An Efficient Optimal Fractional Emotional Intelligent Controller for an AVR System in Power Systems

Transcription

1 Journal of AI and Data Mnng Vol 7, No 1, 2019, DOI: /JADM An Effcent Optmal Fractonal Emotonal Intellgent Controller for an AVR Sytem n Power Sytem M. Morad Zrkoh Department of Electrcal Engneerng, Behbahan hatam Alanba Unverty of Technology, Behbahan, Iran. Receved 18 February 2018; Reved 26 October 2018; Accepted 02 November 2018 *Correpondng author: morad@bkatu.ac.r(m. Morad). Abtract In th paper, a hgh-performance optmal fractonal emotonal ntellgent controller propoed for an Automatc Voltage Regulator (AVR) n a power ytem ung the Cuckoo optmzaton algorthm (COA). AVR the man controller wthn the exctaton ytem that preerve the termnal voltage of a ynchronou generator at a pecfed level. The propoed control trategy baed upon bran emotonal learnng, whch a elf-tunng controller o-called bran emotonal learnng-baed ntellgent controller (BELBIC), and baed on the enory nput and emotonal cue. The major contrbuton of th paper the ue of the mert of the fractonal order PID (FOPID) controller; an FOPID controller employed to formulate the tmulant nput (SI) gnal. Th a dtnct advantage over the paper publhed n the lterature, n whch a PID controller ha been reported to be ued to generate the SI gnal. Another remarkable feature of the propoed approach that t a model-free controller. The propoed control trategy can be a promng controller n term of mplcty of degn, eae of mplementaton, and le tme-conumpton. In addton, n order to enhance the performance of the propoed controller, t parameter are tuned by COA. COA a novel advanced optmzaton algorthm proved to have a hgh effcency. In order to degn a BELBIC controller for an AVR ytem, a mult-objectve optmzaton problem ncludng overhoot, ettlng tme, re tme, and teady-tate error formulated. The mulaton tude confrm that the propoed controller, compared to the clacal PID and FOPID controller ntroduced n the lterature, how a uperor performance regardng the model uncertante. Havng appled the propoed controller, the re tme and the ettlng tme were found to be mproved by 47% and 57%, repectvely. eyword: Bran Emotonal Learnng-baed Intellgent Controller, Cuckoo Optmzaton Algorthm, Fractonal Order PID, Automatc Voltage Regulator. 1. Introducton Unlke the DC generator, alternator cannot be compounded to alter the voltage-load charactertc [1]. Moreover, due to change n the load power factor, the output voltage varaton are ncreaed. That why the automatc voltage regulator (AVR) are generally ued wth alternator [1]. In addton, for the dampng power ytem ocllaton, power ytem tablzer (PSS) can be ued. The reference voltage of an AVR ytem modfed ung the PSS output gnal, and then AVR defne the prmary voltage regulaton of ynchronou machne [2]. Therefore, The AVR ytem play a decve role n the ndutre to mantan the contant termnal voltage n the ynchronou generator under all the condton. However, due to the hgh nductance of the generator feld wndng and load varaton, achevng a dered repone dffcult [3]. Therefore, t mportant to mprove the AVR performance and enure a table and effcent repone to tranent change n termnal voltage. That why an optmal controller n an AVR ytem requred [4]. So far, varou control method have been propoed for the AVR ytem. One of the preferable controller propoed n the lterature the proportonal plu ntegral plu dervatve (PID) controller becaue of t mple tructure. Therefore, many optmzaton algorthm ncludng partcle warm optmzaton (PSO) [5],

2 harmony earch algorthm (HSA) [6], and artfcal bee colony algorthm (ABC) [7, 8] have been employed to regulate the PID gan of the AVR ytem. Recently, n order to mprove the performance of the PID controller, fractonal calculu ha played an mportant role n many control applcaton. The fractonal order PID (FOPID) controller are known a a generalzaton of a tandard PID controller ntroduced by Podlubny [9, 10]. It hould be noted that the FOPID controller n many control applcaton have hown a better control performance than a tandard nteger order PID controller due to extra degree of freedom [10]. Compared to the PID controller, the FOPID controller ha two extra parameter [11]. Th mean that t characterzed by fve parameter [12]. However, the electon of the value for t parameter tll a challengng tak. So far, everal ntellgent technque have been propoed for an effcent tunng of the FOPID controller. Such algorthm nclude Partcle warm optmzaton (PSO) [10, 13, 14], Genetc Algorthm (GA) [14, 15], frut fly optmzaton algorthm [16], and Cuckoo Optmzaton Algorthm (COA) [17]. In the lterature, t ha been tated that the FOPID controller ha been employed for controllng a practcal AVR a well. Zhang [5] ha uggeted the PSO approach for the optmum degn of the FOPID controller n the AVR ytem. The reult obtaned ung th method have been compared wth the conventonal PID controller. Zhang et al. [7] have ued the Artfcal Bee Colony (ABC) algorthm to optmze the parameter of the FOPID controller for an AVR ytem. The reult obtaned have been compared wth PSO and GA, ndcatng that the propoed controller ha a better performance. Ynggan et al. [12] have employed the Chaotc Ant Swarm (CAS) optmzaton algorthm to regulate the parameter of the FOPID controller. The reult obtaned have ndcated that the propoed FOPID controller robut to model uncertante. Recently, a new evolutonary algorthm called COA ha been propoed [18]. The comparon of COA wth the tandard veron of PSO and GA have revealed the uperorty of COA n term of fat convergence and accuracy [18, 19]. Therefore, COA ha been ued for the optmzaton of the FOPID parameter for the AVR ytem [17]. In th tate-of-the-art reearch work, t ha been hown that the propoed optmzed controller provde a more mproved dynamc performance compared to the other extng technque. Depte the great effort devoted to AVR control, many of the theoretcal reult cannot be drectly appled to the practcal AVR ytem. The reaon that th ytem nclude model non-lnearty and model uncertante. Recently, ntellgent technque ncludng neural network [20] and fuzzy ytem [21, 22] have been ntroduced to overcome thee ue. The reaon for the fact that n the academc fled the applcaton of ntellgent control ha been ncreaed that they are modelfree control method [23]. Th mean that a perfect dynamc model not requred. Therefore, ome reearcher have employed ntellgent technque n order to control the AVR ytem. For example, n [24], the fuzzy controller ha been ued for the AVR ytem. Recently, the bran emotonal learnng controller ha been ntroduced a a new ntellgent controller [25]. Bran emotonal learnng baed upon a computatonal model of a lmbc ytem n the human bran [23]. Specfcally, BELBIC eentally an acton generaton mechanm baed on enory nput (SI) and emotonal cue (EC), and the mot nteretng concept of BELBIC the flexblty n defnton of SI and EC dependng on the control problem [25]. So far, BELBIC ha been ued n many ndutral applcaton uch a wahng machne [26, 27], power ytem applcaton [27-30], aeropace launch vehcle [31], and mcro-heat exchanger [32]. The mentoned paper have ndcated that BELBIC ha a good robutne and performance [33]. Snce the parameter to be tuned n BELBIC are le that neural network and fuzzy ytem, BELBIC ha a mple tructure. A mentoned earler, mot of the propoed controller for AVR are baed upon the optmzaton algorthm. It hould be noted that n real power ytem, th earch proce take a long tme [34]. In addton, by conderng the tructured and untructured uncertante, the dynamc behavor of real power ytem are dfferent [34]. Therefore, the PID or FOPID controller optmzed by off-lne earch algorthm may not have a good performance under thee condton [34]. The man objectve of th paper wa to addre thee ue. The man contrbuton of th work to propoe a novel emotonal ntellgent controller a the man controller of the AVR ytem, n whch, to ue the mert of the FOPID controller, an FOPID controller ued to formulate the tmulant nput gnal. One advantage of the propoed controller that t a elf-tunng controller. Th mean that dependng on the operatonal condton of the AVR ytem, t behavor modfed. The 192

3 econd contrbuton of th work that COA employed to optmze the control degn parameter and enhance the performance of the control ytem. Snce the number of parameter of COA le than the parameter ued n the other meta-heurtc technque, th fact reult n a fat convergence of the BELBIC parameter. In order to degn a BELBIC controller for an AVR ytem, a mult-objectve optmzaton problem ncludng the overhoot, ettlng tme, re tme, and teady-tate error wa formulated. To the bet of our knowledge, th the frt tme that th tructure ha been preented. Th a dtnct advantage n comparon to the paper publhed n the lterature. The great mert of the propoed control trategy that t uperor to the fuzzy and neural network ytem n term of mplcty of degn, eae of mplementaton, and le tme-conumpton. In addton, t a modelfree controller. Th mean that t doe not requre any nformaton from the ytem dynamc. The performance of the propoed controller wa compared wth ome PID and FOPID controller ntroduced n the prevou reearch work. Alo uncertante n the AVR parameter were taken nto account to how the robutne of the propoed controller. The remanng part of th paper organzed a what follow. In Secton 2, the AVR ytem decrbed. In Secton 3, the fractonal PID controller are brefly ntroduced. The decrpton of the cuckoo earch algorthm dcued n Secton 4. Secton 5 brefly decrbe the bran emotonal learnng. In Secton 6, the mulaton reult are preented. Fnally, Secton 7 conclude the paper. 2. AVR ytem model To mantan the termnal voltage magntude at a contant pecfed level n ynchronou generator, the AVR ytem ued [3, 35]. The AVR ytem cont of four man component, namely amplfer, excter, enor, and generator [3]. For mathematcal modelng and the tranfer functon of the four component, thee component mut be lnearzed, whch take nto account the major tme contant and gnore the aturaton or other non-lnearte [5]. From a practcal vewpont, by gnorng the aturaton or other non-lnearte, the PID and FOPID controller may not gve a good performance n practcal applcaton. That why n th work, a elf-tunng controller called BELBIC wa developed to deal wth the uncertante. The tranfer functon of the mentoned component can be repreented a what follow [3, 5, 12]. The tranfer functon repreentaton of amplfer a follow: A GA ( ) )1( 1 A where, the amplfer gan ( ) and tme contant ( ) are gven a and A 0.02 A 0.1. The tranfer repreentaton of the excter model a follow: E GE ( ) )2( 1 E where, the excter gan ( ) and tme contant ( E E ) are gven a 10 E 400 and 0.5 E 1, repectvely. The tranfer functon repreentaton of the generator gven a follow: G GG ( ) )3( 1 G where, the excter gan ( ) and tme contant ( G G ) are gven a 0.7 G 1 and 1 G 2, repectvely. The tranfer functon repreentaton of the enor model a follow: G ( ) )4( 1 where, the feedback gan ( A A ) and tme contant ( ) are gven a and , repectvely. The block dagram of the AVR ytem component ncludng BELBIC hown n fgure 1. Fgure 2 how the voltage repone of the AVR ytem wthout conderng the controller. A een, t exhbt hgh ocllaton wth M 65.43%, t 0.75, t 6.97, t In the teady tate condton, the ytem termnal voltage devate from the nomnal value of In a power ytem wth a hgh operatng voltage, th repone completely underable [36]. For th reaon, a controller requred to be ncorporated n the AVR ytem. 3. Fractonal PID controller The fractonal calculu a name of the theory of ntegraton and dervatve of arbtrary order [37]. The FOPID controller a fractonal order tructure, whch provde more flexblty compared wth the PID controller [38]. So far, FOPID ha been appled for control purpoe [3, 7, 12, 37]. p p r 193

4 Fgure 1. Block dagram of an AVR ytem wth controller. Fgure 2. Step repone of an AVR ytem wthout controller. Fgure 3. Graphcal repreentaton of FOPID controller [17]. Generally, the FOPID controller gven a: FOPID ( ) G p D where,, and D are the proportonal, p ntegral, and dervatve gan, repectvely. In the meanwhle, the fractonal order of the ntegral part and the fractonal order of the dervatve part of the FOPID controller. Fgure 3 how the graphcal repreentaton of the FOPID controller. A een, dependng upon the value for λ and μ, the conventonal P, PI, PD, and PID controller can be obtaned from the FOPID controller. )5( 4. Cuckoo optmzaton algorthm (COA) Recently, a novel evolutonary algorthm called COA ha been ntroduced, whch npred by the lfe of cuckoo [18, 39]. Lke the other evolutonary algorthm, COA tart wth an ntal populaton of cuckoo called habtat [18]. To olve the optmzaton problem, a canddate habtat matrx of ze N N mut be generated. Meanwhle, N p the maxmum number of cuckoo that can lve at the ame tme and N var the number of parameter to be optmzed. Further, each cuckoo layng egg wthn a dtance called the egg layng radu (ELR) defned baed upon the followng equaton [18, 40]: Number of current cukoo ' egg ELR Total number of egg )6( (var var ) h low where, var h the upper lmt and var low the lower lmt of varable. In the meanwhle, an nteger, defned to handle the maxmum value of ELR. Each cuckoo tart layng egg randomly n ome other hot brd net wthn her ELR [41]. The new egg-layng proce can be defned a follow: X next X current F ( X goal X current ) )7( where, X and F are the poton and the moton coeffcent, repectvely. The peudo-code of the COA algorthm can be found a n [18]. To evaluate the performance of the AVR ytem, an objectve functon mut be defned. Therefore, n th paper, a tme doman performance crteron defned a [5]: p var 194

5 J (1 e )( M e ) e ( t t ) p r where, the weghtng factor, and M, e, t, p and t r denote the maxmum overhoot, teady tate error, ettlng tme, and re tme, repectvely. It hould be noted that n th objectve functon, the tme repone pecfcaton are ncluded. 5. Bran emotonal learnng-baed ntellgent controller (BELBIC) A hown n fgure 4, the man component of the lmbc ytem nvolved n emotonal procee are amygdala, orbtofrontal cortex, thalamu, enory cortex, hypothalamu, and hppocampu [30]. learnng [42]. Th model hown n fgure 5. The emotonal learnng take place motly n the amygdala. In fact, the amygdala reponble for long-term memory and emotonal tmul. It receve gnal from the enory cortex and n nteracton wth the orbtofrontal cortex [23]. A een n fgure 5, BELBIC ha two tate for each enory nput. Fgure 6 how the computatonal model of emotonal learnng n more detal a well. A een n th fgure, the vector S how the tmul nput to the ytem. There a node for each tmulu S. Suppoe the th enory nput a. )8( A een, A th an nput to the amygdala part, whch the maxmum of tmul nput (SI), gven by [43]. A V (max( S ) S ) )9( th th th In the meanwhle, the weght accordng to the followng equaton: V k (max(0, S ( EC A ))) th th th V th updated )10( where, k th the learnng tep. In addton, the amygdala and orbtofrontal cortex output are, repectvely, gven by: A S V )11( O S W )12( where, V and W are two tate that are updated accordng to the followng equaton [43]: V th k 1 (max(0, S ( EC A ))) )13( W k S E A ( ( ' ))) 2 )14( where, k 1 and k 2 are the learnng tep n the amygdala and orbtofrontal cortex, repectvely [44]. Fgure 6. Graphcal depcton of the BEL proce [42]. Fgure 4. The lmbc ytem of the bran [28]. Fgure 5. Block dagram of the preented computatonal model of human bran learnng mechanm. Eventually, the model output (MO) or output ' node E and E are gven a: E A Ath O )15( ' E A O )16( It hould be noted that the weght V n Eq. (13) cannot be decreaed. It can be concluded that the nformaton n the amygdala part not forgotten. In th paper, to mprove the performance of the propoed approach, the thalamu alo modeled by Eq. (10). A a reult, dfferent learnng tep are condered n Eq. (10) and (13). 195

6 Table 1. Searchng range of parameter. Parameter Mn. value Max. value Fgure 7. Block dagram of the propoed tructure. 6. Propoed BELBIC The enory nput (SI) a knd of control gnal that n BELBIC renforced or punhed baed on an emotonal cue o t hould be choen a a functon of error, jut lke a PID controller [32]. In mot of the publhed paper, the reearcher have utlzed the PID controller to form the tmulant nput gnal [32, 34]. For mplcty, th named a PID-BELBIC. In the preent paper, a a novelty, owng to the value and hgh performance of elf-tunng FOPID controller, a FOPID controller employed to formulate the tmulant nput gnal, a gven by (17). For mplcty, th named a FOPID-BELBIC. SI ( ) ( p D ) E ( ) )17( where,,, and D are the proportonal, p ntegral, and dervatve gan, repectvely. In the meantme, 0 and 0 are not nteger. The tmulant nput gnal n the tme doman : SI ( t ) pe( t ) D e( t ) DD e( t ) )18( In addton, the emotonal gnal (EC) generally mut how the cloed-loop ytem performance. Therefore, EC can be wrtten a a weghted combnaton of prmary/econdary objectve n the applcaton doman, a follow [34]: EC ac e ac e& ac E )19( where, E gven n Eq. (15) and e the dfference between the reference voltage and meaured output voltage, namely e V V. In the meanwhle, a, a, and are the weght c1 c2 coeffcent to be determned. The block dagram of the propoed tructure hown n fgure 7. a c3 ref 7. Smulaton reult and dcuon In th ecton, the propoed controller wa teted n controllng the AVR ytem. To th end, to evaluate the performance and effcacy of the propoed controller, a practcal hgh order automatc voltage regulator wa condered wth the followng pecfcaton: 10, 0.1, 1, 0.4, 1, 1, A A E E G G 1, The block dagram of the control ytem preented n fgure 1. The number of parameter to be optmzed by COA 14, namely k, k, k,,, k, k, k, a, a, a, m, m and p d 1 2 th c1 c2 c3 a o m th. Note that ma, mo and m th are the ntal condton of the memory ued n (13), (14), and (10), repectvely. In addton, the range of thee parameter are gven n table 1. Further, the parameter of COA are gven a the maxmum number of egg for each cuckoo, 10; mnmum number of egg, 5; maxmum number of cuckoo that lve at the ame tme, 20; 5 and F 9. The maxmum number of teraton wa alo et to 50 a a toppng crteron. In the objectve functon, wa et to 1.5. In our mulaton, two cenaro were conducted to confrm the effectvene of the propoed approach. Scenaro 1: In th cae, parametrc uncertante were not condered. Both FOPID-BELBIC and PID-BELBIC were optmzed. A a reult of applyng COA, the tep repone of AVR are depcted n fgure 8. A een, the performance of the propoed controller (FOPID-BELBIC) wa better than PID-BELBIC. One reaon that the number of parameter to be tuned n FOPID- BELBIC 14, wherea the number of parameter to be tuned n PID-BELBIC 12. FOPID- BELBIC ha the advantage over the PID-BELBIC n that t ha greater degree of freedom n the controller parameter. Thee extra degree of freedom can help to enhance the performance of the propoed approach. To have a better comparon, the value for the performance crtera for dfferent controller are ummarzed n table 2. The reult obtaned confrm that FOPID- BELBIC ha a better performance n term of all four ndce. Fgure 9 how the convergence of 196

7 objectve functon at each generaton. A een, COA well-convergent. Type of controller Table 3. Comparatve performance of the controller. MOEO-FOPID [10] CS-FOPID [17] Fgure 8. Step repone of the AVR ytem ung FOPID- BELBIC and PID-BELBIC. Table 2. Comparon of FOPID-BELBIC and PID- BELBIC. Type of controller FOPID-BELBIC PID-BELBIC Fgure 9. Objectve functon veru teraton. It wa oberved that COA converged to n an around 19 teraton. In addton, the correpondng control parameter trajectore of SI parameter regardng the optmzaton algorthm hown n fgure 10, whch ndcate the convergence of the oluton. To further demontrate the effectvene of the propoed FOPID-BELBIC, we gave the comparatve performance of FOPID-BELBIC wth dfferent optmzed FOPID controller recently publhed n the lterature uch a MOEO-FOPID [10] and CS-FOPID [17]. The reult obtaned are gven n table 3 and the correpondng tep repone alo hown n fgure 11. Accordng to the preented reult, the reult obtaned ndcate that the AVR ytem exhbt a better performance, a compared wth the other well-known controller avalable n the lterature. Scenaro 2: In th cae, to tet the robutne and powerfulne of the propoed controller, parametrc uncertante were condered. To ave pace, the uncertante of the AVR model were gven n term of varaton n the excter generator. The varaton range of the tme contant wa choen to be ±50% of t nomnal value wth a 25% tep ze. The reult obtaned are depcted n fgure 12. A can be oberved n th fgure, the devaton of repone curve. (±50% and ±25%) from the nomnal repone for the elected tme contant parameter wthn a mall range. Th can enure the robutne of the propoed controller agant uch large varaton. To have a better vew, the tep repone of the AVR ytem by conderng 50% devaton n excter generator a a reult of applyng propoed controller and MOEO- FOPID/PID are hown n fgure 13. A een, the propoed controller gve a better repone. The ytem repone ung BELBIC rather fater wth le overhoot, even n the face of uncertanty. To um up, from Scenaro 1 and 2, t can be concluded that the propoed controller outperform other FOPID controller wth parameter varaton n the AVR model. Fgure 10. Correpondng control parameter trajectore of SI parameter veru teraton. 197

8 Fgure 11. Comparon of unt tep repone of AVR ytem wth dfferent controller. Fgure 13. Step repone of propoed controller and MOEO-PID/FOPID by conderng uncertanty n excter. The practcal mplcaton of the propoed method a part of our future work. 9. Acknowledgment The author gratefully apprecate the upport of the Behbahan hatam Alanba Unverty of Technology. Fgure 12. Comparon of unt tep repone of AVR ytem by conderng uncertanty n excter. 8. Concluon In th paper, a elf-tunng controller baed on bran emotonal learnng ha been preented. The propoed approach nvolve an FOPID controller to generate an SI gnal. Th the unque feature of the propoed controller. Applcaton of th method to a practcal AVR ytem how that the developed control cheme outperform the PID and FOPID controller. To enhance the performance of control ytem, COA wa ued to tune the control degn parameter. In addton, to how the robutne of the propoed controller, model uncertante were alo condered. The mulaton reult confrmed that, compared to the PID and FOPID controller, the propoed controller had more robut tablty and performance charactertc. Havng appled the propoed controller, compared to PID-BELBIC, the re tme and ettlng tme were mproved by about 47% and 57%, repectvely. In addton, compared to the CS-FOPID method preented n [17], the ettlng tme wa mproved by about 75%. In partcular, the propoed cheme could be a promng controller n term of mplcty of degn, eae of mplementaton, and le tmeconumpton. It worthy of note that from the mulaton reult t could be concluded that the theoretcal reult obtaned had a potental n applcaton. Reference [1] eljk, J., (2013), Electrcty 3: power generaton and delvery, Nelon Educaton. [2] Mlano, F., (2010), Power ytem modellng and crptng, Sprnger Scence & Bune Meda. [3] Sahb, M. A. (2015). A novel optmal PID plu econd order dervatve controller for AVR ytem. Engneerng Scence and Technology, an Internatonal Journal, vol. 18, no. 2, pp [4] BINGUL, Z., &ARAHAN, O. (2018). A Novel Performance Crteron Approach to Optmum Degn of PID Controller Ung Cuckoo Search Algorthm for AVR Sytem. Journal of the Frankln Inttute, vol. 355, no. 13, pp [5] Gang, Z.-L. (2004). A partcle warm optmzaton approach for optmum degn of PID controller n AVR ytem. IEEE tranacton on energy converon, vol. 19, no. 2, pp [6] Sambarya, D., &Palwal, D., (2016), Optmal degn of PIDA controller ung harmony earch algorthm for AVR power ytem, Power Sytem (ICPS), 2016 IEEE 6th Internatonal Conference on, IEEE, pp [7] ZHANG, D.-L., Yng-Gan, T., &Xn-Png, G. (2014). Optmum degn of fractonal order PID controller for an AVR ytem ung an mproved artfcal bee colony algorthm. Acta Automatca Snca, vol. 40, no. 5, pp [8] Gozde, H., &Taplamacoglu, M. C. (2011). Comparatve performance analy of artfcal bee colony algorthm for automatc voltage regulator (AVR) ytem. Journal of the Frankln Inttute, vol. 348, no. 8, pp

9 [9] Polubny, I. (1999). Fractonal-order ytem and PIλDμ controller. IEEE Tran. Automatc Control, vol. 44, no., pp [10] Zeng, G.-Q., et al. (2015). Degn of fractonal order PID controller for automatc regulator voltage ytem baed on mult-objectve extremal optmzaton. Neurocomputng, vol. 160, no., pp [11] Chen, Z., et al. (2014). Degn of a fractonal order PID controller for hydraulc turbne regulatng ytem ung chaotc non-domnated ortng genetc algorthm II. Energy Converon and Management, vol. 84, no., pp [12] Tang, Y., et al. (2012). Optmum degn of fractonal order PI λ D μ controller for AVR ytem ung chaotc ant warm. Expert Sytem wth Applcaton, vol. 39, no. 8, pp [13] Cao, J.-Y., &Cao, B.-G., (2006), Degn of fractonal order controller baed on partcle warm optmzaton, Indutral Electronc and Applcaton, ST IEEE Conference on, IEEE, pp [14] Elyagomar, V., et al. (2015). Cancer clafcaton ung a novel gene electon approach by mean of hufflng baed on data cluterng wth optmzaton. Appled Soft Computng, vol. 35, no., pp [15] Cao, J.-Y., Lang, J., &Cao, B.-G., (2005), Optmzaton of fractonal order PID controller baed on genetc algorthm, Machne Learnng and Cybernetc, Proceedng of 2005 Internatonal Conference on, IEEE, pp [16] Sheng, W., &Bao, Y. (2013). Frut fly optmzaton algorthm baed fractonal order fuzzy- PID controller for electronc throttle. Nonlnear Dynamc, vol. 73, no. 1-2, pp [17] Skander, A., et al. (2018). A novel technque to degn cuckoo earch baed FOPID controller for AVR n power ytem. Computer & Electrcal Engneerng, vol. 70, no. 1, pp [18] Rajaboun, R. (2011). Cuckoo optmzaton algorthm. Appled oft computng, vol. 11, no. 8, pp [19] Lahkar, M., &Moattar, M. (2017). Improved COA wth Chaotc Intalzaton and Intellgent Mgraton for Data Cluterng. Journal of AI and Data Mnng, vol. 5, no. 2, pp [20] Sngh, P., et al., (2017), Sldng Mode Control of Uncertan Nonlnear Dcrete Delayed Tme Sytem Ung Chebyhev Neural Network. Advance n Computer and Computatonal Scence, Sprnger, pp [21] Zrkoh, M. M., umbaar, T., &Ln, T. C. (2017). Hybrd Adaptve Type 2 Fuzzy Trackng Control of Chaotc Ocllaton Dampng of Power Sytem. Aan Journal of Control, vol. 19, no. 3, pp [22] Hamza, M. F., Yap, H. J., &Choudhury, I. A. (2017). Recent advance on the ue of meta-heurtc optmzaton algorthm to optmze the type-2 fuzzy logc ytem n ntellgent control. Neural Computng and Applcaton, vol. 28, no. 5, pp [23] horahadzadeh, S., &Mahdan, M., (2016), Voltage trackng control of DC-DC boot converter ung bran emotonal learnng, Control, Intrumentaton, and Automaton (ICCIA), th Internatonal Conference on, IEEE, pp [24] Pryambada, S., Sahu, B.., &Mohanty, P.., (2015), Fuzzy-PID controller optmzed TLBO approach on automatc voltage regulator, 2015 Internatonal Conference on Energy, Power and Envronment: Toward Sutanable Growth (ICEPE), pp [25] Luca, C., Shahmrzad, D., &Shekholelam, N. (2004). Introducng BELBIC: bran emotonal learnng baed ntellgent controller. Intellgent Automaton & Soft Computng, vol. 10, no. 1, pp [26] Luca, C., Mla, R. M., &Araab, B. N. (2006). Intellgent modelng and control of wahng machne ung locally lnear neuro fuzzy (llnf) modelng and modfed bran emotonal learnng baed ntellgent controller (BELBIC). Aan Journal of Control, vol. 8, no. 4, pp [27] Shekholelam, N., et al. (2006). Applyng bran emotonal learnng algorthm for multvarable control of HVAC ytem. Journal of Intellgent & Fuzzy Sytem, vol. 17, no. 1, pp [28] Qutubuddn, M., &Yadaah, N. (2017). Modelng and mplementaton of bran emotonal controller for Permanent Magnet Synchronou motor drve. Engneerng Applcaton of Artfcal Intellgence, vol. 60, no., pp [29] Dehkord, B. M., et al. (2011). Senorle peed control of wtched reluctance motor ung bran emotonal learnng baed ntellgent controller. Energy Converon and Management, vol. 52, no. 1, pp [30] Ale Aghaee, S., Luca, C., &Amr Zadeh,. (2012). Applyng Bran Emotonal Learnng Baed Intellgent Controller (Belbc) to Multple Area Power Sytem. Aan Journal of Control, vol. 14, no. 6, pp [31] Mehraban, A. R., Luca, C., &Rohanan, J. (2006). Aeropace launch vehcle control: an ntellgent adaptve approach. Aeropace Scence and technology, vol. 10, no. 2, pp [32] Rouhan, H., et al. (2007). Bran emotonal learnng baed ntellgent controller appled to neurofuzzy model of mcro-heat exchanger. Expert Sytem wth Applcaton, vol. 32, no. 3, pp [33] Rav, R., &Mja, S., (2014), Degn of bran emotonal learnng baed ntellgent controller (BELBIC) for uncertan ytem, Advanced Communcaton Control and Computng Technologe 199

10 (ICACCCT), 2014 Internatonal Conference on, IEEE, pp [34] halghan, M. R., &hooban, M. H. (2014). A novel elf-tunng control method baed on regulated bobjectve emotonal learnng controller' tructure wth TLBO algorthm to control DVR compenator. Appled Soft Computng, vol. 24, no. 2, pp [35] Sarker, J., &Gowam, S. (2016). Optmal Locaton of Unfed Power Qualty Condtoner n Dtrbuton Sytem for Power Qualty Improvement. Internatonal Journal of Electrcal Power & Energy Sytem, vol. 83, no. 1, pp [36] Chatterjee, S., &Mukherjee, V. (2016). PID controller for automatc voltage regulator ung teachng learnng baed optmzaton technque. Internatonal Journal of Electrcal Power & Energy Sytem, vol. 77, no. 1, pp [37] Da, S., (2008), Functonal fractonal calculu for ytem dentfcaton and control, Sprnger-Verlag, Berln, Hedelberg. [38] Zaman, M., et al. (2009). Degn of a fractonal order PID controller for an AVR ung partcle warm optmzaton. Control Engneerng Practce, vol. 17, no. 12, pp [39] Jadd, Z., Muthukkumaraamy, V., &Sthraenan, E., (2013), Metaheurtc algorthm baed flow anomaly detector, Communcaton (APCC), th Aa-Pacfc Conference on, IEEE, pp [40] Shokr-Ghaleh, H., &Alf, A. (2014). Optmal ynchronzaton of teleoperaton ytem va cuckoo optmzaton algorthm. Nonlnear Dynamc, vol. 78, no. 4, pp [41] Balochan, S., &Ebrahm, E. (2013). Parameter optmzaton va cuckoo optmzaton algorthm of fuzzy controller for lqud level control. Journal of Engneerng, vol. 2013, no. 1, pp [42] Morén, J., &Balkenu, C. (2000). A computatonal model of emotonal learnng n the amygdala. From anmal to anmat, vol. 6, no. 1, pp [43] Markadeh, G., et al. (2011). Speed and flux control of nducton motor ung emotonal ntellgent controller. IEEE Tranacton on Indutry Applcaton, vol. 47, no. 3, pp [44] Jafar, E., et al. (2013). Degnng an emotonal ntellgent controller for IPFC to mprove the tranent tablty baed on energy functon. Journal of Electrcal Engneerng and Technology, vol. 8, no. 3, pp

11 نش رهی هوش مصنوعی و داده کاوی کنترل هوشمند عاطفی کسری بهینه برای سیستم تنظیم کننده ولتاژ اتوماتیک در سیستمهای قدرت مجید مرادی زیرکوهی گروه مهندسی برق دانشگاه صنعتی خاتم االنبیاء بهبهان بهبهان ایران. ارسال 0281/20/81 بازنگری 0281/82/02 پذیرش 0281/88/20 چکیده: در این مقاله یک کنترل کننده هوشمند عاطفی کسری بهینه برای تنظیم کننده ولتاژ اتوماتیک با استفاده از الگوریتم بهینه سازی فاخته پیشنهاد می- شود. تنظیم کننده ولتاژ اتوماتیک کنترل کننده اصلی درون سیستم تحریک است و هدف آن ثابت نگه داشتن ولتاژ تحریک ژنراتور میباشد. کنترل کننده پیشنهادی بر مبنای یادگیری عاطفی در مغز است که یک کنترل کننده خود تنظیم بوده و به کنترل کننده معروف است. نوآوری مقاله استفاده از کنترل کننده PID مرتبه کسری برای تولید سیگنال تحریک ورودی است در صورتی که در مقاالت منتشر شده از کنترل کننده PID استفاده شده است. یکی از مزایای کنترل کننده پیشنهادی وابسته نبودن به مدل سیستم است. کنترل کننده پیشنهادی یک کنترل کننده رضایت بخش از نقطه نظر سادگی طراحی سادگی پیاده سازی و حجم محاسبات کم میباشد. از طرفی از آنجا که عملکرد کنترل کننده پیشنهادی وابسته به مقدار پارامترهای آن دارد برای افزایش عملکرد آن از الگوریتم بهینه سازی فاخته برای تنظیم پارامترها استفاده شده است. از این رو یک تابع هزینه چند هدفه شامل اورشوت زمان نشست زمان صعود و خطای حالت ماندگار طراحی میشود. نتایج شبیه سازی نشان میدهد که روش پیشنهادی در حضور عدم قطعیتها دارای عملکرد بهتری نسبت به کارهای قبلی که از کنترل کننده PID استفاده شده است دارد. با اعمال کنترل کننده پیشنهادی زمان صعود و زمان نشست بترتیب کلمات کلیدی: %74 و %74 بهبود پیدا کرده است. کنترل کننده هوشمند عاطفی الگوریتم بهینه سازی فاخته کنترل کننده PID مرتبه کسری تنظیم کننده ولتاژ اتوماتیک.