Inter national J our nal of Electrical and Computer Engineering (IJECE) V ol. 10, No. 2, April 2020, pp. 1187 1199 ISSN: 2088-8708, DOI: 10.11591/ijece.v10i2.pp1187-1199 r 1187 Antlion optimization algorithm f or optimal non-smooth economic load dispatch Thanh Pham V an 1 , V ´ acla v Sn ´ a ˇ sel 2 , Thang T rung Nguy en 3 1 F aculty of Electrical Engineering and Computer Science, T echnical Uni v ersity of Ostra v a, Czech Republic 1 European Cooperation Center , T on Duc Thang Uni v ersity , V iet Nam 2 T echnical Uni v ersity of Ostra v a, Czech Republic 3 Po wer System Optimization Research Group, F aculty of Electrical and Electronics Engineering, T on Duc Thang Uni v ersity , V iet Nam Article Inf o Article history: Recei v ed May 26, 2019 Re vised Oct 5, 2019 Accepted Oct 11, 2019 K eyw ords: Antlion optimization Multi fuel sources Ramp rate Spinning reserv e V alv e point loading ef fects ABSTRA CT This paper presents applications of Antlion optimization algorithm (ALO) for han- dling optimal economic load dispatch (OELD) problems. Electricity generation cost minimization by controlling po wer output of all a v ailable generating units is a major goal of the problem. ALO is a metaheuristic algorithm based on the hunting process of Antlions. The ef fect of ALO is in v estig ated by solving a 10-unit system. Each studied case has dif ferent objecti v e function and comple x le v el of restraints. Three test cases are emplo yed and arranged according to the comple x le v el in which the first one only considers multi fuel sources while the second case is more complicated by taking v alv e point loading ef fects into account. And, the third case is the highest challenge to ALO since the v alv e ef fects together with ramp rate limits, prohibited operating zones and spinning reserv e constraints are tak en into consideration. The comparisons of the result obtained by ALO and other ones indicate the ALO algorithm is more potential than most methods on the solution, the stabilization, and the con v er gence v elocity . Therefore, the ALO method is an ef fecti v e and promising tool for systems with multi fuel sources and considering complicated constraints. Copyright c 2020 Insitute of Advanced Engineeering and Science . All rights r eserved. Corresponding A uthor: Thang T rung Nguyen, Po wer System Optimization Research Group, F aculty of Electrical and Electronics Engineering, T on Duc Thang Uni v ersity , 19 Nguyen Huu Tho street, T an Phong w ard, District 7, Ho Chi Minh City , V iet Nam. Email: nguyentrungthang@tdtu.edu.vn 1. INTR ODUCTION Minimizing electricity generation fuel cost in thermal po wer plants (TPPs) is e xtremely import ant because it accounts for a high rate of total electricity generation cost. So, the OELD problem has been widely applied for this purpose. So f ar solutions which ha v e been just achie v ed by the OELD problem is to decide the po wer output of each thermal generating unit (TGU) so that the electricity generation fuel cost can decrease as much as possible. In addition, the OELD problem also tak es man y constraints into account. The constraints are po wer balance, spinning reserv e, po wer output limits of generators, prohibited operating zones, and ramp rate limits. Furthermore, fuel consuming characteristics of TGU such as multi fuel sources and v alv e point loading ef fects are also considered as main issues in the OELD problem. The OELD problem has attracted man y researchers because of its importance in using fuel for the TPPs reasonably . T w o main method groups ha v e used by man y authors including traditional numerical methods and modern methods based on artificial intelligence. J ournal homepage: http://ijece .iaescor e .com/inde x.php/IJECE Evaluation Warning : The document was created with Spire.PDF for Python.
1188 r ISSN: 2088-8708 T raditional numerical methods ha v e been applied to handle the OELD problem such as Lagrangian relaxation (LR) method [1], Linear programming techniques (LPT) [2], F ast Ne wton raphson (FNR) method [3]. Among the methods, LR is one of the earliest methods which has been applied to systems with a quadratic fuel cost function. The constraint s of the problem were rather simple such as po wer balance considering trans- mission losses, v oltage limitations, and generation limits. This method w as also tested on the 10-unit system and achie v ed results also met the requirements of technology at that time. The LPT method w as combination of the Lagrangian method and linear program (LP) method in which duty of the LP method w as to linearize non-linear functions and the Lagrangian method w as used as usual. Therefore, when there were man y non- linear constraints, this method w ould f ace errors due to linearization. In general, all of the traditional numerical methods only ha v e considered basic constraints of the OELD problem. Furthermore, these methods had to tak e the partial deri v ati v e. Consequently , traditional numerical methods will ha v e some restrictions if the y handle the OELD problem with comple x constraints. Unlik e traditional numerical methods abo v e, the modern methods ha v e been proposed to handle the OELD problem more successfully including ANN-based methods (ANN: artificial neural netw ork) and metaheuristic-based methods. ANN-based methods are comprised of Hopfield neural netw ork (HNN) [4], Adapti v e Hopfield neural netw orks (AHNN) [5], Augmented Lagrange Hopfield netw ork (ALHN) [6] and Enhanced Augmented Lagrange Hopfield netw ork (EALHN) [7]. Metaheuristic-based methods ha v e been widely and more successfully handling OELD problem. Dif ferential e v olution algorithm (DEA) [8], Quantum Ev olutionary Algorithm (QEA) [9], Hybrid inte ger coded dif ferential e v olution dynamic programming (HICDE-DP) [10], Impro v ed dif ferential e v olution algorithm (IDEA) [11], Colonial competi- ti v e dif ferential e v olution (CCDE) [12], Stud dif ferential e v olution (SDE) [13] and Hybrid dif ferential e v o- lution and Lagrange theory (HDE-LH) [14]. Real-coded genetic algorithm (RCGA) and Impro v ed RCGA (IRCGA) [17]. Hybrid real coded genetic algorithm (HRCGA) [15], P article sw arm optimization (PSO) [16], Modified particle sw arm optimization (MPSO) [18], Quantum-inspired particle sw arm optimization (QIPSO) [19], Distrib uted sobol PSO and T ab u Search algorithm (TSA) (DSPSO-TSA) [20], Fuzzy and self adapti v e particle sw arm optimization (FSAPSO) [21], -P article sw arm optimization ( -PSO) [22], Cuck oo search algorithm (CSA) [23], [24], Impro v ed CSA (ICSA) [25], Modified CSA (MCSA) [26], Kill herd algorithm (KHA) [27], Impro v ed KHA (IKHA) [28], Artificial immune systems (AIS) [29], Biogeograph y- based optim ization (BBO) [30], Chaotic firefly algorithm (CF A) [31], Gre y w olf optimizer (GW O) [32], [33], Crisscross optimization (CO) [34], Exchange mark et algorithm (EMA) [35], ALO [36], [37], Impro v ed fire- fly algorithm (IF A) [38], Whale Optimization Algorithm (W O A) [39], Cro w search algorithm (CrSA) [40]. Among the DEA method and its dif ferent v ersions, SDE [13] is the best method. This method w as created by using stud crosso v er operator with intent to restrict the search around lo w quality solutions. The most complicated conditions that this method has solv ed include multi fuel sources and v alv e point loading ef fects. The results sho w that the SDE method has pro vided the highest quality results compared to all other methods including the GA and TSA methods in [20] and HDE-LH in [14]; ho we v er , more complicated constraints lik e ramp rate, spinning reserv e, and prohibited operating zones were not tak en into account for challenging the method. The traditional GA method w as dif ficult to solv e OELD problem, b ut its v ariants ha v e been applied to this problem usefully . IRCGA [17] w as the strongest method in GA f amily . In this method, an ef ficient real-coded genetic algorithm (RCGA) with arithmetic-a v erage-bound crosso v er and w a v elet mutation w as presented. The solv ed system by method consists of 10 TGUs with v alv e point loading ef fects, multi fuel sources, ramp rate limits, prohibited operating zones, and spi nning reserv e. It has pro v e n to be the m ost ef fec- ti v e when comparing to other methods including PSO, ne w PSO, DEA, an impro v ed genetic algorithm (IGA). DSPSO–TSA [20] is better than se v eral methods such as TSA, GA, PSO, and other PSO v ariants b ut it w as only considering multi fuel sources and v alv e point loading ef fects b ut po wer loss constraints, prohibited oper - ating zones as well as other complicated constraints were not consist of. The IF A method in [38] seems to be the best method since it w as applied for solving systems with the most complicated constraints including both constraints considered in mentioned w ork and all const raints in transmission po wer netw orks. All t est cases could demonstrate the outstanding performance of the method. As a ne w approach method, the ALO method w as introduced the first time by Mirjalili in 2015 [41]. Unlik e the other algorithms, ALO has tw o population sets are an Ant colon y and an Antlion colon y . According to paper [41], the ALO method has handled se v eral mathematical functions and some engineering problems such as three classical engineering problems including three-bar truss design, cantile v er beam design, and gear train design. The ALO method also has been proposed for handling the OELD problem. F or e xample, in [36] the OELD problem has handled with simple constraints Int J Elec & Comp Eng, V ol. 10, No. 2, April 2020 : 1187 1199 Evaluation Warning : The document was created with Spire.PDF for Python.
Int J Elec & Comp Eng ISSN: 2088-8708 r 1189 such as po wer balance and generator limits constraint, or in [37] the method has tested on small-scale po wer systems considering v alv e point ef fects. The ALO method has sho wn its potential search including se v eral systems of the OELD problem as in [36], [37]. Ho we v er , the comple x le v el of considered systems w as not lar ge a n d complicated enough to decide its performance. So, in order to clarify further for the ef ficienc y of ALO need to be more research. In this paper , the ALO method has been applied for handling the OELD problem with the most con- straints and dif ferent fuel consuming characteristics. The set of constraints is po wer balance, spinning reserv e, po wer output limits of the TGU, prohibited operating zones and ramp rate limits. Fuel consuming characteris- tics are directly related to objecti v e functions such as piece wise quadratic functions and non-con v e x piece wise function. The method has been tested on three study cases and obtained results ha v e been compared to other methods for in v estig ation of the ALO method. 2. PR OBLEM FORMULA TION 2.1. Major objecti v e of the pr oblem In operation process of TPPs using fossil fuels, electricity generation cost is required to be optim ized. It can be mathematically formulated by: F = n X s =1 F s ( P s ) (1) where n is number of TGUs; F s ( P s ) is the fuel cost function of the s th TGU. When the system has one fuel type, the fuel cost function of the TGU can be presented in a s ingle quadratic form as follo ws: F s ( P s ) = s P 2 s + s P s + s (2) where s , s , s are the cost coef ficients of the s th TGU; and P s is po wer output of the s th TGU. In the case of the multi fuel sources, the fuel cost function of each generator should be represented by a piece wise quadratic function as sho wn in (3). Ho we v er , with the case of v alv e ef fect s, the cost function is more complicated as gi v en in (4): F s ( P s ) = 8 > > > < > > > : s 1 P 2 s + s 1 P s + s 1 ; fuel 1 s 2 P 2 s + s 2 P s + s 2 ; fuel 2 ::: sm P 2 s + sm P s + sm ; fuel m (3) F s ( P s ) = 8 > > > < > > > : s 1 P 2 s + s 1 P s + s 1 + j s 1 x sin ( s 1 ( P s;min P s )) j ; fuel 1 s 2 P 2 s + s 2 P s + s 2 + j s 2 x sin ( s 2 ( P s;min P s )) j ; fuel 2 ::: sm P 2 s + sm P s + sm + j sm x sin ( sm ( P s;min P s )) j ; fuel m : (4) where sm , sm , sm , sm , sm are the cost coef ficients of the s th TGU; m is number of the fuel types; and P s;min is the minimum po wer output of the s th TGU. 2.2. Constraints of po wer system and generator 2.2.1. Real po wer balance The total po wer generation should meet the total load demand P demand plus transmission losses P l oss as the follo wing rule: n X s =1 P s = P demand + P l oss (5) where P l oss is calculated by using Kron’ s formula belo w: P l oss = n X s =1 n X h =1 P s B sh P h + n X s =1 B 0 s P s + B 00 (6) where B sh , B 0 s , and B 00 are loss coef ficients. Antlion optimization algorithm for optimal non-smooth... (Thanh Pham V an) Evaluation Warning : The document was created with Spire.PDF for Python.
1190 r ISSN: 2088-8708 2.2.2. Spinning r eser v e constraint All TGUs are required that total acti v e po wer reserv e of them should be more than or equal to the lar gest generating unit. The constraint requires total acti v e po wer reserv e of al l units P r s must be at least equal to the po wer system requirement P pr . n X s =1 P r s P pr (7) 2.2.3. Generating capacity constraint The po wer output of each generator must not e xceed its operating limits des cribed by the follo wing rule: P s;min P s P s;max ; for s = 1, 2, ... n (8) where P s;max is the highest acceptable w orking po wer of the s th TGU. 2.2.4. Ramp rate constraint Because each TGU cannot change its po wer output with a high step compared to its pre vious genera- tion. Thus, tw o major conditions are added as the follo wing inequalities: P s P 0 s P r u s for the case of increasing po wer (9) P 0 s P s P r d s for the case of decreasing po wer (10) where P 0 s is initial po wer from the pre vious operating hour of generating unit; P r u s and P r d s are ramp up limit and ramp do wn limit of the s th TGU. 2.2.5. Pr ohibited operating zones Due to engineering reason that generating units must a v oid operating in se v eral operating zones as sho wn in the follo wing model: P min s;h P s P max s;h (11) where P min s;h and P max s;h are the minimum and maximum po wer output of the s th TGU in the h th prohibited operating zone. 3. ANTLION OPTIMIZA TION ALGORITHM Initialization: In initialization of ALO algorithm, the population of Antlions is randomly produced within the upper and lo wer limitations as the follo wing model: ALO d = C V min + r and ( C V max C V min ) ; d = 1,..., N pop (12) where N pop is the population size, and C V max and C V min are maximum and minimum limitations of control v ariables. Random walk of Ant: The mo v ement direction of Ants does not follo w an y rules and it is also a random w alk as described in the follo wing model: R W C I = [ 0 ; 1 X C I =1 (2 C I 1) ; 2 X C I =1 (2 C I 1) ; 3 X C I =1 (2 C I 1) ; : : : ; G max X C I =1 (2 C I 1) ] (13) where CI is an ordinal number of the current iteration; G max is the maximum number of iterations; C I is considered as a mo ving f actor; and calculated by: C I = ( 1 if 0 : 5 0 otherwise (14) Int J Elec & Comp Eng, V ol. 10, No. 2, April 2020 : 1187 1199 Evaluation Warning : The document was created with Spire.PDF for Python.
Int J Elec & Comp Eng ISSN: 2088-8708 r 1191 The r estricted space of Ant: The acti v e radius of the j th Ant w ould be more and more decreased adapti v ely when the current number of iterations is increased. F or mathematically modeling this beha vior , the follo wing equations are used: c j ;C I = C V min j and d j ;C I = C V max j (15) where c j ;C I and d j ;C I are do wn and up limitations in the acti v e range of the j th Ant at iteration CI; and is a f actor obtained by: = 10 C I G max (16) where is a constant defined based on the current iteration CI ( = 2 when CI > 0.1* G max , = 3 when CI > 0.5* G max , = 4 when CI > 0.75* G max , = 5 when CI > 0.9* G max , = 6 when CI > 0.95* G max ) Sliding Ant to ward Antlion: The range of acti vity of Ant is af fected by beha vior of shooting sands of Antlion. This made Ant sliding to the bottom of the trap where the massi v e ja w w as w aiting to catch pre y . T o describe the assumption, the tw o follo wing equations are necessary: X min j ;C I = ALO j ;C I + c j ;C I and X max j ;C I = ALO j ;C I + d j ;C I (17) where ALO j ;C I is the position of the j th Antlion at the C I th iteration; X max j ;C I and X min j ;C I are corresponding to ne wly updated upper and lo wer limits of control v ariables included in the position of Antlions. The mo v ement e v ery Ant: The mo v ement of ants is corresponding to the determination of search zones since random w alk in (13) only produces an updated st ep size that is not related to ne w solutions. The random w alk position of ant can be updated by the follo wing model: X j ;C I = ( R W C I R W min ) x ( X max j ;C I X min j ;C I ) R W max R W min + X min j ;C I (18) where R W max and R W min are the minimum and maximum v alues of R W C I respecti v ely . As a result, the real position of Ant is updated by using tw o random w alks around X j ;C I and the current best solution. The purpose is to use information e xchange between tw o other positions. The real position is obtained by: Ant j ;C I = X R W j ;C I + Gbest R W C I 2 (19) where X R W j ;C I is a ne w solution around X j ;C I by using random w alk; Gbest R W C I is a ne w solution nearby the best current solution Gbest C I . The phase of catching pr ey: In the process, the assumption is that the action of catching pre y happens when Ants goes inside sand. The follo wing equation is proposed in this re g ard: Antl ion j ;C I = Ant j ;C I if F F ( Ant j ;C I ) < F F ( Antl ion j ;C I ) (20) where F F ( Ant j ;C I ) and F F ( Antl ion j ;C I ) are the fitness function of Ant j ;C I and Antl ion j ;C I respecti v ely . All steps ALO method has been summarized as Figure 1. 4. IMPLEMENT A TION OF THE ALO ALGORITHM FOR OELD PR OBLEMS 4.1. V ariables of each indi vidual of the algorithm The position of each Antlion includes all control v ariables and is i nitialized within limits as the fol- lo wing model: X d = C V min + r and ( C V max C V min ) ; d = 1 ; :::; N pop (21) where C V min = f P 1 ;min ; :::; P n 1 ;min g and C V max = f P 1 ;max ; :::; P n 1 ;max g (22) As a result, the po wer output P n;d is obtained by equation (23) follo wing: P n;d = P demand + P l oss n 1 X i =1 P i;d (23) Antlion optimization algorithm for optimal non-smooth... (Thanh Pham V an) Evaluation Warning : The document was created with Spire.PDF for Python.
1192 r ISSN: 2088-8708 Start Initializing the population of Antlions by using (12) Calculating fitness function for all Antlions Determine the best Antlions and set CI = 1 Determine R W C I using (13) Calculate restrict space for Ants using (15) Determine lo wer and uper limits for position of Ants using (17) Find ne w position of Ants using (18) and (19) Ev aluate such ne w position by calculate fitness function Compare Antlion and ne w Ant to update ne w Antlion by using (20) Determine the best Antlion CI = G max CI = CI + 1 no Stop yes Figure 1. The flo wchart of ALO algorithm 4.2. Punishment of dependent v ariable violations In process of optimization, as P n;d is outside upper and lo wer bounds, it is penalized and determined by: C ost P un = 8 > < > : P n;d P n;max if P n;d > P n;max P n;min P n;d if P n;d < P n;min 0 if P n;min P n;d P n;max (24) 4.3. Compatible function The compatible function is added to the product of the square of punishment v alue and a punishment f actor ( F a ), as follo wing equation: C F d = n X i =1 F i ( P i ) + F a ( C ost P un ) 2 (25) 5. NUMERICAL RESUL TS The ef ficienc y proposed method is judged in this section. The ALO algorithm has been tested on the 10-unit system with constraints of the po wer net w ork and the generators, and dif ferent fuel consuming characteristics of thermal units. The detail is as follo ws: (a) Case 1: 10-unit po wer system using multi fuel sources and without v alv e point loading ef fects. (b) Case 2: 10-unit po wer system using multi fuel sources and v alv e point loading ef fects. Int J Elec & Comp Eng, V ol. 10, No. 2, April 2020 : 1187 1199 Evaluation Warning : The document was created with Spire.PDF for Python.
Int J Elec & Comp Eng ISSN: 2088-8708 r 1193 (c) Case 3: 10-unit po wer system using multi fuel sources, v alv e point loading ef fects and complicated constraints such as spinning reserv e, prohibited operating zones and ramp rate limits. Data which are used for the three cases is tak en from [42]. In addition, The ALO algorithm has been coded in Matlab platform and personal computer with processor 2.0 GHz and Ram of 2.0 Gb . 5.1. In v estigating impact of contr ol parameters on obtained r esults Three cases abo v e of the OELD problem ha v e been e x ecuted by the ALO algori thm to i n v est ig ate the impact of dif ferent v alues of cont rol parameters in the ef fecti v eness, rob ustness, and stability of the search process of ALO. P arameters ha v e been used in the in v estig ation are population size ( N pop ) and the number of iterations ( G max ). 5.1.1. Case 1: 10-unit po wer system using multi fuel sour ces and without v alv e point loading effects There are four studied subcases with four load cases from 2,400 to 2,700 MW with a change of 100 MW . The obtained results with respect to the minimum fuel cost for 100 trial runs with dif ferent cases of N pop and G max ha v e been reported in T able 1. Experimentation has been di vided into tw o parts. In the first part, the population size has been k ept at N pop = 10 and the number of iterations has been changed as the right first column of t he T able 1. While the population size has been k ept at N pop = 20 and the number of iterations as the T able 1 in the second part. Observing the table can see that the same v alue of N pop , when G max rises, there is a corresponding decline in the minimum fuel cos t. When the minimum fuel cost equals 481.7226 $ =h in both of tw o parts then it is impossible to decrease although G max still increases. In the first part at the first six ro ws of T able 1, the best c osts of subcase 1.1, subcase 1.3 and subcase 1.4 are respecti v ely 481.7426 $/h, 574.3808 $/h, 623.8092 $/h corresponding with N pop = 10, G max = 250. P articularly , subcase 1.2 get the best cost at N pop = 10 and G max = 200 with v alue of 526.2388 $/h. In the second part at the last four ro ws of T able 1 all four subcases reach the best c o s t at G max = 150. This point out that when the population size is set to higher v alue, the number of iterations can be set to lo wer v alue b ut the best optimal solution can be found. T able 1. The lo west cost ($/h) obtained from 100 runs for dif ferent v alues of N pop and G max Load of Load of Load of Load of N pop G max 2,400 MW 2,500 MW 2,600 MW 2,700 MW 491.2418 533.6331 579.9704 632.5862 10 50 483.0403 526.5162 574.5738 624.9104 10 100 481.7637 526.2820 574.3815 623.8695 10 150 481.7424 526.2388 574.3852 623.8103 10 200 481.7226 526.2388 574.3808 623.8093 10 250 481.7226 526.2388 574.3808 623.8092 10 300 483.5568 526.8289 576.4240 629.4534 20 50 481.7424 526.2494 574.5813 623.8402 20 100 481.7226 526.2388 574.3808 623.8092 20 150 481.7226 526.2388 574.3808 623.8092 20 200 The results of the distrib ution of the fuel costs for subcase 1.4 o v er 100 trials are sho wn in Figure 2. The results of the tests sho w that ALO can find the best optimal solution for dif ferent setting of control parameters and the de viation among obtained minimums is v ery small. Thus, ALO is stable and ef fecti v e for the first case with four dif ferent loads. 5.1.2. Case 2: 10-unit po wer system using multi fuel sour ces and v alv e effects The second study case only considers the load demand of 2,700 MW . Meantime, N pop has bee n k ept at the v alue of 20 b ut G max has been adj u s ted within 8 v alues from 50 to 400 with a small change of 50. Numbers yielded from the test including minimum cost, a v erage cost, maximum c ost are presented in T able 2. As sho wn in T able 2, once G max decrease, the fuel cost function will decrease in the first v e ro ws. Ho we v er , ro w 7 indicates the fuel cost function incre ases unintentionally although G max increases equaling 300. This is also repeated one more time at the last ro w . The o v ervie w on T able 2 point out that the best cost of 623.8709 $/h is obtained at G max = 350. Meanwhile, the minimum cost at G max = 400 is higher than 623.8709 $/h. Clearly , this phenomenon has been caused by the impact of v alv e point loading ef fects on the stability of the ALO search process. The most of the fuel costs for case 2 o v er 100 trials ha v e distrib uted nearby the minimum Antlion optimization algorithm for optimal non-smooth... (Thanh Pham V an) Evaluation Warning : The document was created with Spire.PDF for Python.
1194 r ISSN: 2088-8708 v alue as sho wn in Figure 2. This sho ws that ALO has good stability for solving the OELD problems on the 10-unit system with multi fuel sources and v alv e point loading ef fects. T able 2. Result obtained by ALO for case 2 with dif ferent v alues of control v ariables Lo west cost ($/h) A v erage cost ($/h) Highest cost ($/h) N pop G max 629.9271 653.2830 721.8110 20 50 624.1303 638.9358 690.4222 20 100 624.0333 631.6475 653.1102 20 150 623.9216 628.4258 643.9007 20 200 623.8958 625.8331 643.8601 20 250 623.9309 626.1910 644.3901 20 300 623.8709 625.6935 636.3510 20 350 623.8878 625.2053 635.7175 20 400 5.1.3. Case 3: 10-unit po wer system using multi fuel sour ces, v alv e effects and complicated constraints In the third studied case , input parameters and obtained results are presented in T able 3. As observ ed from T able 3, the minimum fuel costs obtained by ALO can drop significantly from G max = 50 to G max = 400 and it reaches the best minimum at G max = 400 with the cost of 624.3894 $/h. Ho we v er , the minimum cannot be reduced since setting G max to 450 and 500, corresponding to the cost of 624.3976 $/h and 624.4035 $/h. Clearly , the phenomenon is similar to that in case 2 b ut totally dif ferent from 4 s u bc ases in case 1. Ob viously , v alv e point loading ef fects and complicated constraints ha v e a highly significant impact on the stability of ALO. Figure 2 is sho wn the fuel costs after 100 trials. The y are w a v ered between tw o numbers 625 $/h and 630 $/h. T able 3. Result obtained by ALO for case 3 with dif ferent v alues of control v ariables Lo west cost ($/h) A v erage cost ($/h) Highest cost ($/h) N pop G max 625.0314 634.9521 658.0532 30 50 624.6915 628.9171 639.8262 30 100 624.5606 627.2422 637.3534 30 150 624.4409 627.0395 634.9398 30 200 624.4920 626.2745 632.8379 30 250 624.4626 626.3989 630.5901 30 300 624.4287 626.2675 630.5357 30 350 624.3894 625.9337 630.5276 30 400 624.3976 625.7483 630.2841 30 450 624.4035 625.6773 629.0156 30 500 Figure 2. The best cost of 100 trials from sub-case 1.4, case 2 and case 3 Int J Elec & Comp Eng, V ol. 10, No. 2, April 2020 : 1187 1199 Evaluation Warning : The document was created with Spire.PDF for Python.
Int J Elec & Comp Eng ISSN: 2088-8708 r 1195 5.2. Result comparison obtained by ALO and other methods In this section comparison of results obtained by the ALO method and other onces has been performed. In order to in v estig ate the real performance of ALO method, an important f actor must be concerned to be a number of fitness function e v aluations N f e , which is calculated by: N f e = ! N pop G max (26) The ALO method has only one ne w solution generation for each iteration, so ! is 1 for ALO. F or other methods such as F A and IF A in [38], N f e has another model, that is: N f e = N pop ( N pop + 1) G max 2 (27) 5.2.1. Comparisons of T est Case 1 This section compares fuel cost obtained by ALO and dif ferent methods from four subcases of case 1. The best cost together with N f e are reported in T able 4. Numbers from T able 4 point out that the proposed method can harv est optimal solutions as good as others. While ALHN [6] and EALHN [7] are not metaheuristic methods. These methods will f ace to the dif culties of applying for OELD problem with non-con v e x fuel cost function and comple x constraints. ALO has an approximate minimum to most methods and has better cost than F A [38] for all subcases and CF A [31] for subcase 1.4. Ho we v er , as comparing N f e v alues, ALO has tak en N f e = 2,500 for seeking and reachi ng the best optimal solutions while other methods ha v e emplo yed high v alue of N f e , from 5,500 to 156,000. Namely , N f e w as respecti v ely set to 12,000, 30,000 and 156,000 for DEA, HRCGA, and CF A. In summary , ALO has found approximate or better results than compared methods b ut it has o wned v ery f ast con v er gence speed compared to other ones. In s ummary , the applied ALO method is really ef fecti v e for case 1 with discrete objecti v e function. T able 4. The lo west cost ($/h) of case 1 and compared methods Load (MW) ALHN [6] EALHN [7] DEA [8] HICDE- DP [10] HRCGA [15] CSA [23] AIS [29] CF A [31] F A [38] IF A [38] ALO 2,400 481.723 481.723 481.723 481.7226 481.7226 - 481.723 - 505.2337 481.7226 481.7226 2,500 526.239 526.239 526.239 526.2388 526.2388 - 526.24 - 580.4417 526.2388 526.2388 2,600 574.381 574.381 574.381 574.3808 574.3808 - 574.381 - 639.287 574.3808 574.3808 2,700 623.809 623.809 623.809 623.809 623.8092 623.8092 623.8092 623.8339 679.9525 623.809 623.8093 N f e - - 12,000 8,000 30,000 6,000 4,000 156,000 11,000 5,500 2,500 5.2.2. Comparisons of T est Case 2 The section is carrying our comparison of results from the applied ALO and other methods for case 2. According to reported data in T able 5, KHA [27] is the best one; ho we v er , checking optimal solution reported in [27] sees that incorrect type of fuel w as reported and fuel cost is much higher than reported v alues. Thus, KHA [27] is not a competitor of the applied ALO method. When compared to accepted methods lik e GA, TSA, PSO [20], F A and IF A [38], the proposed method is better with respect to the best cost. On the contrary , the ALO method is less ef fecti v e than remaining methods lik e in DSPSO-TSA [20], CSA [23], ICSA [25]. Ho we v er , it should be noted that CSA and ICSA ha v e used N f e equaling 10,000 and 12,000 while that of the applied ALO method w as 7,000. Compared to GA, TSA, and PSO in [20], ALO has better cost b ut higher N f e since those from ALO are 623.8708 and 7,000 while those from these methods are higher than 624.3045 and 3,000. Ho we v er , results of ALO re p or ted in T able 2 sees that ALO found the best cost of 624.1303 at N pop = 20 and G max = 100, corresponding to N f e = 2,000. Thus, ALO could find better optimal solutions with f aster con v er gence than GA, TSA, and PSO in [20]. In summary , ALO has yielded better results b ut its search speed has been f aster than these methods while other ones with better results than ALO were slo wer for con v er ging to the best optimal solution. In other w ords, ALO is a promising method for case 2 considering with 10 units taking multi fuel sources and v alv e point loading ef fects into account. Antlion optimization algorithm for optimal non-smooth... (Thanh Pham V an) Evaluation Warning : The document was created with Spire.PDF for Python.
1196 r ISSN: 2088-8708 T able 5. The comparison of results for case 2 Cost ($/h) CCDE [12] GA [20] TSA [20] PSO [20] DSPSO- TSA [20] CSA [23] ICSA [25] KHA [27] F A [38] IF A [38] ALO Lo west 623.8288 624.505 624.3078 624.3045 623.8375 623.8684 623.8684 605.7582 664.5306 623.8768 623.8708 A v erage 623.8574 624.7419 635.0623 624.5054 623.8625 623.9495 623.9495 605.8043 675.5344 625.2704 625.6935 Highest 623.8904 624.8169 624.8285 625.9252 623.9001 626.3666 626.3666 605.9426 679.426 629.2765 636.3510 N f e 7,000 3,000 3,000 3,000 3,000 10,000 12,000 10,000 11,000 5,500 7,000 5.2.3. Comparisons of T est Case 3 The section presents the com parison of fuel cost and N f e from the applied ALO and other methods for case 3. The results obtained for case 3 are gi v en in T able 6. According to the results, the ALO method is only less ef fecti v e than IRCGA [17] while it is better than all other methods; ho we v er , the applied ALO method has the most ef fecti v e con v er gence speed since it has used N f e = 12,000 b ut that from other w as much higher . F or instance, the v alue is 33,000 for IF A [38], 66,000 for F A [38] and 90,000 for both RCGA and IRCGA. Clearly , ALO can be f aster than these methods approximately from 3 times to 8 times. In summary , ALO has found a better optimal solution than three methods b ut less ef fecti v e optimal solution than one method. Ho we v er ALO is f aster than all compared methods from 3 times to 8 times. Thus, ALO is a v ery ef fecti v e method for the case, which has multi fuel sources, v alv e point loading ef fects and man y comple x constraints. The optimal solution obtained by the ALO algorithm for all cases ha v e been presented in Appendix. T able 6. The comparison of results for case 3 (2,700 MW) Cost ($/h) RCGA [17] IRCGA [17] F A [38] IF A [38] ALO Lo west 624.6605 624.355 673.5544 624.4950 624.3894 A v erage 625.9201 624.5792 685.2872 625.2647 625.6773 Highest 628.9253 624.7541 699.2855 629.3951 629.0156 N f e 90,000 90,000 66,000 33,000 12,000 In this section, it is e xplained the results of research and at the same time is gi v en the comprehensi v e discussion. Results can be presented in figures, graphs, tables and others that mak e the reader understand easily [2, 5]. The discussion can be made in se v eral sub-chapters. 6. CONCLUSION In this article, the proposed ALO method is ef fectually implemented to handle the OELD problem. The studied system has 10 TGUs with dif ferent types of fuel consuming characteristic and almost comple x operation constraints of the po wer grid practiced in three tested cases. The method has been pro v ed to be stable, ef fecti v e and rob ust. The obtained results ha v e been compared with man y other methods. The comparison can imply that the ALO method is better than most other methods in term of lo wer fuel cost and smaller number of fitness e v aluations. Ho we v er , ALO has not found all better resul ts than all methods for study cases, especially in comparison with impro v ed v ersions of original meta-heuristic algorithms. Thus, it can conclude that ALO method can be selected as an optimization tool for dealing with OELD problem b ut it needs more impro v ements for enhancing optimal solution quality and con v er ge speed. REFERENCES [1] Jonathan F . Bard, ”Short-T erm Scheduli n g of Thermal-Electric Generators Using Lagrangian Relaxation, IEEE T ransactions on Po wer Systems , v ol. 36(5), pp. 756-766, 1988. [2] A. F arag, S. Al-Baiyat and T . C. Cheng, ”Economic load dispatch multiobjecti v e optimization procedures using linear programming techniques, IEEE T ransactions on Po wer Systems , v ol. 10(2), pp. 731-738, May 1995. [3] Jiann-Fuh Chen and Shin-Der Chen, ”Multiobjecti v e po wer dispatch with line flo w constraints using the f ast Ne wton-Raphson method, IEEE T ransactions on Ener gy Con v ersion , v ol. 12(1), pp. 86-93, March 1997. Int J Elec & Comp Eng, V ol. 10, No. 2, April 2020 : 1187 1199 Evaluation Warning : The document was created with Spire.PDF for Python.