Journal Pre-proof

A new metaheuristic optimization algorithm inspired by human dynasties

with an application to the wind turbine micrositing problem

Shafiq-ur-Rehman Massan, Asim Imdad Wagan, Muhammad

Mujtaba Shaikh

PII: S1568-4946(20)30116-2
DOI: https://doi.org/10.1016/j.asoc.2020.106176
Reference: ASOC 106176

To appear in: Applied Soft Computing Journal

Received date : 28 April 2019

Revised date : 10 February 2020
Accepted date : 12 February 2020

Please cite this article as: S.-u.-R. Massan, A.I. Wagan and M.M. Shaikh, A new metaheuristic
optimization algorithm inspired by human dynasties with an application to the wind turbine
micrositing problem, Applied Soft Computing Journal (2020), doi:
https://doi.org/10.1016/j.asoc.2020.106176.

This is a PDF file of an article that has undergone enhancements after acceptance, such as the
addition of a cover page and metadata, and formatting for readability, but it is not yet the definitive
version of record. This version will undergo additional copyediting, typesetting and review before it
is published in its final form, but we are providing this version to give early visibility of the article.
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.

© 2020 Elsevier B.V. All rights reserved.

 Journal Pre-proof

1 A new metaheuristic optimization algorithm inspired by human dynasties with an

2 application to the wind turbine micrositing problem

of
3 Shafiq-ur-Rehman Massan1, 2, a, *, Asim Imdad Wagan2, b and Muhammad Mujtaba Shaikh3, 4, c
a
4 ([email protected]), b ([email protected]), c ([email protected])
1
5 Shaheed Zulfikar Ali Bhutto Institute of Science and Technology, Sindh, Pakistan.

pro
2
6 Mohammad Ali Jinnah University, Karachi, Sindh, Pakistan.
3
7 Department of Basic Sciences and Related Studies, Mehran University of Engineering and Technology,
8 Jamshoro, Pakistan.
4
9 Supply Chain and Operations Management Research Group, Mehran University of Engineering and Technology,
10 Jamshoro, Pakistan

11

12
*
re-
For correspondence: Email: ([email protected]) Mobile: +92-333-234 432 2

13 Abstract

14 Optimization is an art that is best performed by a well-tuned algorithm. Nature – instead of being fully
lP
15 deterministic – is evolutionary, vibrant and resourceful. The nature-inspired algorithms use the best
16 combination and evolution strategy in a given situation. In this work, a new metaheuristic algorithm is
17 developed by using social behavior in human dynasties. The motivation, conceptual framework,
18 mathematical model, pseudocode and working of the algorithm are described in this paper and the
19 adjoining papers. The proposed dynastic optimization algorithm (DOA) has evolved with the wind
rna

20 turbine micrositing (WTM) problem in mind. The proposed DOA has been successfully applied to the
21 traditional WTM and encouraging results have been obtained. It is demonstrated that the proposed
22 approach is equally viable as other existing algorithms, like the Genetic algorithm (GA) and Differential
23 evolution algorithm (DEA). The main advantage of the proposed DOA is that it is simple, unique, fast,
24 unbiased and versatile in comparison with others. The validation of results has been made with respect
25 to a few other mainstream algorithms in the literature, besides statistical sensitivity analysis is also
Jou

26 performed. The 95% confidence interval forecasts for the power enhancement and cost reduction by
27 using DOA against GA and DEA are encouraging and guarantee an adequate amount of mean increase
28 in power output and a considerable average cost reduction.

29 Keywords

1
 Journal Pre-proof

30 Soft Computing, Genetic Algorithm, Differential Evolution Algorithm, Wind turbine Micrositing,
31 Engineering optimization.

of
32 1) Introduction

33 The main target of optimization is to successfully find the maximum or the minimum of a given physical
34 system that is defined by an objective function and is bound by some real-valued constraints. Therefore

pro
 
35 for any objective function f (x ) , the goal is to find the best decision variable vector x in R n so that the
36 lower and upper bound constraints are satisfied. Mathematically, we can write:
 
37 optimize f ( x ) : x  R n (1)

38 where the lower bounds and upper bounds are defined as in (2),
  
39 l  xu (2)

40
41
re-
There exists a solution satisfying the constraints that are optimal for the function and may be approached
by a suitable technique.

42 Some of the most cited algorithms in the stochastic and metaheuristic domain are the Genetic algorithm
43 [2]–[7], Grey-wolf optimizer [8], Monte Carlo simulation [6], Differential Evolution Algorithm [9]–
lP
44 [12], the stochastic hill-climbing [13], Ant colony optimization algorithms [13], the Lion pride algorithm
45 [14], the Firefly algorithm [15]–[18], hybrid algorithms [19], bio-inspired Salp swarm algorithm [20],
46 etc.

47 These algorithms have some sort of a search strategy that is utilized to navigate the search space. The
rna

48 search strategy is combined with a stochastic component so as to minimize the effect of localization of
49 the search and such methods are known as metaheuristic methods.

50 There are two classes of algorithms in the metaheuristic domain; either deterministic or stochastic. Thus,
51 the deterministic algorithms are the ones that shall always give a unique and single answer given that
52 the initial conditions and initial guess are the same. However, in the case of stochastic algorithms, the
53 solution may differ with every implementation even if the same initial conditions and initial guess are
Jou

54 used [16].

55 The main reason for choosing a stochastic implementation is that it is less prone to local optima. A
56 deterministic algorithm may be prone to converge at a local optima [16]. This property is evident when
57 we compare the hill climbing and stochastic hill-climbing algorithms [13].

2
 Journal Pre-proof

58 The means of propagation of a stochastic algorithm are as diverse as ant pheromones to genes [13].
59 These algorithms have two main components, the main engine which is deterministic and the stochastic

of
60 component. The form and the implementation of the stochastic component may vary, it may work inside
61 the algorithm by changing an allele randomly [10] or it may work separately as a unique solution or a
62 random walk [21].

pro
63 In the past, researchers have approached multi-objective non-linear optimization problems and solved
64 them by using metaheuristic techniques and are referenced as [1], [2], [12]–[16], [18]–[20], [3]–[8], [10],
65 [11]. There also exist computational fluid dynamics techniques for the solution of the wind turbine
66 Micrositing problem [22], [23], etc.

67 The development of new techniques in the field of Nature-Inspired Algorithms is an interesting and
68 challenging task. It requires analogy from nature, from different ecosystems, to emulate and simulate.

69
70
re-
The field of optimizers is an emerging field, and most of the algorithms already available in the literature
suffer from one or more of the problems mentioned in [13]. Some techniques are not completely
71 described in the literature, and it is difficult for researchers to piece the parts together to replicate the
72 technique. There are many inconsistencies in techniques and a variety of implementations exist. Thus,
73 a given formulation of pseudocode may vary from one technique to another. Furthermore, new scripts
lP
74 and languages render the older implementations void as the new implementations, such as in Python,
75 are quicker, shorter and less processor intensive. There is a vast difference in the parameterization,
76 operations and data structures for the same technique in a number of papers spanning over its history.
77 Thus, an unambiguous way of writing the technique is important for leading the researchers towards
rna

78 successful application, formulation, further investigation and a possible extension of the technique.

79 The objective of this paper is to present a unique and robust metaheuristic algorithm that is focused on
80 solving non-linear bound-constrained optimization problems. The idea of the proposal has been a
81 consequence of inspiration from dynasties and revolutions in social sciences. The proposed algorithm
82 is thus named as the Dynastic optimization algorithm (DOA). The proposed algorithm has been
83 established mathematically in a formal way with a description of its working and pseudocode. The
Jou

84 performance of the proposed DOA has been checked along-with other existing algorithms on a well-
85 known engineering problem of WTM. The validation of the results of DOA has been made with other
86 approaches, and it is demonstrated that the application of DOA on the WTM problem yields a higher
87 value of the power produced and a lower value of cost per installation per unit turbine after installing
88 100 turbines. The efficiency of the DOA has been discussed, and results are also encouraging.

3
 Journal Pre-proof

89 2) Proposed Algorithm

90 Here, the proposed optimization algorithm DOA is presented in detail. The motivation for the

of
91 formulation of this algorithm was to present a simple and unbiased method of solution of optimization
92 problems. The proposed method has a low computational cost and is comparable with other methods.

93 The proposal, its mathematical model, discussion and example working are discussed. Further, the

pro
94 basics of WTM problem and a corresponding mathematical model is described for the implementation
95 and validation of the performance of DOA against some other algorithms in the literature.

96 2.1) The proposed Dynastic optimization algorithm (DOA)

97 The DOA has been written in an unambiguous method and is versatile for use in any type of application.
98 It would be very useful to future researchers and is a contribution of its own kind. Further, the analogy
99 may be derived from the synthesis of any data sample in the process of optimization. In every sample,
re-
100 there are troughs, peaks, and clusters. Therefore, the DOA utilizes the peaks to create local optima, then
101 the nearby or adjacent data points move towards these peaks in clusters. Whereas, a third significant
102 portion of the data points is generated stochastically that enriches the search experience.

103 It should be clearly noted that this algorithm is not a variant of the PSO as it is much simpler in the
lP
104 application. The PSO has to calculate the three directions in search space to update the next position. In
105 DOA, this is even simpler as the Euclidean distance is calculated to the rulers' position where all the
106 followers move. The purpose of randomness is added by randomly moving separate actors known as
107 explorers in the search space.

108 2.2) Inspiration from dynasties and revolutions

rna

109 The inspiration for the proposed algorithm is derived from human nature and from the social sciences
110 in particular. It is derived from the society that is a product of nature itself. Therefore, it may also be
111 termed as a nature-inspired algorithm. Here is a brief introduction to the algorithm.

112 At the very beginning of time mankind was comprised of groups of hunter-gatherers or explorers. They
113 scavenged the wilderness in search of food, shelter, and clothing.
Jou

114 With the passage of time, the concept of belongings and holding on to the belongings for extended
115 periods of time was born. Mankind chose to live in tribes.

116 This expanse in society gave rise to local nobles or the tribesmen with the best clothing, from the best
117 food gatherers. From amongst these tribesmen rose the Chieftains of the tribes.

4
 Journal Pre-proof

118 With the passage of time, the role of the best hunter-gatherer has never been lost. This explorer has
119 albeit changed its form from Columbus to Bill Gates. They can be very powerful indeed and leave a

of
120 deep impact on society.

121 Next, the normal populace is the workers. These workers always gather around the nobles or rich men.
122 They follow the lifestyle of the nobles and emulate their way of life.

pro
123 The nobles choose the king at the highest strata of society. Thus, the algorithm has two methods of
124 propagation, inspiration and stochastic.

125 Hence, the trio of explorers, workers, and nobles select the king. The King thrives in the dynasty and
126 succeeds in the next generation whereas the nobles and workers change with every iteration of the
127 population.

128 In the event of an explorer becoming a king, we can safely say that the new best solution has been
re-
129 attained. The inspiration of this algorithm may be derived from the various dynasties that ruled the
130 Indian sub-continent.

131 2.3) Mathematical model of the proposed DOA

132 Step 1. Consider generating a random population Np, which is then evaluated.
lP
133 Np = {1, 2, 3….m} (3)

134 m  I , and

135 N r  N p , N w  N p and N e  N p ,
rna

136 Where, Nr, Nw and Ne are the number of rulers/kings, workers/followers and explorers, respectively.

137 Step 2. Now, for any objective function,

138 F(x) where x =(x1, x2, x3, x4….xm)T (4)

139 Step 3. The rr percentage of the population shall be considered as the Rulers, and they shall be ranked
140 and their positions will be fixed. Then about 60% or a random percentage of the population will be
Jou

141 considered as workers. These workers will only perform local search around a particular ruler with a
142 fixed radius. The rest of the remaining population will be considered as the Explorers which will
143 randomly explore the space for optimization. The total ratio of the three percentages should add to 100%.
144 For example, if rr = 20%, then:

145 rr = 0.2, rw=0.6 and re=0.2

5
 Journal Pre-proof

146 Nr=rr*Np N r  I and so on

147 Step 4. The rulers shall be ranked with each iteration, such that,

of
148 Rank(Nr) = max {|Np| N r  N p , N r  0.2 * N p } and so on

149 Step 5. Rulers’ position shall be fixed. The workers shall perform the localized stochastic search around

pro
150 the rulers. The Workers shall move around the closest rulers such that they move in a fixed radius around
151 the rulers,

152 radw = 0.4 ss rad w  1 , (5)

153 ss is the search space.

154 The distance between the rulers and the workers shall be governed by the cartesian equation,

155 Distance =
2
xr ,i  x w, j
2
re- (6)

156 where, r = Rulers and w = Workers and i,j  I

157 Step 6. The explorers shall be randomly generated and shall move at random positions in the unexplored
lP
158 space.

159 xe  rand ( xi )i  I (7)

160 All the population shall be ranked and the rulers shall again be selected. Then steps 3 to 6 shall be
161 repeated.
rna

162 Step 7. After the “n” number of iterations or reaching a minimum error, the algorithm will return the
163 ruling class as the best solution.
 
164 xbest  max F ( x ) (8)

165 Thus, the best ruler, from the best class of rulers, which is significantly better than the workers and the
166 explorers is selected [24].
Jou

167 2.4) Discussion about the DOA

168 The proposed DOA has a number of similarities with the GA, however, it is less complex, as it does not
169 have the processes of crossover and mutation.

170 The DOA followers search for possible optimal points in the vicinity of the rulers in a given society. It
171 introduces the concept of parallel search which has a broader scope as compared to some narrowly tuned
6
 Journal Pre-proof

172 algorithms. The DOA can handle multi-modal and volatile optimal functions in an efficient manner. It
173 can be deduced that the DOA is a special class belonging to the swarm type of algorithms.

of
174 The DOA is much simpler than the standard particle swarm optimization algorithm [1]. The migration
175 of the workers in the DOA is adjustable so that they may approach a given ruler from multiple directions
176 and converge on it. Varying the number of rulers assures that the search space is sufficiently explored.

pro
177 It should be noted that the distance between the rulers and the workers is not limited to their Euclidian
178 distance. This is because other measures of distance can also be taken into account such as time lag as
179 in job scheduling problems on the internet by the length of a connection and its associated cost.

180 It has been experienced that as the iterations are increased, the DOA improves its algorithmic
181 performance. However, the right combination of the mix of rulers, workers, and explorers has to be
182 judged for the problem under consideration. It is evident that the DOA is quick to converge on the local
183

184
re-
optima as well as on the global optima.

Another advantage of the DOA is that it can search in the vicinity of the global best solution as it
185 approaches it from all sides. This is because the workers can move independently towards their nearest
186 ruler. This is not possible in the implementation of the GA [2] and the PSO [1].
lP
187 Another advantage of the DOA is that the king or the global best is the penultimate solution.

188 2.5) The wind turbine micrositing (WTM) problem

189 It was well quoted by Holland [25], that living organisms are some of the best problem solvers. Hence,
190 it may be attributed to this property of living organisms that we have emulated for the solution of the
rna

191 WTM problem using the proposed DOA.

192 2.6) The Mathematical model of WTM problem

193 This Jensen model was formulated by N.O Jensen in the year 1983 at the famous Riso National
194 Laboratory Denmark [26]. The Jensen model is known for its simplicity and ease of deployment in
195 mathematical calculations for turbines arranged in open fields. It is possible to emulate and simulate
Jou

196 multiple wakes generated one behind the other with this model.

197 In this work, we make the same assumptions as in [11], [12], [18], [19] such as Radius of the wind
198 turbine rotor = 40 m, Height of the wind turbine hub = 60 m, and Thrust coefficient = 0.88. Thus, we
199 may depict our simulation for a simple scenario as shown in Fig. 1.

7
 Journal Pre-proof

of
pro
200
re-
201 Fig. 1. The simple drawing of the Jensen model illustrating the wake at a distance behind the turbine
lP
202 In Fig. 1. the initial wind velocity u0, is incident on a wind turbine having a rotor radius rr. The effects
203 of the wake can be felt at a distance of x behind the turbine. The radius of the wake effect at this
204 distance is r1, which is governed by the illustrated equation.

205 This model is based on the assumption that the total momentum is conserved within a wake. It also
rna

206 stipulates that this wake is spread in a linear fashion downstream from the turbine. Hence, by
207 cascading turbines in this linear model, we may accommodate multiple wakes and their interaction.
208 Here, we may utilize these variables,

209 rr = Initial wake radius which is also the Turbine radius

Jou

210 r1 = Wake radius at a distance of x

211 x = downstream distance where wake radius is calculated

212 Thus, Betz’s theory gives the wind speed after the rotor as,

8
 Journal Pre-proof

213 𝑈 𝑢 1 (9)

of
214 The free stream wind speed or mean speed is denoted by U0. Moreover, the axial induction factors, a
215 can be found by the following equation as,

216 CT = 4 𝑎 (1 – 𝑎) (10)

pro
217 The equation governing the relationship between the turbine radius rr and the incident radius r1 located
218 at a distance of x is,

219 𝑟 𝑟 (11)

220
re-
Whereas, α, the entertainment constant is given by the equation,

.
221 𝛼 (12)

222 Where,
lP
223 z = Hub height

224 z0 = Terrain roughness (for flat areas = 0.3)

rna

225 Hence the Jensen model may be elaborated to yield the equation for wind turbines that have multiple
226 wakes incident on them, given by,

227 𝑢 𝑢 1 (13)

228 Where,
Jou

229 Nt = the number of Turbines in incident wake

230 u0 = original wind speed,

231 UI = wind speed at the ith turbine,

9
 Journal Pre-proof

232 u is the resultant wind speed at incident wind turbine

233 Hence, we may reach the conclusion that the number of wind turbines and their subsequent placement

of
234 is equally important.

235 Further, the Available Power is yielded as per the equation,

pro
236
237 Available Power = 𝜌 𝐴 𝑢 (14)

238

239 If we multiply the efficiency of the turbine in this equation, we have,

240 Available Power = η 𝜌 𝐴 𝑢

re- (15)

241

242 The maximum power produced is thus,

lP
243 Power Produced = 0.3𝑢 Kilowatts (16)

244 The efficiency, η, may be reached by utilizing the aerodynamic efficiency of rotors. From the Betz
245 limit we have,
246
rna

Nt

 0.3  u
i 1
3
i
247 η= (17)
N t (0.3  u03 )

248
Jou

249 Hence,

250 𝐸𝑓𝑓𝑖𝑐𝑖𝑒𝑛𝑐𝑦 (18)

.

10
 Journal Pre-proof

251 The cost model used in this study is dimensionless and empirical and has also been referenced in
252 earlier literature. The cost model predicts a maximum reduction of cost by one third with the

of
253 installation of every new turbine.

254 The cost model is as follows,

pro
.
255 𝐶𝑜𝑠𝑡 𝑁 (19)

256 Thus, the cost reduces by one-third as every new turbine is added to the system.

257 Consider Fig. 2 showing the positioning of different turbines (nine as an example) with the wake
258 effect of each turbine. The case of two turbines at the same position, for example, turbines “1” and “3”
259 is shown in Fig. 3. In such a case, if both x-distance and y-distance between the turbines are less than
re-
260 199 then the solution is discarded. Whereas, if either one of x distance or y distance is less than 199, it
261 means that the turbines are adequately separated.

262 On the other hand, the velocity and power of a turbine experiencing wake can be calculated. For
263 example, consider turbines “1” and “7” as shown in Fig. 4, where “7” is fed by the wake of “1”. The
lP
264 radius of the wake is given by the equation,

265 radius  rr  Y (20)

266 If xmin < x and xmax > x, then, at a given distance Y, if x lies between the values of xmin or xmax then
rna

267 it experiences a wake. The velocity is calculated by using initial wind speed u0, location of the turbine.
268 The position of both turbines is checked and the velocity ratio is calculated by first calculating the
269 vertical distance between the two turbines, the final equation of velocity is given in (13). This velocity
270 ratio is multiplied with the velocity to get the incident velocity at the turbine.

271 In the case of a cascaded turbine, like “9” with “4” and “2” in Fig. 5, the turbines are eliminated one by
272 one as they approach the final turbine. The initial wind speed u0 decreases with every wind turbine that
Jou

273 it encounters.

274

11
 Journal Pre-proof

of
pro
275

276 Fig. 2. Wake effect of different turbines Fig. 3. Two turbines at the same position (1 and 3)

re-
lP

277
rna

278 Fig. 4. Velocity and power for a turbine in wake Fig. 5. Cascaded turbines in wake

279 3) Results and discussion

280 3.1) Numerical results and discussion of the performance of DOA for WTM problem

281 To check the performance of the proposed DOA for the WTM problem, the parameters used for the
282 numerical simulation are given in Table 1 and [27]. In a square-grid of 2 km by 2 km area, an attempt
Jou

283 has been made to install 100 turbines. The coordinates in the search region mesh are divided into 100
284 points for x and for y, thus in total a 100 by 100 mesh. The maximum number of iteration used to search
285 for the optimal positioning of the wind turbines has been set at 10,000. The ratio of the kings, followers,
286 and explorers has been set as 5%: 55%:40%, respectively. The constant wind speed of 12 m/s has been
287 used from the succession of similar simulations in the literature [11], [12], [18], [19]. The optimal layout
288 of the wind farm after installing 100 turbines with optimal placement of turbines is shown in Fig. 6 (f).
12
 Journal Pre-proof

289 The successive installation of turbines starting from a single turbine to 25, 50, 54, 75 is also shown in
290 Fig. 6 (a)–(e), respectively. The total power produced and the cost of inclusion after installation of every

of
291 next turbine in the grid have been noted for the proposed DOA, and the results are shown in Figs. 7 and
292 8, respectively. The efficiency of the proposed DOA has been almost ideal, but due to stochastic nature
293 at instances it dropped slightly, the results are shown in Fig. 9.

pro
294

295 Table 1. The parameters for simulation of the DOA

Number Parameter Values

1) Niter, Number of iterations 10,000

2) Np, Number of population 100

3)

4)
re-
rr, Ratio of rulers/kings 0.05

rw, Ratio of workers/followers 0.55

5) re, Ratio of explorers 0.4

6) radw, Radius of workers 0.4

lP
7) U0, wind speed 12 m/s

296

297
rna
Jou

13
 Journal Pre-proof

(a) DOA frame after installation of 1 turbine (b) DOA frame after installation of 25 turbines

of
pro
turbines
re-
(c) DOA frame with optimal positioning of 50 (d) DOA frame with optimal positioning of 54
turbines
lP
rna

(e) DOA frame after installation of 75 (f) DOA frame after installation of 100
turbines turbines
Jou

298 Fig. 6. Optimal layouts of different positioning of successive turbines by using DOA

14
 Journal Pre-proof

of
pro
re-
299

300 Fig. 7. The power produced by DOA after installing every successive turbine
lP
rna
Jou

15
 Journal Pre-proof

of
pro
re-
lP
301

302 Fig. 8. Cost per installation of the unit turbine using proposed DOA
rna
Jou

16
 Journal Pre-proof

of
pro
re-
303
lP
304 Fig. 9. The efficiency of DOA in the simulations after installing a new turbine successively

305
rna
Jou

17
 Journal Pre-proof

306 3.2) Comparison and validation of results of proposed DOA against other existing algorithms

307 The optimal statistics in terms of total power produced and cost per unit turbine have been compared in

of
308 detail with a comparison to a few similar studies in the literature [2], [11], [12]. Some other algorithms
309 from literature like the works of Mosetti et al [4], Grady et al. [5], Mittal et al. [3] and Marmidis et al.
310 [6] have been used for validation. The parameters used for GA [2], DEA [11], [12], and other studies

pro
311 [3]–[6] have been taken from the corresponding papers.

312 The power produced by using DOA was found higher than the GA method for nearly all the instances
313 of the algorithm run. However, the power produced by the DOA is lesser than the GA at the instance of
314 turbines 18, 19 and 25 [2]. At the installation of the 18th turbine, the GA returns a total power output of
315 9,331.20 kW whilst the DOA returns a value of 9,328.22 kW [2]. Similarly, at the installation of the
316 19th turbine, the GA returns the value of 9,849.60 kW and the DOA returns a value of 9,845.28 kW [2].
317
318
319
re-
And, at the installation of the 25th turbine, the GA returns a value of 12,862.08 kW while the DOA
returns a value of 12,805.65 kW [2]. However, the power difference is evident from the installation of
the 20th turbine to the 24th turbine and from the 26th turbine to the 100th turbine. Hence, for the 20th
320 turbine, the DOA returns a value of 10,359.23 kW and the GA lags at a value of 10,351.68 kW [2].
321 Similarly, at the installation of the 24th turbine, the DOA results in a value of 12,429.13 whereas the GA
lP
322 lags at a value of 12,360.00 kW [2]. And at the installation of the 26th turbine, the DOA gives a value
323 of 13,453.78 kW while the GA gives a value of 13,364.16 kW [2]. Finally, at the installation of the
324 100th turbine, the DOA gives a value of 49,831.45 kW while the GA lags behind at a value of 48,452.26
325 kW. Similarly, the power produced by the proposed DOA remained higher as compared to DEA [11],
326 [12] except in a few instances. The improvement in power by the proposed DOA against GA and DEA
rna

327 are shown in Fig. 10. The negative power differences in Fig. 10 highlight a few cases where the GA and
328 DEA were slightly taking edge over the results of DOA in the numerical simulation. However, after a
329 complete run of algorithms for installation of 100 turbines, the final positioning of turbines by the
330 proposed DOA has been found more efficient than GA and DEA in terms of power produced and cost
331 reduction. This effect is attributed to the stochastic nature of both the algorithms [2], [11], [12].
Jou

332 From the results in [2], [11], [12] and of Fig. 10, it is evident that the DOA gives a lower cost for nearly
333 all instances of the simulation. This holds true with GA for all instances except for the installation of
334 the 18th, 19th, and 25th turbines. At the installation of the 18th turbine, the GA has a lower cost of
335 0.0016519 whereas the DOA lags at a value of 0.0016524. Similarly, at the installation of the 19th turbine,
336 the GA returns a cost of 0.0016291 whereas the DOA returns a higher value of 0.0016298. And for the
337 installation of the 25th turbine, the GA results in a cost of 0.0015142 and the DOA results in a higher
18
 Journal Pre-proof

338 cost of 0.0015209. A similar situation arises when comparing cost reduction by DOA against DEA. The
339 reduction in cost per unit turbine by the proposed DOA against GA and DEA are shown in Fig. 11. The

of
340 difference in cost is minimum when DOA is compared with both algorithms, but given the magnitude
341 of the lifetime of the project, the cost would become significant. This statement assumes that the
342 operations and maintenance costs do not affect the total cost by a significant factor.

pro
2000.00
Power produced difference (kWh)

1500.00

1000.00

500.00

0.00
1
6
11
16
21
26
31
36
41
46
51
56
61
66
71
76
81
86
91
96
‐500.00

‐1000.00

Power difference (DOA‐GA)
re-
Number of turbines

Power difference (DOA‐DEA)
343

344 Fig. 10. Increase in power produced by proposed DOA against GA [1] and DEA [9-10]
lP
0.00006000
Cost per unit turbine  difference 

0.00004000

0.00002000
rna

0.00000000
1
6
11
16
21
26
31
36
41
46
51
56
61
66
71
76
81
86
91
96

‐0.00002000
Number of turbines

Cost difference (GA‐DOA) Cost difference (DEA‐DOA)
345

346 Fig. 11. Decrease in cost per unit turbine by proposed DOA against GA [1] and DEA [9-10]
Jou

347 From the results, it is evident that the optimal solution for GA exists at the instance when 54 turbines
348 are installed at a power output of 27,169.52 kW and a cost of 0.0013292. However, the power output of
349 DOA is higher at the installation of 54 turbines at 27,748.88 kW at a lower cost of 0.0013039. These
350 results for DOA are also higher than the results of DEA. Similarly, the optimal point in the case of the
351 DOA occurs at the installation of the 61st turbine at a power output of 31,278.57 kW and at a cost of

19
 Journal Pre-proof

352 0.0013011. Here, the GA lags behind at a power output of 30,512.21 kW and at a higher cost of
353 0.0013338, and the DEA exhibits a power output of 31,149.41 kW and at a higher cost of 0.0013065.

of
354 Fig. 12 shows the comparison of power outputs of proposed DOA, GA [2] and DEA [11], [12] at optimal
355 points. The power output by the proposed remained highest of all.

356 Fig. 13 (a) illustrates that the DOA gives higher power output as compared to Mosetti et al. [4], GA [2]

pro
357 and DEA [11], [12] at the installation of 26 turbines as mentioned limit in Mosetti et al. [4]. Whilst, Fig.
358 13 (b) illustrates that the DOA gives lower cost as compared to Mosetti et al [4], GA [2] and DEA [11],
359 [12] when 26 turbines are installed. Thus, it may be safely concluded that the proposed DOA gives better
360 performance as compared to Mosetti et al [4], GA [2] and DEA [11], [12]. Fig. 14 illustrates that the
361 DOA gives higher power output and lower cost per unit turbine as compared to Grady et al. [5], GA [2]
362 and DEA [11], [12] at the installation of 30 turbines. It is evident that the results of the DOA simulation
363
364
re-
are superior to the results of Grady et al. [5], GA [2] and DEA [11], [12], as the power produced is
higher and the cost is lower.

365 Figs. 15 (a) and 15 (b) illustrate that the proposed DOA has a superior performance as compared to
366 Marmidis et al. [6], GA [2] and DEA [11], [12] at the installation of 32 turbines in terms of higher power
367 production at lower cost per installation of unit turbine. A similar comparison of the performance of
lP
368 proposed DOA with studies of Mittal et al. [3] and other used algorithms GA [2] and DEA [11], [12] is
369 shown in Fig. 16. At the installation of 44 turbines, the DOA has a better performance as it results in
370 higher power obtained shown in Fig. 16 (a). On the other hand, from Fig. 16 (b), it is evident that the
371 lowest cost is reported when the DOA is as compared to Mittal et al. [3], GA [2] and DEA [11], [12] at
rna

372 the installation of 44 turbines.

373 3.3) Contribution

374 The novel contribution of the proposed DOA indicated that this algorithm gave better results than DEA
375 [11], [12], GA [2], Mittal et al. [3], Marmidis et al. [6], Mosetti et al. [4] and Grady et al. [5]. This factor
376 may be attributed to the simplistic, unassuming swarm-based structure and efficient operation of the
377 proposed DOA. It is anticipated that this algorithm shall continue to grow and improve with every
Jou

378 successful implementation over its lifetime. Hence, the above results strongly suggest that though this
379 algorithm is a new entrant in this field of study however, it is giving competitive results with regard to
380 the WTM problem.

20
 Journal Pre-proof

32000
31000
30000

of
29000
28000
27000
26000
25000

pro
54 Turbines 61 Turbines
Power by GA Power by DOA Power by DEA
381

382 Fig. 12. Comparison of optimal point of GA [1], DEA [9-10] and proposed DOA

14000 0.00165
13500
13000
12500
12000
11500
re- 0.0016
0.00155
0.0015
0.00145
0.0014
26 Turbines installed 26 Turbines installed

Mosetti et al. Power by GA
Mosetti et al. Cost by GA
Power by DOA Power by DEA
lP
Cost by DOA Cost by DEA
383

384 (a) Power produced in kW (b) Cost per unit turbine

385 Fig. 13. Comparison of power produced and cost per unit turbine at the installation of 26 turbines by
386 Mosetti et al. [4], GA [2], DEA [11], [12] and proposed DOA
rna

16000 0.0016
15500 0.00155
15000 0.0015
0.00145
14500 0.0014
14000 0.00135
13500 0.0013
30 Turbines installed 30 Turbines installed
Jou

Grady et al. Power by GA

Power by DOA Power by DEA
Grady et al. Cost by GA Cost by DOA Cost by DEA
387

388 (a) Power produced in kW (b) Cost per unit turbine

21
 Journal Pre-proof

389 Fig. 14. Comparison of power produced and cost per unit turbine at the installation of 30 turbines by
390 Grady et al. [5], GA [2], DEA [11], [12] and proposed DOA

of
16600 0.001425
16550 0.00142
0.001415
16500 0.00141

pro
16450 0.001405
16400 0.0014
0.001395
16350 0.00139
16300 0.001385
16250 0.00138
32 Turbines installed 32 Turbines installed

Marmidis et al. GA DOA Power by DEA

Marmidis et al. GA DOA Cost by DEA
391

392 (a) Power produced in kW (b) Cost per unit turbine

393
394
re-
Fig. 15. Comparison of power produced and cost per unit turbine at the installation of 32 turbines by
Marmidis et al [6], GA [2], DEA [11], [12] and proposed DOA
lP
22800 0.00137
22600 0.00136
22400 0.00135
0.00134
22200 0.00133
22000 0.00132
21800 0.00131
21600 0.0013
21400 0.00129
44 Turbines installed 44 Turbines installed
rna

Mittal et al. GA DOA Power by DEA

Mittal et al. GA DOA Cost by DEA
395

396 (a) Power produced in kW (b) Cost per unit turbine

397 Fig. 16. Comparison of power produced and cost per unit turbine at the installation of 44 turbines by
398 Mittal et al. [3], GA [2], DEA [11], [12] and proposed DOA

399
Jou

400 3.4) Statistical analysis, sensitivity, and forecasts of results

401 Table 2 summarizes the descriptive statistics for the power produced and cost per unit turbine by using
402 the proposed DOA, DEA and GA. It appears from Table 2 that all methods start with a power of 518.4
403 kW from the installation of the first turbine whereas the proposed DOA power reaches the highest level
404 after installation of the 100th turbine. The power after installation of the 100th turbine for the DEA and
22
 Journal Pre-proof

405 GA are, respectively, the second and third in rank. Similarly, the proposed DOA beats all other methods
406 from the viewpoint of being cost-efficient as the cost required per installation of a unit turbine in the

of
407 process of installing 100 turbines for the DOA is smaller than those in DEA and GA. While looking at
408 average power and cost values in Table 2, the following rating of methods appears clearly in ascending
409 levels of power: GA, DEA, DOA. On the other hand, from the viewpoint of cost efficiency, the rating
410 of methods is DOA, DEA, GA. The standard errors of estimate and the standard deviations in the power

pro
411 and cost levels for 100 turbines are also reported in Table 2.

412 While it appears clearly from descriptive stats of data, i.e. Table 2, that the maximum power level and
413 minimum cost level for DOA are achieved as compared to DEA and GA, we also give results of the t-
414 test to demonstrate our observations and hypotheses about power and cost of all methods more clearly.

415 In the student’s t-test, we compare the average power and average cost by using DOA for the WTM
416

417
re-
problem against other methods. For this, we set the following hypotheses:

The null hypothesis, Ho: the average power (cost) produced by DOA and any other method X is the
418 same. Mathematically,

419 𝜇 𝜇 (21)
lP
420 The alternative hypothesis, Ha: the mean power (cost) by DOA is significantly different than that by any
421 other methods X. Mathematically,

422 𝜇 𝜇 (22)

423 at 0.05 level of significance. Here X = {GA, DOA}.

rna

424 Whenever we see a significance level of more than 0.05 in a difference, the null hypothesis will be true,
425 otherwise alternative is accepted. Moreover, if the sign of a difference is positive then the average power
426 (cost) produced by DOA is considered higher than a method X, otherwise lower at the same level of
427 significance. The mathematics of the procedure of statistical comparison in energy saving and cost
428 reduction can be found in [28], [29].

429 It appears from Table 3, at 0.05 level of significance, that there is a significant difference in the power
Jou

430 produced, as well as cost, by using proposed DOA and other methods. So, the alternative hypothesis is
431 accepted. This shows that the average power (cost) levels achieved by using DOA in installing 100
432 turbines are significantly different than the levels attained by other methods, at a 5% level.

433 Moreover, it is clear from Table 3 that the power achieved by DOA on average is higher than DEA and
434 GA due to positive differences. Similarly, the average cost level of DOA after installing 100 turbines is
23
 Journal Pre-proof

435 lower than DEA and GA due to negative signs. At a 5% level of significance, it can be concluded that
436 the DOA produces higher power output on average than the DEA and GA, and it is also more cost-

of
437 efficient than these methods. The 95% confidence intervals for the expected differences in the power
438 and cost levels of DOA against other methods in the similar installation of 100 turbines in the future are
439 shown more explicitly in Figs. 17 and 18, respectively. We have followed the procedure of forecasts
440 and similar statistical models from [28], [29]. Fig. 18 displays the present average increase in power and

pro
441 reduction in cost by using proposed DOA against GA and DEA; the future forecasts of the upper and
442 lower bounds of the confidence intervals are also shown. The encouraging confidence interval forecasts
443 for the power enhancement and cost reduction are encouraging, and indicate that, out of the 100 times
444 the next installations of the similar wind turbines under the assumptions of this study, at least 90 times
445 the computed bounds will hold, with a type-I error of only 0.01.

446

447
re-
448

449
lP
rna
Jou

24
 Journal Pre-proof

450 Table 2 Descriptive statistics for the power produced and cost incurred by all algorithms

Standard
N Mean / Standard

of
Range Minimum Maximum error in Variance
Variables (size) average Deviation
mean
DEA 100 49208.22 518.40 49726.62 25510.2400 1433.55896 14335.58962 2.055E+8
POWER DOA 100 49313.05 518.40 49831.45 25699.3844 1453.41711 14534.17105 2.112E+8

pro
GA 100 47933.86 518.40 48452.26 25137.0871 1405.11270 14051.12695 1.974E+8
DEA 100 0.0006253 0.0013026 0.0019279 0.001441852 0.00001934 0.00019347 0.000
COST DOA 100 0.0006267 0.0013011 0.0019278 0.001435473 0.00001971 0.00019718 0.000
GA 100 0.0005987 0.0013292 0.0019279 0.001457067 0.00001851 0.00018514 0.000
451

452

453

Standard
re-
Table 3 t-test details to see significant difference in power and cost of DOA against other methods

Standard 95% confidence interval

Degrees
Difference Mean deviation in error in forecasts of the mean
differences t-value of p-value
variables differences mean mean
freedom
differences differences Lower Upper
lP
P DOA–DEA 189.14440 347.11954 34.71195 120.26835 258.02045 5.449 0.000
O
W
E DOA–GA 562.29730 507.74907 50.77491 461.54887 663.04573 11.074 0.000
R 99
C DOA–DEA -6.37810E-6 1.073491E-5 1.07349E-6 -8.5081E-6 -4.2480E-6 -5.941 0.000
rna

O
S DOA–GA -2.15931E-5 1.516020E-5 1.51602E-6 -2.4601E-5 -1.8584E-5 -14.243 0.000
T
454

455

456
Jou

25
 Journal Pre-proof

700

600

of
500 95% confidence
400
DOA–DEA
300
DOA–GA
200

pro
100

0
Mean differences in power outputs
457

458 Fig. 17. 95% confidence interval forecasts for the increase in power output by DOA against GA and
459 DEA

‐1.00E‐07
‐2.60E‐06
‐5.10E‐06
re-
Mean differences in costs per unit turbine installation

‐7.60E‐06
‐1.01E‐05
DOA–DEA
‐1.26E‐05
DOA–GA
lP
‐1.51E‐05 95% confidence
‐1.76E‐05
‐2.01E‐05
‐2.26E‐05
‐2.51E‐05
460
rna

461 Fig. 18. 95% confidence interval forecasts for the cost difference by using DOA against GA and DEA

462 3000

463 2500

464 2000

1500 Upper bound
465 Lower bound
Jou

1000
466 Average
500
467 0
P_DOA‐DEA P_DOA‐GA C_DOA‐DEA C_DOA‐GA
468 *10^8 *10^8

26
 Journal Pre-proof

469 Fig. 19. The average difference, lower and upper bounds of the expected differences in parameters using
470 DOA against GA and DEA.

of
471 3.5) Application to test functions

472 The DOA, DEA and GA were applied to the Rosenbrock’s (f1) and Easom’s (f2) test functions which
473 are considered as conventional test functions for optimization methods.

pro
474
2

f1  x1 , x2   1  x1   100 x2  x1 
2 2
(23)

f 2 x1 , x2    cos x1 cos x2 e  x1    x2   

2 2
475 (24)

476 The minimum values of these functions are respectively 0 and -1 at (1,1) and (π, π).

477 We used same parameters as used in previous sections for WTO problem.
re-
lP
rna
Jou

478

479 Fig. 20. Comparison of minima attained versus number of generations by all methods for Rosenbrock’s
480 function

27
 Journal Pre-proof

of
pro
re-
481
lP
482 Fig. 21. Comparison of minima attained versus number of generations by all methods for Easom’s
483 function

484 It is evident from Figs. 20-21 that DOA obtains the minimum value of f1 more frequently than GA and
485 DEA for a range of number of generation points. Proposed DOA exhibits substantial accuracy with
rna

486 regards to the results closer to expected minimum value for both functions, as in Figs. 20-21. On the
487 other hand, GA and DEA in some cases are less likely to attain the expected minimum value. More on
488 the mathematical insights and application of other widely used test functions can be found in near future.

489 4) Main achievements, recommendations, and future insights

490 The DOA algorithm has been developed for the solution of the WTM problem, but it is versatile enough
491 for application to a variety of non-linear optimization problems with a finite number of constraints. It is
Jou

492 a stochastic algorithm belonging to the class of nature-inspired metaheuristic algorithms. The term
493 stochastic refers to the fact that two different instances of the same algorithm may yield two entirely
494 different solutions. It has widespread applicability to a vast number of fields of science. The application
495 of the algorithm is open for debate and its further advantages are yet to be discovered. However, it is
496 concluded that the solution provided by the DOA is superior to the solution offered by the GA [2], DEA
28
 Journal Pre-proof

497 [11], [12], Mittal et al. [3], Marmidis et al. [6], Mosetti et al. [4] and Grady et al. [5]. Further research
498 and innovation would open many more vistas of application of this algorithm.

of
499 The DOA may find applications for developing a number of multi-agent systems and robotics, in
500 unmanned vehicles. It has demonstrated that group dynamics can achieve group goals. This is because
501 the collective convergence is more favorable as compared to individual excellence. The application of

pro
502 DOA may also be found in training of ANN weights and for the study of interstellar systems where the
503 entropy of the system may change due to unforeseen circumstances and the global optima is volatile.

504 The following recommendations have been made during the course of this study,

505 1) The DOA can be tested on a wide number of problems in the optimization domain.
506 2) The DOA can be utilized in other real-world problems as it is anticipated to do so.
507 3) The convergence of the proposed DOA can be further improved by using higher values of the
508
509
characteristic parameters.
re-
4) An increase in computational power of the future would enable more extensive search of the
510 solution space and further evaluation of algorithmic efficiency.
511 5) Real-world scenarios and uneven terrains can be built and simulated using mainframe computers.
512 6) The metaheuristic approach can be combined with the computational fluid dynamics (CFD)
lP
513 approach to yield superior layouts of wind farms.
514 7) Presently, the layout of the wind farms is being proposed by humans that are run for a number
515 of simulations to reach the optimal values. In our future work, we would seek to automate this
516 final frontier so as enable the computer to suggest by itself the best placement of the wind
rna

517 turbines.
518 8) The DOA can be further tested for getting the best possible combination for such type of
519 problems.
520 9) It is also deemed feasible to purchase computing power from the internet cloud to run higher-
521 order simulations.

522 5) Conclusion
Jou

523 In this paper, a new metaheuristic algorithm has been formulated and applied to the WTM problem. It
524 successfully optimized the placement of 100 wind turbines in a 2 km by 2 km area. The comparisons
525 were drawn with earlier results obtained by using modern algorithms of GA, DEA, Mittal et al., Monte
526 Carlo – Marmidis et al., Mosetti et al. and Grady et al. It is evident that the stochastic solution of the

29
 Journal Pre-proof

527 WTM works out more solutions as compared to the human trial and error method indicated by Rajper
528 et al. [2]. This is due to the sheer number of iterations evaluated by the computer.

of
529 The simulation and the results clearly demonstrate that the DOA gives a superior performance as
530 compared to the other approaches in the literature. Moreover, it is a novel and validated contribution in
531 the field of metaheuristics. It is evident at the installation of the 100th turbine the power produced using

pro
532 DOA was recorded as higher than using than the DEA and the GA. Whereas the cost per installation of
533 unit turbine after 100th turbine was smaller than GA and DEA.

534 It is seen that the results are consistent and follow the results obtained by the GA and other studies. The
535 results of DOA outperform the GA and DEA in terms of success rate and efficiency. The final results
536 suggest that the new algorithm, DOA has good robustness and accuracy as compared to the mainstream
537 algorithms.

538
539
re-
The proposed DOA as a new metaheuristic algorithm is expected to open up a new era in stochastic
metaheuristic computing. A new class of optimization algorithm with reference to the problem of wind
540 turbine micrositing, has been introduced. This algorithm shall find widespread usage in all fields of
541 science. It has been reported that it is often difficult to evaluate such functions that have uncertainties in
542 their objective functions or the constraints. Such functions may be termed as volatile objective functions
lP
543 and the constraints may be termed as volatile constraints. Therefore, the optimization problem may be
544 termed as a stochastic optimization problem. The DOA shall be very useful in the evaluation of optima
545 for both volatile and non-volatile objective functions. Such types of functions may be found in
546 interstellar studies of nebulae where disruptions in stellar systems change the entropy of the system from
rna

547 one optimal/maximum point to the other.

548 It is suggested that even more rigorous testing of the DOA shall be carried out in the future with a
549 number of test functions in order to test the efficacy of the algorithm. The efficiency of the DOA for
550 solving other NP-hard problems shall be investigated in the future. In the future, an extended version of
551 this algorithm shall also be presented for the scientific community.

552
Jou

553 Acknowledgment

554 The authors wish to thank Dr. Hafiz Abdul Ghani Shaikh for continuous motivation and encouragement
555 while completing this work.

556 References

30
 Journal Pre-proof

557 [1] J. Kenneddy and R. C. Eberhart, “Particle Swarm Optimization,” in Proceedings of 1995 IEEE
558 International Conference on Neural Networks, 1995, pp. 1942–1948.

of
559 [2] S. Rajper and I. J. Amin, “Optimization of wind turbine micrositing: A comparative study,”
560 Renewable and Sustainable Energy Reviews, vol. 16, pp. 5485–5492, 2012.

561 [3] A. Mittal, “Optimization of the layout of large wind farms using genetic algorithm, MS thesis,”

pro
562 Case Western Reserve University, 2010.

563 [4] G. Mosetti, C. Poloni, and B. Diviacco, “Optimization of wind farms positioning in large wind
564 farms by means of a genetic algorithm,” Wind Engineering and Industrial Aerodynamics, no. 51,
565 pp. 105–116, 1994.

566 [5] S. A. Grady, M. Y. Hussaini, and M. M. Abdullah, “Placement of wind Turbines using genetic
567 algorithms,” Renewable Energy, vol. I, no. 30, pp. 259–270, 2005.
re-
568 [6] G. Marmidis, S. Lazarou, and E. Pyrgioti, “Optimal placement of wind turbines in a wind park
569 using Monte Carlo simulation,” Renewable Energy, no. 33, pp. 1455–1460, 2008.

570 [7] A. Emami and P. Nougreh, “New approach on optimization in placement of wind turbines within
571 wind farm by genetic algorithms,” Renewable Energy, no. 35, pp. 169–178, 2010.
lP
572 [8] S. Mirjalili, S. M. Mirjalili, and A. Lewis, “Grey Wolf Optimizer,” Advances in Engineering
573 Software, vol. 69, pp. 46–61, Mar. 2014.

574 [9] K. V. Price, R. M. Storn, and J. A. Lampinen, The differential evolution algorithm. Berlin
575 Heidelberg: Springer, 2005.
rna

576 [10] R. Storn and K. Price, “Differential Evolution – A Simple and Efficient Heuristic for Global
577 Optimization over Continuous Spaces,” Journal of Global Optimization, no. 11, pp. 341–359,
578 1997.

579 [11] S.-R. Massan, A. I. Wagan, M. M. Shaikh, and M. S. Shah, “Application of Differential Evolution
580 Algorithm for Wind Turbine Micrositing,” Mehran University Research Journal Of Engineering
Jou

581 & Technology, vol. 36, no. 2, pp. 133–146, 2017.

582 [12] S.-U.-R. Massan, A. I. Wagan, M. M. Shaikh, and M. S. Shah, “Numerical data concerning wind
583 farm layout optimization using differential evolution algorithm at different wind speeds,” Data
584 in Brief, vol. 15, pp. 244–248, 2017.

585 [13] J. B. Lee, Clever Algorithms - Nature Inspired Programming Recipes. Melbourne: Jason Brown

31
 Journal Pre-proof

586 Lee, 2011.

587 [14] A. Kaveh and S. Mahjoubi, “Lion pride optimization algorithm: A meta-heuristic method for

of
588 global optimization problems,” Scientia Iranica, vol. 25, pp. 3113–3132, 2018.

589 [15] X.-S. Yang and S. Deb, “Eagle strategy using Levy walk and firefly algorithms for stochastic
590 optimization,” Nature Inspired Cooperative Strategies for Optimization (NICSO 2010), vol. 284,

pro
591 pp. 101–111, 2010.

592 [16] X.-S. Yang, “Firefly algorithm, stochastic test functions and design optimisation,” International
593 Journal of Bio-Inspired Computation, pp. 1–12, 2010.

594 [17] X.-S. Yang, Nature-Inspired Metaheuristic Algorithms, Second. Luniver press, 2010.

595 [18] S.-R. Massan, A. I. Wagan, M. M. Shaikh, and R. Abro, “Wind turbine micrositing by using the
596

597
re-
firefly algorithm,” Applied Soft Computing, vol. 27, pp. 450–456, Nov. 2015.

[19] S.-R. Massan, A. I. Wagan, and M. M. Shaikh, “A New Hybrid Metaheuristic Algorithm for
598 Wind Farm Micrositing,” Mehran University Research Journal Of Engineering & Technology,
599 vol. 36, no. 3, pp. 635–648, 2017.

600 [20] S. Mirjalili, A. H. Gandomi, S. Z. Mirjalili, S. Saremi, H. Faris, and S. M. Mirjalili, “Salp Swarm
lP
601 Algorithm: A bio-inspired optimizer for engineering design problems,” Advances in Engineering
602 Software, vol. 114, pp. 163–191, Dec. 2017.

603 [21] X.-S. Yang and S. Deb, “Multiobjective cuckoo search for design optimization,” Computers &
604 Operations Research, vol. 40, no. 6, pp. 1616–1624, 2013.
rna

605 [22] Ryanking, “OpenWindFarm,” 2015, 2015. [Online]. Available:

606 https://github.com/ryannking/OpenWindFarm. [Accessed: 25-Nov-2015].

607 [23] S.-R. Massan, A. I. Wagan, and M. M. Shaikh, “Power Optimization of Wind Turbines by the
608 Adjoint Method,” Sindh University Research Journal, vol. 48, no. 3, pp. 559–562, 2016.

609 [24] Wikipedia, “Rank of a Group,” 2016, 2016. [Online]. Available:

Jou

610 https://en.wikipedia.org/wiki/Rank_of_a_group.

611 [25] J. H. Holland, “Genetic Algorithms,” Scientific American, vol. 267, no. 1, pp. 66–72, 1992.

612 [26] N. O. Jensen, “A Note on Wind Generator Interaction,” RISO-M-2411, 1983.

613 [27] S.-R. Massan, A. I. Wagan, and M. M. Shaikh, “Numerical Data for wind turbine micrositing
614 inspired by human dynasties by use of the Dynastic Optimization Algorithm (DOA),” Data in
32
 Journal Pre-proof

615 Brief , (submitted), 2020.

616 [28] A. J. Memon and M. M. Shaikh, “Confidence bounds for energy conservation in electric motors:

of
617 An economical solution using statistical techniques,” Energy, vol. 109, pp. 592–601, Aug. 2016.

618 [29] M. M. Shaikh, A. J. Memon, and M. Hussain, “Data on electrical energy conservation using high
619 efficiency motors for the confidence bounds using statistical techniques,” Data in Brief, vol. 8,

pro
620 pp. 529–535, Sep. 2016.

621

re-
lP
rna
Jou

33
 Journal Pre-proof

*Declaration of Interest Statement

To the Editor

Applied Soft Computing Journal

of
www.elsevier.com

pro
Subject: Declaration of interest " A new metaheuristic optimization algorithm inspired by
human dynasties with an application to wind turbine micrositing problem"

Dear Sir,

I wish to publish my above named paper in your journal. I wish to state that,

1) We have received no funding of any kind for this paper from any source

re-
2) There is no conflict of interest involved in the publishing of this paper

3) This work is of academic nature

4) This paper has not been submitted anywhere presently

5) This work is original and novel

lP
6) This work has been accomplished by the efforts of the authors who take full responsibility of the
novelty and uniqueness of the endeavor

Regards,
rna

Shafiq-ur-Rehman Massan
Corresponding Author
[email protected]
Mobile: +92 333 234 432 2
Jou

Phone: +92 21 3530 4530

2B/1, 24th Street
Defence Housing Authority, Phase 5,
Karachi, Pakistan
*Highlights (for review) Journal Pre-proof

HIGHLIGHTS

 The proposed metaheuristic optimization algorithm is inspired from human dynasties

 Full application is made to the Wind Turbine Micrositing (WTM) problem.

of
 The mathematics, pseudocode, working, advantages and application are discussed
 The DOA reveals superior results as compared to the GA and DEA
 95% confidence bounds indicate that future forecasts are encouraging

pro
re-
lP
rna
Jou
 Journal Pre-proof

Author credit statement

The following authors have contributed to the papers named below [1] & [2],

[1] Massan, S.-R., Wagan, A. I., and Shaikh, M. M., “A new metaheuristic optimization

of
algorithm inspired by human dynasties with an application to the wind turbine micrositing
problem,” Applied Soft Computing, (submitted), 2020.

[2] Massan, S.-R., Wagan, A. I., and Shaikh, M. M., “Numerical Data for wind turbine

pro
micrositing inspired by human dynasties by use of the Dynastic Optimization Algorithm
(DOA),” Data in Brief , (submitted), 2020.

Authors,

1) Shafiq-ur-Rehman Massan

([email protected]) re-
a) Shaheed Zulfikar Ali Bhutto Institute of Science and Technology, Sindh, Pakistan.
b) Mohammad Ali Jinnah University, Karachi, Sindh, Pakistan.
lP
2) Asim Imdad Wagan

([email protected])

Mohammad Ali Jinnah University, Karachi, Sindh, Pakistan.

rna

3) Muhammad Mujtaba Shaikh

([email protected])

a) Department of Basic Sciences and Related Studies, Mehran University of Engineering and
Technology, Jamshoro, Pakistan.
b) Supply Chain and Operations Management Research Group, Mehran University of
Jou

Engineering and Technology, Jamshoro, Pakistan