Use of ANFIS Control Approach for SSSC based Damping Controllers Applied in a Two-area Power System

Transcription

1 Use of ANFIS Control Approach for SSSC base ampng Controllers Apple n a Two-area Power System. Mural, M. Rajaram 2 epartment of Electrcal an Electroncs Engneerng, Government College of Engneerng, Salem, Tamlnau, Ina. *mural36@yahoo.com 2 Anna Unversty, Chenna, Tamlnau, Ina. ABSTRACT In an nterconnecte power system, low frequency electromechancal oscllatons are ntate by normal small changes n system loas, an they become much worse followng a large sturbance. Flexble AC Transmsson System (FACTS) evces are wely recognze as powerful controllers for ampng power system oscllatons. The stanar FACTS controllers are lnear controllers whch may not guarantee acceptable performance or stablty n the event of a major sturbance. To overcome the rawbacks of conventonal controllers, ANFIS (Aaptve Neuro-Fuzzy Inference System) control scheme has been evelope n ths paper, an t has been apple for the external coornate control of seres connecte FACTS controllers known as Statc Synchronous Seres Compensators (SSSCs) employe n a two-area power system. In neuro-fuzzy control metho, the smplcty of fuzzy systems an the ablty of tranng n neural networks have been combne. The tranng ata set the parameters of membershp functons n fuzzy controller. Ths ANFIS can track the gven nput-output ata n orer to conform to the esre controller. Smulaton stues carre out n MATLAB/SIMULIN envronment emonstrate that the propose ANFIS base SSSC controller shows the mprove ampng performance as compare to conventonal SSSC base ampng controllers uner fferent operatng contons. eywors: ANFIS, FACTS, low frequency electromechancal oscllatons, MATLAB/SIMULIN, SSSC.. Introucton Wth ncreasng power transfer an heaver loang, power systems become graually more complex to operate an they may become less secure for rng out major power outages [, 2]. As a result, large power flows wth naequate control may be observe an excessve reactve power an large ynamc swngs may be experence n fferent parts of the system whch wll prevent the transmsson nterconnectons from beng fully utlze. Power system exhbts varous moes of oscllatons ue to nteracton among varous components. Most of the oscllatons are ue to synchronous generator rotors swngng relatve to each other. Stresse power systems are known to exhbt nonlnear behavor. Loa changes or faults are the man causes of power oscllatons. If the oscllatons are not controlle properly, t may lea to a total or partal system outage. If no aequate ampng s avalable, these oscllatons may sustan an grow to cause system separaton [3]. In the past three ecaes, power system stablzers (PSSs) have been extensvely use to ncrease the system ampng for low frequency oscllatons. The power utltes worlwe are currently mplementng PSSs as effectve exctaton controllers to enhance the system stablty [4]-[6]. However, there have been problems experence wth PSSs over the years of operaton. Some of these were ue to the lmte capablty of PSSs n ampng only local an not nter-area moes of oscllatons. In aton, PSSs can cause great varatons n the voltage profle uner severe sturbances an they may even result n leang power factor operaton an losng system stablty. Ths stuaton has necesstate a Journal of Apple Research an Technology 895

2 Use of ANFIS Control Approach for SSSC base ampng Controllers Apple n a Two area Power System,. Mural / revew of the tratonal power system concepts an practces to acheve a larger stablty margn, greater operatng flexblty, an better utlzaton of exstng power systems. In recent years, the concept of flexble ac transmsson systems (FACTS) brought racal changes n the power system operaton an control. The FACTS evces lnke to the mprovements n semconuctor technology opene new opportuntes for controllng power an enhancng the usable capacty of exstng transmsson lnes. As supplementary functons, ampng the nter-area moes an enhancng power system stablty usng FACTS controllers have been extensvely stue an nvestgate. In these years, Voltage Source Converter (VSC)-base seres connecte FACTS controllers, known as the new FACTS generaton, whch can nject a voltage wth controllable magntue an phase angle at the funamental frequency, are foun to be more capable of hanlng power flow control, transent stablty an oscllaton ampng enhancement. In a recent lterature [7], a novel methoology for tunng statc synchronous compensator (STATCOM) base ampng controller n orer to enhance the ampng of system low frequency oscllatons has been escrbe. Ths paper nvestgates the ampng capabltes of statc synchronous seres compensator (SSSC), whch s one of the VSC-base seres connecte FACTS controllers. SSSC s a sol-state controllable voltage source nverter that s connecte n seres wth power transmsson lnes. Wth the njecte voltage n quarature wth the lne current an the capablty of ynamcally changng ts reactance characterstc from capactve to nuctve, SSSC becomes an effectve tool for power flow control [8]. In aton, an auxlary stablzng sgnal can be supermpose on ts power flow control functon to mprove power system oscllaton stablty [9]. An attempt has been mae to apply hybr neurofuzzy approach for the coornaton between the conventonal power oscllaton ampng (PO) controllers for multmachne power systems. Wth the help of MATLAB, a class of aaptve networks, that are functonally equvalent to fuzzy nference systems, s propose. The propose archtecture s referre to as ANFIS (Aaptve Neuro-Fuzzy Inference System) []-[4]. In ths paper, ANFIS base SSSC controllers are evelope where each controller uses the spee of a synchronous machne an ts ervatve as the nputs. The ANFIS base SSSC controller uses a frst-orer Sugeno-type fuzzy logc controller whose membershp functons an consequences are tune by backpropagaton metho. Fuzzy rules an membershp functons of the controller can be tune automatcally by learnng algorthm. The propose technque s llustrate on a 3-machne, 9-bus power system. MATLAB/SIMULIN an fuzzy logc toolbox have been use for system smulaton. The results emonstrate that the propose self-learnng ANFIS base SSSC controller proves a goo ampng performance over a we range of operatng contons as compare to conventonal SSSC base ampng controllers. Ths paper s organze as follows. The snglemachne nfnte-bus (SMIB) an multmachne power system moels are escrbe n Secton 2. The esgn aspects of conventonal SSSC base ampng controller an fuzzy logc coornate SSSC base ampng controller, an the concept of ANFIS control scheme are scusse n Secton 3. Smulaton results an scussons are llustrate n Secton 4. Some conclusons are gven n Secton Power system moel The recommene state space moel for SSSC to stuy ynamc stablty of a sngle-machne nfntebus power system (SMIB) s gven by Eqn. () [5]. δ ω E q E f V C M 4 T A 5 T A 7 pu qu u vu m ω b M 2 M 3 T A 6 T A 8 T T A pc δ M ω qc E T q E f A vc V T C A 9 () 896 Vol., ecember 23

3 Use of ANFIS Control Approach for SSSC base ampng Controllers Apple n a Two area Power System,. Mural / where, A vm qm vu, qu, T A T o pm pu, M u m There are two conventonal SSSC base ampng controllers n the system; one nstalle between bus 5 an bus 7, an another between bus 6 an bus 9 respectvely. The system ata are gven n Appenx. δ ω pc q E f V C y E u 2 pm (2) In the above equatons, the output sgnal s assume to be the actve power elvere along the lne, so that y P e. Open-loop egen values can be calculate usng the state matrx of Eqn. (). The characterstcs of the SSSC nstalle n an n-machne power system are the same as the case of SMIB system. The Phllps-Heffron moel for ths system s erve from the followng lnearze equatons [5, 6]: I Q I T E Y q δ F q E q G q V H q m C (3) Y δ F E q I E q I E I Q Q q G V C (X Q Q X )I H m I (X Q Q (4) X )I (5) where the varables are n-orer vectors an the matrces are n-orer agonal matrces. Usng the above fferental equatons, the Phllps-Heffron moel for an n-machne power system can be expresse. It shoul be notce that n ths case, the -coeffcents are n-orer vectors [9, 6]. In ths paper, a 3-machne 9-bus power system moel as shown n Fgure s use to examne nter-area oscllaton control problem. In Fgure, the generator G s consere as reference bus. Ths two-area system, where Area conssts of generator G2 an Area 2 conssts of generator G3, s create especally for the analyss an stuy of the nter-area oscllaton problem [7]. The base MVA s an the system frequency s 5 Hz. Fgure. Sngle lne agram of a 3-machne 9-bus system. All mpeances are n per unt on MVA base 3. esgn methoology In ths secton, the esgn aspects of conventonal SSSC base ampng controller, fuzzy coornate SSSC controller an ANFIS control scheme are escrbe. 3. esgn of ampng controller The ampng controller s esgne to mprove the ampng torque. The structure of an SSSC base ampng controller [5] s shown n Fgure 2. It conssts of gan, sgnal wash-out an phase compensator blocks. The block of sgnal wash-out s a hgh pass flter that mofes the SSSC nput sgnal an prevents steay changes n actve power. Therefore, T W shoul have a large value to allow sgnals assocate wth actve power oscllatons to pass unchange. The value of T W s not crtcal an may be n the range of to 2 secons. Here t s assume to be equal to 3 secons. Fgure 2. Structure of SSSC base ampng controller Journal of Apple Research an Technology 897

4 Use of ANFIS Control Approach for SSSC base ampng Controllers Apple n a Two area Power System,. Mural / esgn of fuzzy logc coornate ampng controller Most of the FACTS base ampng controllers belong to the PI (Proportonal + Integral) type an work effectvely n sngle machne system [8]. However, the performance of the above mentone ampng controllers eterorates n multmachne power systems. The ampng performance of the FACTS base ampng controllers n multmachne power systems can be mprove by usng fuzzy coornate esgn [9]. The structure of the propose fuzzy coornaton controller s shown n Fgure 3, where the nputs are spee evaton of synchronous machnes an ther acceleraton. Thus, the conventonal ampng controllers are tune by usng fuzzy logc controllers. Fgure 3. SSSC base fuzzy-coornaton controller The fuzzy logc controller [2] comprses of four stages: fuzzfcaton, a knowlege base, ecson makng an efuzzfcaton. The fuzzfcaton nterface converts nput ata nto sutable lngustc values that can be vewe as label fuzzy sets. In ths paper, the nputs are fuzzfe nto seven fuzzy sets: Postve Bg (PB), Postve Meum (PM), Postve Small (PS), Zero (ZE), Negatve Small (NS), Negatve Meum (NM) an Negatve Bg (NB) as shown n Fgure 4. The knowlege base comprses knowlege of applcaton oman an attenant control goals by means of set of lngustc control rules. The ecson makng s the aggregaton of output of varous control rules that smulate the capablty of human ecson makng. In ths paper, the rules are trane usng ANFIS technology. Table shows the rule base of the fuzzy logc controller. To obtan a etermnstc control acton, a efuzzfcaton strategy s requre. efuzzfcaton s a mappng from a space of fuzzy control actons efne over an output unverse of scourse nto a space of nonfuzzy (crsp) control actons. There are fferent technques for efuzzfcaton of fuzzy quanttes such as Maxmum metho, Heght metho, an Centro metho. Here, Centro metho s use for efuzzfcaton. Fgure 4. Gaussan membershp functons Spee Acceleraton ev. NB NM NS ZE PS PM PB NB NB NB NB NB NM NS ZE NM NB NB NM NM NS ZE PS NS NB NM NS NS ZE PS PM ZE NM NM NS ZE ZE PM PM PS NM NS ZE ZE PS PM PB PM NS ZE PS PM PM PM PB PB ZE ZE PM PS PB PB PB Table. Rules extracte from the conventonal SSSC controller 3.3 ANFIS control scheme for SSSC The propose ANFIS controller also uses seven lngustc varables such as: Postve Bg (PB), Postve Meum (PM), Postve Small (PS), Zero (ZE), Negatve Small (NS), Negatve Meum (NM) an Negatve Bg (NB). The membershp functons are chosen to be Gaussan as shown n Fgure 4. The efuzzfcaton of the varables nto crsp outputs s teste by usng the weghte average metho. In MATLAB, the ANFIS etor graphcs user nterface s avalable n Fuzzy Logc Toolbox [3]. Usng a gven nput/output ata set, the toolbox constructs a fuzzy nference system (FIS) whose membershp functon parameters are ajuste usng ether a backpropagaton algorthm alone, or n combnaton wth a least squares type of metho. Ths allows the fuzzy systems to learn from the ata they are moelng. For the backpropagaton-base neurofuzzy approach, t nclues the Sugeno s moel wth the followng format: 898 Vol., ecember 23

5 Use of ANFIS Control Approach for SSSC base ampng Controllers Apple n a Two area Power System,. Mural / If the spee evaton error s ω an the acceleraton s ω, then U SSSC p ω q ω r (6) where, {, n*m} refers to the rule numbers; j {, n} refers to the spee evaton error terms n the fuzzy set; n, m refers to the number of terms generate; k {, m} refers to the acceleraton terms n the fuzzy set; {p,q,r } are the th consequent (SSSC output) parameters. The nput sgnals to the ANFIS controller for SSSC are ω an ω. In the ANFIS etor, the fuzzy nference can be generate usng two partton methos; gr parttonng an subtractve clusterng. Here, gr parttonng metho s use. For gr parttonng, t uses the Fuzzy C-means (FCM) clusterng ata clusterng technque. FCM s a ata clusterng algorthm n whch each ata pont belongs to a cluster wth a egree specfe by a membershp grae. After generatng the fuzzy nference, the generate nformaton escrbng the moel s structure an parameters of both the nput an output varables are use n the ANFIS tranng phase. Ths nformaton wll be fne-tune by applyng the hybr learnng or the backpropagaton schemes. The generate moel s of a frst-orer Sugeno s form an the generate rules are n the form escrbe n Eqn. (6). After ths stage, the membershp functons wll be ajuste to optmze the controller acton. 4. Smulaton results an scusson The robustness of the propose ANFIS base SSSC controllers to mprove the stablty of multmachne power systems s checke by conserng the 3- machne 9-bus power system moel shown n Fgure. The generator G s consere as reference bus. Each machne of ths power system has been represente by a fourth orer two-axs nonlnear moel. etals of the system ata [2, 22] are gven n Appenx. To evaluate the performance of the propose controller, the system response s compare wth conventonal SSSC base ampng (lag-lea) controller. The comparson s carre out uner two fferent kns of operatng ponts: (). Total real power of loa P =.7 p.u, Total reactve power of loa Q =.8 p.u, Termnal voltage V t =.5 p.u, an (). Total real power of loa P =.8 p.u, Total reactve power of loa Q =.9 p.u, Termnal voltage V t =.5 p.u) an power sturbances. Here, for llustraton, the frst set of operatng pont s consere. Wth conventonal SSSC base ampng (lag-lea) controllers, one nstalle between bus 5 an bus 7 an another between bus 6 an bus 9 respectvely, the system response curves ue to a power (or torque) sturbance of T m =. p.u an sturbance clearng tme of 5 secons are shown n Fgures 5-. From these Fgures, t s observe that the system ampng n Area an Area 2 s poor an the system s hghly oscllatory. Therefore, t s necessary to nstall ANFIS base SSSC controllers n orer to have goo ampng performance. The fuzzy rules are trane usng ANFIS technology. From the Fgures 5-, t s nferre that the ANFIS base SSSC controller can prove better ampng of the spee evaton, power angle, an real power oscllatons than the conventonal SSSC base ampng controller. The aaptve neuro-fuzzy base SSSC controller s able to track the system operatng contons, an thus as seen from the results shown n Fgures 5-, t s able to ajust an prove a unformly goo performance over a we range of operatng contons an sturbances. Spee evaton, (p.u).5 x Area Conventonal SSSC Controller ANFIS base SSSC Controller Fgure 5. Varaton of spee evaton for a torque sturbance of. p.u wth ANFIS base SSSC controller for Area (P =.7 p.u, Q =.8 p.u) Journal of Apple Research an Technology 899

6 Use of ANFIS Control Approach for SSSC base ampng Controllers Apple n a Two area Power System,. Mural / Area Conventonal SSSC Controller ANFIS base SSSC Controller.6.4 Area 2 Conventonal SSSC Controller ANFIS base SSSC Controller.2 Power angle, (p.u) Power angle, (p.u) Fgure 6. Varaton of power angle for a torque sturbance of. p.u wth ANFIS base SSSC controller for Area (P =.7 p.u, Q =.8 p.u) Fgure 8. Varaton of power angle for a torque sturbance of. p.u wth ANFIS base SSSC controller for Area 2 (P =.7 p.u, Q =.8 p.u) 5 x -4 4 Area 2 Conventonal SSSC Controller ANFIS base SSSC Controller Area Conventonal SSSC Controller ANFIS base SSSC Controller 3 74 Spee evaton, (p.u) 2 Real Power, (MW) Fgure 7. Varaton of spee evaton for a torque sturbance of. p.u wth ANFIS base SSSC controller for Area 2 (P =.7 p.u, Q =.8 p.u) Fgure 9. Varaton of real power for a torque sturbance of. p.u wth ANFIS base SSSC controller for Area (P =.7 p.u, Q =.8 p.u) 9 Vol., ecember 23

7 Use of ANFIS Control Approach for SSSC base ampng Controllers Apple n a Two area Power System,. Mural / Fgure. Varaton of real power for a torque sturbance of. p.u wth ANFIS base SSSC controller for Area 2 (P =.7 p.u, Q =.8 p.u) 5. Concluson To mprove the performance of conventonal SSSC base ampng controllers, ANFIS technology has been employe n ths paper. The fuzzy rules are trane usng ths technology. The propose metho has been evaluate on a 3-machne 9-bus power system. From the smulaton results, t s unerstoo that the ANFIS base SSSC controllers can prove better ampng of the spee evaton, power angle an real power oscllatons n terms of reuce settlng tmes than the conventonal SSSC base ampng controllers uner fferent operatng contons. The ANFIS control scheme s easy to tune an qute robust. Also such a nonlnear aaptve SSSC controller wll yel better an fast ampng uner small an large sturbances even wth changes n system operatng contons. Appenx (). Reuce Y- bus Matrx: Real P ower, (M W ) j j.53.2 j.226 Area j j j.88.2 j j j2.368 Lag-lea NFSSSC (2). Generators ata: Generator G2: Rate MVA = 92, Rate voltage = 8 kv, H(s) = 6.4, T 6, T q.535, x.8958, x.98, xq.8645, xq.969. Generator G3: Rate MVA = 28, Rate voltage = 3.8 kv, H(s) = 3., T 5.89, T q.6 x.325,, x.83, xq.2578, xq.25. (3). Excter ata: a a2 4, T a T a2.5. (4). SSSC ampng controller ata: T a., T 2a.5, Tw 3, Saturaton: Upper lmt =.5, Lower lmt = -.5. (5). Transmsson lne ata: Bus No. Impeance R X (6). Shunt amttances ata: Bus No. Amttance G B Note: All mpeance an amttance values are n per unt on MVA base. All tme constants are n secons. Journal of Apple Research an Technology 9

8 Use of ANFIS Control Approach for SSSC base ampng Controllers Apple n a Two area Power System,. Mural / References [] P. Asare et al. An Overvew of Flexble AC Transmsson Systems, ECE Techncal Reports PURUE Unversty, 994. [2] J. Paserba, How FACTS Controllers Beneft AC Transmsson Systems, Transmsson an strbuton Conference an Exposton, PES 23. [3] Graham Rogers, Power System Oscllatons, luwer Acaemc Publshers, Boston, 22. [4] T. Hyama, Robustness of Fuzzy Logc Power System Stablzers Apple to Multmachne Power System, IEEE Transactons on Energy Converson, vol. 9, no. 3, 994, pp [5] A. S. Venugopal et al., An Aaptve Neuro Fuzzy Power System Stablzer for ampng Inter-area Oscllatons n Power Systems, Proceengs of 36th the Southeastern Symposum on System Theory, 24, pp [6]. Mural an M. Rajaram, Neuro-Fuzzy Base Power System Stablzers for ampng Oscllatons n Mult-machne Power Systems, Journal of Electrcal Engneerng, vol.2, no. 2, June 22, pp [7]. A. Safar et al., Controller esgn of STATCOM for Power System Stablty Improvement Usng Honey Bee Matng Optmzaton, Journal of Apple Research an Technology, vol., no., February 23, pp , [8] L. Gyugy et al., Statc Synchronous Seres Compensator: A Sol-State Approach to the Seres Compensaton of Transmsson Lnes, IEEE Transactons on Power elvery, vol. 2, no., 997, pp [9] H. F. Wang, esgn of SSSC ampng Controller to Improve Power System Oscllaton Stablty, AFRICON, IEEE, no., 999, pp [] W. Qao an R. G. Harley, Inrect Aaptve External Neuro-control for a Seres Capactve Reactance Compensator Base on a Voltage Source PWM Converter n ampng Power Oscllatons, IEEE Transactons on Inustral Electroncs, vol. 54, no., February 27, pp [] J. R. Jang, ANFIS Aaptve-netwok-Base Fuzzy Inference System, IEEE Transactons on Systems, Man an Cybernetcs, vol. 23, no. 3, 993, pp [3] Fuzzy Logc Toolbox, Avalable: com. [4] T. R. Sumthra an A. Nrmal umar, Elmnaton of Harmoncs n Multlevel Inverters Connecte to Solar Photovoltac Systems Usng ANFIS: An Expermental Case Stuy, Journal of Apple Research an Technology, vol., no., February 23, pp [5] A. azem et al., Optmal Selecton of SSSC Base ampng Controller Parameters for Improvng Power System ynamc Stablty Usng Genetc Algorthm, Iranan Journal of Scence & Technology, Transacton- B, Engneerng, vol. 29, no. B, 25, pp. -. [6] H. F. Wang et al., A Unfe Moel for the Analyss of FACTS evces n ampng Power System Oscllatons, II. Mult-machne Power Systems, IEEE Transactons on Power elvery, vol. 3, no. 4, 998, pp [7] Y. Y. Hsu an C. R. Chen, Tunng of Fuzzy Power System Stablzers Usng Artfcal Neural Network, IEEE Transactons on Energy Converson, vol. 6, no. 4, 99, pp [8] Ljun Ca an Istvan Erlch, Fuzzy Coornaton of FACTS Controllers for ampng Power System Oscllatons, Proc. of Moern Electrc Power Systems-22. [9] P.. ash et al., ampng Mult-moal Power System Oscllaton Usng a Hybr Fuzzy Controller for Seres Connecte FACTS evces, IEEE Transactons on Power Systems, vol. 5, no. 4, November 2, pp [2] Neeraj Gupta an Sanjay. Jan, Comparatve Analyss of Fuzzy Power System Stablzer usng fferent Membershp Functons, Internatonal Journal of Computer an Electrcal Engneerng, vol. 2, no. 2, Aprl 2, pp [2] P. M. Anerson an A. A. Foua, Power System Control an Stablty, Wley-IEEE Press, 23. [22] Y. N. Yu, Electrc Power System ynamcs, Acaemc Press, 983. [2] S. P. Ghoshal, Mult-Area Frequency an Te-Lne Power Flow Control wth Fuzzy Logc Base Integral Gan Scheulng, IE (I) Journal-EI, vol. 84, ecember 23, pp Vol., ecember 23