitive for use in
440 the energy industry and market, with the ultimate aim of accelerating renewable energy
442 Acknowledgment
443 This project was funded by the Deanship of Scientific Research (DSR) at King
444 Abdulaziz University, Jeddah, under grant no. (G/664/135/1437). The authors, therefore,
445 acknowledge with thanks DSR for technical and financial support.
446
29
ACCEPTED MANUSCRIPT
447 References
448 [1] Ellabban O, Abu-Rub H, Blaabjerg F. Renewable energy resources: current status,
449 future prospects and their enabling technology. Renew Sustain Energy Rev 2014;
450 39:748–64.
451 [2] Bayod-Rújula A A. Future development of the electricity systems with distributed
456 [4] Hatziargyriou N, Asano H, Iravani R. Microgrids. Power and Energy Magazine
457 2007;5(4):78-94.
461 [6] Rehman S. Wind energy resources assessment for Yanbo, Saudi Arabia. Energy
463 [7] Rehman S, Ahmad A. Assessment of wind energy potential for coastal locations of
465 [8] Al-Abbadi NM. Wind energy resource assessment for five locations in Saudi
467 [9] Rehman S, El-amin IM, Ahmad F, Shaahid SM, Al-Shehri AM, Bakhashwain JM.
468 Wind power resource assessment for Rafha, Saudi Arabia. Renew Sustain Energy
30
ACCEPTED MANUSCRIPT
470 [10] Rehman S, Sahin AZ. Wind power utilization for water pumping using small wind
471 turbines in Saudi Arabia: A techno-economical review. Renew Sustain Energy Rev
472 2012;16(7):4470–8.
473 [11] Hepbasli A, Alsuhaibani Z. A key review on present status and future directions of
474 solar energy studies and applications in Saudi Arabia. Renew Sustain Energy Rev
475 2011;15(9):5021–50.
476 [12] Dalton GJ, Lockington DA, Baldock TE. Case study feasibility analysis of
479 [13] Dalton GJ, Lockington DA, Baldock TE. Feasibility analysis of renewable energy
480 supply options for a grid-connected large hotel. Renew Energy 2009;34(4): 955–
481 64.
482 [14] Himri Y, Arif S, Stambouli AB, Himri S, Draoui B. Review and use of the
483 Algerian renewable energy for sustainable development. Renew Sustain Energy
485 [15] Malik AQ. Assessment of the potential of renewable for Brunei Darussalam.
491 [18] Ramli MAM, Hiendro A, Twaha S. Economic analysis of PV/diesel hybrid system
31
ACCEPTED MANUSCRIPT
495 [20] Li Y, Yue D, Liu H, Liu Y. Wind-solar complementary power inverter based on
496 intelligent control. In: 4th IEEE Conference on Industrial Electronics and
498 [21] Takle ES, Shaw RH. Complimentary nature of wind and solar energy at a
500 [22] Himri Y, Stambouli AB, Draoui B, Himri S. Techno-economical study of hybrid
502 [23] Bekele G, Palm B. Feasibility study for a standalone solar-wind-based hybrid
504 [24] Dali M, Belhadj J, Roboam X. Hybrid solar-wind system with battery storage
505 operating in grid-connected and standalone mode: control and energy management
507 [25] Abdullah MO, Yung VC, Anyi M, Othman AK, Hamid KBA, Tarawe J. Review
508 and comparison study of hybrid diesel/solar/hydro/fuel cell energy schemes for a
510 [26] Strunz K, Kristina Brock E. Stochastic energy source access management:
514 renewable energy system based on energy storage as hydrogen. IEEE Trans
32
ACCEPTED MANUSCRIPT
518 2008;23(3):957–67.
519 [29] Chen H, Cong TN, Yang W, Tan C, Li Y, Ding Y. Progress in electrical energy
521 [30] HOMER energy modeling software. National Renewable energy laboratory.
522 http://en.openei.org/wiki/HOMER.
523 [31] Erdinc O, Uzunoglu M. Optimum design of hybrid renewable energy systems:
525 25.
527 for stand-alone renewable energy systems with hydrogen storage. Renew Energy
528 2007;32(7):1102–26.
529 [33] Mohammadi M, Hosseinian SH, Gharehpetian GB. GA-based optimal sizing of
530 microgrid and DG units under pool and hybrid electricity markets. Int J Elect
535 [35] Schutte JF. The particle swarm optimization algorithm – structural optimization
33
ACCEPTED MANUSCRIPT
539 306.
540 [37] Daud AK, Ismail MS. Design of isolated hybrid systems minimizing costs and
542 [38] Elliott DL, Holladay CG, Barchet WR, Foote HP, Sandusky WF. Wind energy
543 resource atlas of the United States. Taylor & Francis Group, LLC; 2006. p. 44.
544 [39] Justus CG. Wind energy statistics for large arrays of wind turbines (New England
546 [40] Rehman S, Al-Abbadi NM. Wind shear coefficients and energy yield for Dhahran,
548 [41] Jaramillo OA, Borja MA. Wind speed analysis in La Ventosa, Mexico: a bimodal
550 [42] Farrugia RN. The wind shear exponent in a Mediterranean island climate. Renew
553 validation of the methodology used in the extrapolation of wind speed data at
555 [44] Bechrakis DA, Sparis PD. Simulation of the wind speed at different heights using
557 [45] Patel MR. Wind and solar power systems. CRC Press. 1999. p. 50–54.
558 [46] International Electrotechnical Committee IEC. 61400–1, Wind Turbines Part 1:
34
ACCEPTED MANUSCRIPT
560 [47] International Electrotechnical Committee IEC. 61400–3, Wind Turbines Part 3:
561 Design Requirements for Offshore Wind Turbines, 1st ed., IEC, Geneva,
564 hybrid power system with multiple renewable sources of energy. Proceeding
566 [49] Korpaas M, Holen A T, Hildrum R. Operation and sizing of energy storage for
567 wind power plants in a market system. Int J Elect Power Energy Syst 2003;25(8):
568 599–606.
569 [50] El-Hefnawi SH. Photovoltaic diesel-generator hybrid power system sizing. Renew
571 [51] Ashari M, Nayar CV. An optimum dispatch strategy using set points for a
573 [52] Skarstein O, Uhlen K. Design considerations with respect to long-term diesel
576 generation for rural and peri-urban populations of sub-Saharan Africa: the “flexy-
578 [54] Deshmukh MK, Deshmukh SS. Modeling of hybrid renewable energy systems.
580 [55] Diaf S, Diaf D, Belhamel M, Haddadi M, Louche A. A methodology for optimal
582 18.
35
ACCEPTED MANUSCRIPT
583 [56] Schmid J, Dincklage RD von. Power conditioning and control. Proceeding
585 [57] Schmid J, Schmid H. Inverter for photovoltaic system. Proceeding 5th Contractor's
587 122–32.
588 [58] Darras C, Sailler S, Thibault C, Muselli M, Poggi P, Hoguet JC, et al. Sizing of
589 photovoltaic system coupled with hydrogen/oxygen storage based on the oriented
591 [59] Rehman S. Prospects of wind farm development in Saudi Arabia. Renew Energy
592 2005;30:447–63.
593 [60] Ramli MAM, Hiendro A, Al-Turki YA. Techno-economic energy analysis of
594 wind/solar hybrid system: Case study for western coastal area of Saudi Arabia
602 [63] Yang H, Zhou W, Lu L, Fang Z. Optimal sizing method for stand-alone hybrid
603 solar/wind system with LPSP technology by using genetic algorithm. Sol Energy
604 2008;82(4):354–67.
36
ACCEPTED MANUSCRIPT
605 [64] Rajkumar RK, Ramachandaramurthy VK, Yong BL, Chia DB. Techno-economical
607 2011;36(8):5148–53.
608 [65] Storn RM, Price KV. Differential Evolution – A simple and efficient adaptive
609 scheme for global optimization over continuous spaces. 1995; pp. 1–12.
612 [67] Price KV, Storn RM. Lampinen JA. Differential Evolution A Practical Approach
614 [68] Huang VL, Qin AK, Suganthan PN. Self-adaptative differential evolution
617 [69] Huang VL, Qin AK, Suganthan PN, Tasgetiren MF. Multi-objective optimization
621 generations in Yanbu, Saudi Arabia. Proceeding World Renewable Energy Forum
622 2012:1–8.
37
ACCEPTED MANUSCRIPT
Highlights
1. Optimal sizing of PV/wind/diesel hybrid system with battery storage was analyzed.
2. The analysis was done using MOSaDE algorithm for the city of Yanbu, Saudi Arabia.
3. The energy management strategy is the main criteria to design and optimize the HMS.
4. The multi-objective optimization approach was used to analyze the cost and reliability.
5. HMS components are selected to meet the minimum energy cost and high reliability.