1 Opposition-based gravitational search algorithm applied to economic

2 power dispatch problems consisting of thermal units with emission

3 constraints

6 Serdar ÖZYÖNa,*, Celal YAŞARa, Burhanettin DURMUŞa, Hasan TEMURTAŞb

a
9 Electrical and Electronics Engineering Department, Dumlupınar University, 43100, Kütahya, TURKEY
b
10 Computer Engineering Department, Dumlupınar University, 43100, Kütahya, TURKEY

11 [email protected], [email protected], [email protected],

12 [email protected]

13

14

15 *Corresponding author. Tel: +90 (274) 265 2031 / 4264; fax: +90 (274) 265 2066

16 E-mail address: [email protected] (S. Özyön)

17

18

19

20

21

22

23

24

25

26

27
28 Abstract

29 In this study, multi objective environmental economic power dispatch problem has been

30 converted into single objective optimization problem using weighted sum method

31 (WSM). For the solution of the converted problem gravitational search algorithm

32 (GSA), which is one of the latest algorithms, has been used. Also in order to increase

33 the performance of GSA opposite positioning quality has been added to the structure of

34 the algorithm (OGSA). The obtained results show that the proposed algorithm has

35 obtained better results and has provided a faster convergence. The 30 bus 6 generator

36 test system has been selected for application of OGSA. The transmission line losses

37 have been added to the problem by using B loss matrix. Optimum solutions of the

38 problem have been obtained for different weights (w) and the results have been

39 disputed.

40

41 Keywords: Environmental economic, Power dispatch, Gravitational search algorithm,

42 Opposition-based learning, Weighted sum method.

43

44 1. Introduction

45 Economic power dispatch problem is defined as the setting of active power outputs of

46 generation units, for the system load to be satisfy by the units under system constraints

47 and at minimum cost [1].

48

49 In general, power generation units use coal, petroleum and natural gas as fossil fuel.

50 And these fossil fuels cause atmospheric waste emission, composed of particles and

51 gases. These emitted waste gases include CO2, SO2 and NOx. These gases are harmful

2
52 for all living beings and they even cause global warming. Among these, SO2 emission

53 only depends on fuel depletion and therefore it's easier to perform mathematical

54 modeling. However, NOx emission depends on several factors like steam boiler

55 temperature and air mixture. Therefore it is extremely hard to perform modeling of NOx

56 gas emission. Besides, NOx emission is more hazardous than other contaminants [2].

57

58 In addition to the minimization of the fuel costs, emitted hazardous gases should also be

59 considered in economic dispatch problems. These multi objective power dispatch

60 problems considering environmental pollution as well are referred to as "environmental

61 economic power dispatch problems" [1,2].

62

63 The problem converts into a multi objective optimization problem when one would like

64 to minimize both fuel cost function and emission amount. These problems are solved in

65 the literature in different two methods. First one is to directly and the second one is to

66 modify, and then apply methods that solve these problems. One of the methods used for

67 modified is WSM [1].

68

69 Today, heuristic algorithms are used for solving complex problems like economic

70 power dispatch problems that are very hard or impossible to solve by numerical

71 methods. Major ones among these algorithms are genetic algorithm, ant colony

72 optimization, differential evolution algorithm, particle swarm optimization, harmony

73 search algorithm and gravitational search algorithm. In recent years, several methods

74 have been developed to improve the performance of heuristic algorithms. One of these

75 methods is adding opposition-based positioning feature to the used algorithm.

3
76 Experimental studies show the utilizing of opposition-based positioning has been

77 improved the convergence rate of optimization algorithms [3-5].

78

79 In the literature, solutions to several economic power dispatch problems have been

80 sought using various algorithms. Genetic algorithm (GA) [1,6,7], chaotic ant swarm

81 optimization algorithm (CASO) [8], bacteria foraging optimization methods (BFA) [9],

82 particle swarm optimization approach (PSO) [10,11], differential evolution and

83 modified differential evolution algorithms (DE, MDE) [12,13], analytic method [14],

84 linear programming [15], artificial bee colony optimization algorithm (ABC) [16],

85 charged system search algorithm [17] and gravitational search algorithm [18-20] can be

86 listed as some of the solution algorithms.

87

88 Recently, a novel heuristic search algorithm called gravitational search algorithm (GSA)

89 has been recommended in [21]. GSA has been confirmed to have higher performance in

90 solving various nonlinear functions, compared with some well-known search methods.

91 Therefore, the GSA algorithm has captured much attention. It has been applied to

92 various optimization problems such as the economic load dispatch problem [22],

93 forecasting of turbine heat rate [23], the digital filter design [24], and so on.

94

95 Additionally, there are modified variants of the GSA proposed in the literature for

96 enhancing the solution accuracy and the convergence rate. Zahiri [25] has presented an

97 fuzzy GSA for data mining. Nobahari et al. has proposed a non-dominated sorting GSA,

98 utilizes the non-dominated sorting concept to update the gravitational acceleration of the

4
 99 particles [26]. Li and Zhou have proposed improved GSA approach for parameters

100 identification of hydraulic turbine governing system [27].

101

102 In the literature, SO2 and NOx emissions are evaluated together or separately in various

103 studies. Only NOx emission was taken into account in this study. Existing multi-

104 objective economic power dispatch problem was converted into a single-objective

105 optimization problem by help of WSM and gravitational search algorithm (GSA) was

106 used for solving. This work aims to accelerate the convergence rate of GSA by utilizing

107 opposite numbers. The opposition-based gravitational search algorithm (OGSA) is

108 introduced to economic power dispatch problem. The experimental results show that the

109 proposed application of may be possible strategy to improve the performance of GSA.

110

111 2. Formulation of the Problem

112 Solution of the environmental economic power dispatch problem is obtained by

113 minimizing the objective function combined with WSM under system constraints.

114 Objective function of the environmental economic power dispatch problem to be

115 minimized combined with weighted sum method is as follows [1].

116 OF  w  Fn ( PG ,n )  (1  w)  E n ( PG ,n ) (1)
nNG nNG

117 In the equation, ($ / h) fuel cost is shown by Fn ( PG ,n ) and (ton / h) NOx emission

118 function is shown by En ( PG ,n ) .  represents scaling factor, w represents weight factor

119 varying as (0  w  1) and N G represents the set of all thermal generation units in the

120 system [3]. Here, w  1.0 value corresponds to the minimization of only fuel cost;

121 where w  0.0 value corresponds to the minimization of only NOx emission.

5
122 Fuel cost of the generation units in the system is taken as the second degree function of

123 the active power generation for each facility [1, 2].

124 Fn ( PG ,n )  an  bn PG ,n  cn PG2,n , ($ / h) (2)

125 NOx emission generated by each thermal facility was defined in terms of the output

126 power of the facility as follows [2].

127 En ( PG ,n )  d n  en .PG ,n  f n PG2,n  gn exp(hn PG ,n ), (ton / h) (3)

128 PG ,n unit in equations (2) and (3) is taken as MW. Power balance constraint in the lossy

129 system is taken as in equation (4)

130 P
nNG
G ,n  Pload  Ploss  0 (4)

131 Operation limit values of thermal generation units are given in equation (4).

,n  PG ,n  PG ,n , (n  N G )
PGmin max
132 (5)

133 Power losses of the system are computed by using B- matrix loss formula in (6) [6].

134 Ploss   P
nNG jNG
G ,n .Bnj .PG , j  B
nNG
0n .PG ,n  B00 (6)

135 Total fuel cost in the system FT ( PG ,n ) and total NOx emission ET ( PG ,n ) are calculated

136 using equations (7) and (8), respectively.

137 FT ( PG ,n )   F (P
nNG
n G ,n ), ($/h) (7)

138 ET ( PG ,n )   E (P
nNG
n G ,n ), (ton / h) (8)

139

140 3. Gravitational Search Algorithm

141 In physics, massive objects tend to accelerate towards each other. In Newton's law of

142 gravitation, each object attracts one another with a particular force that is "gravitational

6
143 force". GSA is one of the newest algorithms, inspired by Newtonian laws of gravity and

144 motion. In GSA, a number of agents referred to as masses are defined in order to find

145 the optimal solution by simulations of Newtonian laws of gravity and motion [21, 28].

146

147 In order to define GSA, let us assume a system of s sets where the position of ith set is

148 defined as follows:

149 X i  ( xi1,..., xid ,..., xin ), i  1,2,....., s (9)

150 Here, xid is the position of ith set in dth dimension, and n is the size of search area. Mass

151 of each agent is represented by its fitness and is calculated in terms of the fitness of

152 other individuals in the population [21, 28].

fiti (t )  worst (t )
153 qi (t )  (10)
best (t )  worst (t )

qi (t )
154 M i (t )  s
(11)
 q (t )
j 1
j

155 In these equations, M i (t ) and fiti (t ) represent the mass and the fitness value of the ith

156 agent at the time t, respectively. For a minimization problem, best (t ) and worst (t ) are

157 defined as follows [21, 28].

158 best (t )  min fit j (t ) (12)

j{1,....., s}

159 worst (t )  min fit j (t ) (13)

j{1,....., s}

160 When calculating the acceleration of an agent, total force applied on it by other agents is

161 first calculated depending on the law of gravity. Total force acting on subject agent is

162 defined in (14). Later, acceleration of the agent is calculated via (15) by using the law of

163 motion. As is seen in equation (16), acceleration value of the agent is added to its

7
164 velocity and a new velocity vector is obtained. Last, the next position of the agent is

165 determined according to (17) [21, 28].

M j (t ) M i (t )
166 Fi d (t )  
jkbest , j 1
rand jG(t )
Rij (t )  
( x dj (t )  xid (t )) (14)

Fi d (t ) M j (t )
167 aid (t ) 
M i (t )
 
jkbest , j 1
rand jG(t )
Rij (t )  
( x dj (t )  xid (t )) (15)

168 vid (t  1)  randi  vid (t )  aid (t ) (16)

169 xid (t  1)  xid (t )  vid (t  1) (17)

170 Here, rand i and rand j are two numbers that are randomly distributed in [0,1] interval.

171  is a small value (used for eliminating undefined results), Rij (t ) is the Euclidean

172 distance between agents i and j, and is defined as Rij (t )  X i (t ), X j (t ) 2 . kbest is a

173 sequence of the first K agents with the biggest mass and the best fitness value, which is

174 a function of time started in K 0 and is reduced in time. Here, K 0 is a set of s (total

175 agent number) and is linearly decreased to 1. In GSA, G (t ) represents gravitational

176 constant, G0 an initial value,  a constant, t the current iteration and tmax the final

177 iteration [21, 28]. GSA algorithm flowchart is displayed in Figure 1.

178 G(t )  G0 e B ( t /tmax ) (18)

179

180 Figure 1. GSA flowchart

181

182 4. Opposition-based Positioning

183 To come up with better solutions, evolutionary algorithms begin calculating process

184 with an initial population. They develop the individuals in the population by using

8
185 various development methods, in order to obtain the optimal solution. However, initial

186 population is generally made up of randomly generated individuals in the search space.

187 Yet, instead of randomly generated values, starting calculation with an initial population

188 where the individuals have better fitness values would speed up the development

189 process. In this sense, the concept of opposition-based learning was put forward.

190 According to this approach, opposition-based status of a number is possibly closer to a

191 solution, compared to a randomly generated number. Therefore, compared to its

192 opposition-based value, a number requires a smaller search space to converge to correct

193 solution [5, 29].

194

195 4.1. Opposition-based number definition

196 Let x be a real number defined in [a, b] interval. This number is defined as follows

197 according to opposition-based positioning theorem.

198 x  ab x (19)

199 This expression can be generalized for multi-dimensional sequences as in Chapter 4.2

200 [5, 29].

201

202 4.2. Opposition-based point definition

203 Let us define a point P  ( x1, x2 ,...., xd ) in a d-dimensional space. Let x1, x2 ,...., xd  R

204 and xi  [ai , bi ]i {1,2,...., d } . Opposition-based position of this point

205 P  ( x1, x2 ,...., xd ) is defined by the following components.

206 xi  ai  bi  xi (20)

9
207 An x point and opposition-based position x are shown for a one-dimensional space in

208 Figure 2. Here, [a, b] represents plane boundaries, where c represents the plane center

209 [5, 29].

210

211 Figure 2. Point and opposition-based position display for one-dimensional space.

212

213 4.3. Opposition-based optimization

214 Let us take a point defined as P  ( x1, x2 ,...., xd ) in d-dimensional search space. This

215 point can be resembled to the candidate solutions of the population in the solution of an

216 optimization problem [11]. By definition of the opposition-based positioned point, the

217 opposition-based position of this point would be P  ( x1, x2 ,...., xd ) . Then, when both

218 individuals are evaluated concerning the objective function, their fitness functions

219 would be f ( P) and f ( P) respectively. If f ( P)  f ( P) then P and P individuals

220 could switch places for a better solution. In this study, opposition-based position of each

221 randomly generated individual in the initial position has been determined for GSA

222 algorithm and individuals with a better fitness value have been switched with

223 individuals with a lower fitness value. Thus, converging speed of GSA has been aimed

224 to be increased by initializing search operation with individuals with a higher fitness

225 value. [5, 29].

226

227 5. Sample Problem Solution

228 IEEE test system with 6 generators and 30 buses has been selected as a sample problem

229 for a load Pload  283.4 MW . System values are obtained from source [5] and B loss

230 matrix has been given in Table 1. In order to show that opposition-based positioning

10
231 enhances the performance of the algorithm, the sample problem has been solved both by

232 GSA and OGSA.   1000 , GSA and OGSA parameter values G0  40 ,   20 ,

233   106 and agent number (s) is taken as 30 in the performed study.

234

235 Table 1. Values of the B loss matrix

236

237 150 iterations have been run for each weight value in a computer with AMD 64 X2 2.31

238 GHz processor and 4 GB RAM, with the program developed in MATLAB R2010 for

239 the solution of the test system.

240

241 Results obtained from the solution of the problem with GSA and OGSA are displayed

242 in tables 2 and 3.

243

244 As seen in Table 2, speed performance of GSA method varies between a minimum of

245 9.4913 seconds and a maximum of 10.7235 seconds for different w values, with an

246 average of 10.225 seconds. When Table 3, is examined it’s seen that speed performance

247 of OGSA method varies between a minimum of 7.9046 seconds and a maximum of

248 9.1005 seconds for different w values, with an average of 8.1999 seconds. OGSA

249 method is better than GSA in terms of speed.

250

251 Table 2. Results obtained by GSA for IEEE test system with 30 buses.

252

253 Table 3. Results obtained by OGSA for IEEE test system with 30 buses.

254

11
255 Minimum fuel cost ( w  1.0 ) and minimum emission ( w  0.0 ) values have been given

256 in Table 4, with other results in the literature.

257

258 Table 4. Results obtained in the literature by different methods for IEEE test system

259 with 30 buses [5]

260

261 Going through Table 4, it is seen that the results obtained by GSA and OGSA have

262 come close to the results in literature, they are even better than those.

263

264 The graphics showing the changes in total fuel cost and total NOx emission values,

265 obtained with the application of GSA and OGSA to the test system for w  1.0 and

266 w  0.0 , with regards to the iterations have been displayed in Figure 3 and 4,

267 respectively.

268

269 Figure 3. Change in total fuel cost with regard to the iterations

270

271 Figure 4. Change in total emission amount with regard to the iterations

272

273 It can be seen in Figure 3 that with regards to total fuel cost, OGSA converges to the

274 optimal value in greater speed compared to GSA. In Figure 4, a similar condition holds

275 for the convergence in total emission amount. Examining both figures, one can see that

276 OGSA reaches optimal value in approximately 35 iterations where GSA reaches in

277 approximately 65 iterations. Another positive effect of opposition-based positioning on

278 GSA has been revealed here.

12
279 The condition of OGSA application, in which w is increased from 0.0 to 1.0 with

280 intervals of 0.1 and the total fuel cost is decreased where total NOx emission amount is

281 increased, has been displayed in Figure 5.

282

283 Figure 5. Change in total NOx emission amount with regard to total fuel cost (OGSA)

284

285 When the value of w is taken as equal to zero (the fuel cost rates are ignored), the total

286 fuel cost rate and the total NOx emission amount become 605.9982 $/h and 0.220718

287 ton/h, respectively. In the same way when w is taken as equal to one (the NOx emission

288 rates are ignored), the total fuel cost rate and the total NOx emission amount become

289 646.20576 $/h and 0.194179 ton/h, respectively.

290

291 6. Conclusions

292 In the study, GSA and OGSA algorithms have been performed to the system. GSA and

293 OGSA have both studied to obtain the best result for all values of weight factor w . In

294 the process, weight factor has been altered in increments as 0.1 starting from w  0.0

295 and reaching w  1.0 . Results obtained by both algorithms have come close to the ones

296 in the literature. OGSA has been found to converge to the solution in a higher speed,

297 compared to GSA.

298

299 7. References

300 [1] Yaşar C, Özyön S. Solution to scalarized environmental economic power

301 dispatch problem by using genetic algorithm. Int J Elect Power Energy Sys

302 2012; 38: 54-62.

13
303 [2] Özyön S, Yaşar C, Temurtaş H. Solution of environmental economic power

304 dispatch problem in systems with limited energy supply thermal units. J

305 Eskişehir Osmangazi University Natural Sciences 2011; 24: 47-71.

306 [3] Rahnamayan S, Tizhoosh HR, Salama MMA. Opposition-based differential

307 evolution. IEEE Trans Evol Com 2008; 12: 64–79.

308 [4] Chatterjee A, Ghoshal SP, Mukherjee V. Solution of combined economic and

309 emission dispatch problems of power systems by an opposition-based harmony

310 search algorithm. Inter J Elec Power & Energy Sys 2012; 39: 9-20.

311 [5] Tizhoosh HR. Opposition-based learning: A new scheme for machine

312 intelligence. In: 2005 International Conference on Computational Intelligence

313 for Modeling Control and Automation; 28-30 November 2005; Vienna, Austria.

314 pp. 695–701.

315 [6] Abido MA. Multiobjective evolutionary algorithm for electric power dispatch

316 problem. IEEE Trans Evo Comp 2006; 10: 315-329.

317 [7] Güvenç U. Combined economic emission dispatch solution using genetic

318 algorithm based on similarity crossover. Sci Res and Ess 2010; 5: 2451-2456.

319 [8] Cai J, Ma X, Li Q, Li L, Peng H. A multi-objective chaotic ant swarm

320 optimization for environmental/economic dispatch. Int J Elect Power Energy Sys

321 2010; 32: 337-344.

322 [9] Panigrahi BK, Ravikumar PV, Sanjoy D, Swagatam D. Multiobjective fuzzy

323 dominance based bacterial foraging algorithm to solve economic emission

324 dispatch problem. Energy 2010; 35: 4761-4770.

325 [10] Zhang Y, Gong D, Ding Z. A bare-bones multi-objective particle swarm

14
326 optimization algorithm for environmental/economic dispatch. Inf Sci 2012; 192:

327 213-227.

328 [11] Abido MA. Multiobjective particle swarm optimization for environmental

329 economic dispatch problem. Elec Power Sys Res 2009; 79: 1105-1113.

330 [12] Yaşar C, Temurtaş H, Özyön S. Differential evolution algorithm applied to

331 environmental economic power dispatch problems consisting of thermal units.

332 In: ELECO 2010 Electrical - Electronics and Computer Engineering

333 Symposium; 2-5 December 2010; Bursa, Turkey. pp. 108-112.

334 [13] Wu LH, Wang YN, Yuan XF, Zhou SW. Environmental/economic power

335 dispatch problem using multi-objective differential evolution algorithm. Elec

336 Power Sys Res 2010; 80: 1171-1181.

337 [14] Palanichamy C, Babu NS. Analytical solution for combined economic and

338 emissions dispatch. Elec Power Sys Res 2008; 78: 1129-1137.

339 [15] Farag A, Al-baiyat S, Cheng TC. Economic load dispatch multiobjective

340 optimization procedures using linear programming techniques. IEEE Trans

341 Power Sys 1995; 10: 731-738.

342 [16] Özyön S, Yaşar C, Özcan G, Temurtaş H. An artificial bee colony algorithm

343 (ABC) approach to environmental economic power dispatch problems. In: FEEB

344 2011 National Electrical-Electronics Computer Symposium; 5-7 October 2011;

345 Elazığ, Turkey. pp. 222-228.

346 [17] Özyön S, Temurtaş H, Durmuş B, Kuvat G. Charged system search algorithm

347 for emission constrained economic power dispatch problem. Energy 2012; 46:

348 420-430.

15
349 [18] Güvenç U, Sönmez Y, Duman S, Yörükeren N. Combined economic and

350 emission dispatch solution using gravitational search algorithm. Sci Ira 2012; 19:

351 1754-1762.

352 [19] Duman S, Güvenç U, Yörükeren N. Gravitational search algorithm for economic

353 dispatch with valve-point effects. Int Rev Elect Eng (IREE) 2010; 5: 2890-2895.

354 [20] Duman S, Güvenç U, Sönmez Y, Yörükeren N. Optimal power flow using

355 gravitational search algorithm. Energy Con & Man 2012; 59: 86-95.

356 [21] Rashedi E, Nezamabadi-pour H, Saryazdi S. GSA: a gravitational search

357 algorithm. Inf Sci 2009; 179: 2232–2248.

358 [22] Swain RK, Sahu NC, Hota PK. Gravitational search algorithm for optimal

359 economic dispatch. Proc Tech 2012; 6: 411-419.

360 [23] Zhang W, Niu P, Li G, Li P. Forecasting of turbine heat rate with online least

361 squares support vector machine based on gravitational search algorithm.

362 Knowledge-Based Sys 2013; 39: 34-44.

363 [24] Rashedi E, Nezamabadi-pour H, Saryazdi S. Filter modeling using gravitational

364 search algorithm. Eng App of Artif Intel 2011; 24: 117-122.

365 [25] Zahiri SH. Fuzzy gravitational search algorithm an approach for data mining.

366 Iranian J Fuzzy Sys 2012; 9: 21-37.

367 [26] Nobahari H, Nikusokhan M, Siarry P. Non-dominated sorting gravitational

368 search algorithm. In: 2011 International Conference on Swarm Intelligence; 14-

369 15 June 2011; Cergy, France. pp. 1-10.

370 [27] Li C, Zhou J. Parameters identification of hydraulic turbine governing system

16
371 using improved gravitational search algorithm. Energy Con & Man 2011; 52:

372 374-381.

373 [28] Rashedi E, Nezamabadi-pour H, Saryazdi S. BGSA: Binary gravitational search

374 algorithm. Nat Comp 2010; 9: 727-745.

375 [29] Ergezer M, Simon D. Oppositional biogeography-based optimization for

376 combinatorial problems, IEEE 2011 Congress on Evolutionary Computation; 5-8

377 June 2011; New Orleans, USA: IEEE. pp. 1496-1503.

378

379

380

381

382

383

384

385

386

387

388

389

390

391

17
392 Figure legends

Start

Enter system information and GSA

parameters.

Generate initial population.

Evaluate the fitness value of each

E
agent.

Update the best and the

worst agent of the
population and G.

Calculate M and a for each agent.

Update velocities and positions.

H Is the stopping
criterion satisfied?

Return to the
best solution.
393

394 Figure 1. GSA flowchart

395

a x c x b
396

397 Figure 2. Point and opposition-based position display for one-dimensional space.

18
 635

GSA
630 OGSA
Total fuel cost ($ / h)

625

620

615

610

605
0 50 100 150
Iteration number

398

399 Figure 3. Change in total fuel cost with regard to the iterations

400

215

GSA
OGSA
210
NOx Emission amount (kg / h)

205

200

195

190
0 50 100 150

Iteration number

401

402 Figure 4. Change in total emission amount with regard to the iterations

19
 0,2235
Total NOx emission amount ( ton / h )
0,2185

0,2135

0,2085

0,2035

0,1985

0,1935

Total fuel cost ( $ / h )

403

404 Figure 5. Change in total NOx emission amount with regard to total fuel cost (OGSA)

405

406

407 Table legends

408

409 Table 1. Values of the B loss matrix

 0.1382 0.0299 0.0044 0.0022 0.0010 0.0008

 0.0299 0.0487 0.0025 0.0004 0.0016 0.0041 

 0.0044 0.0025 0.0182 0.0070 0.0066 0.0066
 B   
 0.0022 0.0004 0.0070 0.0137 0.0050 0.0033 
 0.0010 0.0016 0.0066 0.0050 0.0109 0.0005 
 
 0.0008 0.0041 0.0066 0.0033 0.0005 0.0244 
410  B0    0.0107 0.0060 0.0017 0.0009 0.0002 0.0030
B00  0.00098573

20
411 Table 2. Results obtained by GSA for IEEE test system with 30 buses.

Total
Total fuel NOx Line
Time
w PG,1 PG,2 PG,3 PG,4 PG,5 PG,6 cost emission losses
(s)
($/h) amount ( pu )
( ton / h )
1.0 0.120969 0.286312 0.583557 0.992854 0.523970 0.351899 605.99836 0.220729 0.025562 9.4913

0.9 0.138287 0.297892 0.582043 0.946813 0.531236 0.362879 606.20055 0.216922 0.025150 10.0910

0.8 0.157122 0.310362 0.579907 0.899650 0.537246 0.374603 606.84123 0.213308 0.024890 9.9249

0.7 0.177726 0.323857 0.577181 0.850791 0.542056 0.387194 607.99038 0.209873 0.024805 10.0176

0.6 0.200407 0.338531 0.573889 0.799614 0.545706 0.400787 609.74783 0.206620 0.024933 10.1038

0.5 0.225543 0.354556 0.570053 0.745398 0.548212 0.415565 612.25279 0.203570 0.025327 10.1180

0.4 0.253613 0.372146 0.565695 0.687301 0.549599 0.431706 615.69721 0.200764 0.026060 10.5233

0.3 0.285232 0.391541 0.560856 0.624276 0.549881 0.449450 620.34851 0.198272 0.027236 10.4186

0.2 0.321206 0.413028 0.555602 0.555008 0.549081 0.469083 626.58303 0.196209 0.029007 10.5834

0.1 0.362612 0.436942 0.550052 0.477791 0.547242 0.490954 634.93974 0.194751 0.031594 10.4842

0.0 0.410925 0.463668 0.544419 0.390374 0.544459 0.515485 646.20700 0.194179 0.035330 10.7235
412

413

21
414 Table 3. Results obtained by OGSA for IEEE test system with 30 buses.

Total
Total fuel NOx Line
Time
w PG,1 PG,2 PG,3 PG,4 PG,5 PG,6 cost emission losses
(s)
($/h) amount ( pu )
( ton / h )
1.0 0.121048 0.286208 0.583635 0.992697 0.523973 0.351995 605.99815 0.220718 0.025557 7.9785

0.9 0.138219 0.297970 0.582325 0.946333 0.531611 0.362680 606.20279 0.216901 0.025139 7.9449

0.8 0.157071 0.310360 0.579839 0.899608 0.537240 0.374774 606.84182 0.213305 0.024892 7.9742

0.7 0.177935 0.323526 0.577122 0.850801 0.542166 0.387251 607.99058 0.209872 0.024802 8.0281

0.6 0.200398 0.338440 0.573920 0.799549 0.545863 0.400759 609.74780 0.206620 0.024930 9.1005

0.5 0.225578 0.354596 0.569991 0.745241 0.548349 0.415573 612.25942 0.203563 0.025328 8.1917

0.4 0.253687 0.372078 0.565575 0.687341 0.549726 0.431654 615.69594 0.200765 0.026062 8.3645

0.3 0.285149 0.391683 0.560776 0.624210 0.550119 0.449298 620.34883 0.198272 0.027235 8.4802

0.2 0.321251 0.413077 0.555586 0.555006 0.549048 0.469041 626.58567 0.196208 0.029010 8.0675

0.1 0.362662 0.436899 0.550032 0.477805 0.547261 0.490935 634.93972 0.194751 0.031596 8.1648

0.0 0.410929 0.463664 0.544417 0.390394 0.544443 0.515483 646.20576 0.194179 0.035330 7.9046

22
415 Table 4. Results obtained in the literature by different methods for IEEE test system

416 with 30 buses [5]

Minimum total NOx

Minimum total fuel cost
Method emission amount
( w = 1.0 ) ( $ / h )
( w = 0.0 ) ( ton / h )
GA [2] 607.7800 0.1942
NSBF [4] 607.5901 0.1944
FSBF [4] 607.5081 0.1942
PSO [6] 607.7800 0.1942
MDE [7] 606.4160 0.1942
GSA 605.9983 0.1941
OGSA 605.9981 0.1941
417

23