TELK OMNIKA Indonesian Journal of Electrical Engineering V ol. 14, No . 3, J une 2015, pp . 363 375 DOI: 10.11591/telk omnika.v12.i9.4678 363 Fuzzy Logic PSS Assisted b y Neighboring Signals to Mitigate the Electr omec hanical W a ve Pr opa gation in P o wer Systems Mahmoud N. Ali * F aculty of Engineer ing, Shoubr a, Benha Univ ersity 108 Shoubr a st., Cairo , Egypt. * e-mail: mahmoud.nour@f eng.b u.edu.eg Abstract This paper deals with the mitigation of e lectromechanical w a v e propagation in po w er systems . Dif- f erent con v entional controllers addressed this prob lem, such as the con v entional PSS and the con v entional fuzzy logic PSS . In this paper , the fuzzy logic PSS is assisted b y auxiliar y signals from the fuzzy logic PSS of the interconnected machines to augment the damping of electromechanical w a v e propagation and the associated oscillations . The neighbor ing machines speed d e viation and its der iv ativ es signals are e xploited through the fuzzy logic PSS to assist the local fuzzy logic PSS . The disturbance propagation and reflection phenomena are considered in the desig n of the adopted str ategy . The efficacy of the proposed assistance of the con v entional fuzzy logic PSS are e xamined through diff erent sim ulation results . K e yw or ds: Electromechanical w a v e propagation, fuzzy logic PSS , interconnected machines Cop yright c 2015 Institute of Ad v anced Engineering and Science . All rights reser v ed. 1. Intr oduction P o w er systems are contin ually subjected to diff erent co ntingencies and r andom e v ents . The po w er balance betw een gener ation and loads is essential to get a stab le oper ation of a po w er system. The mismatch betw een the gener ator mechanical and el ectr ical po w er f orces the gener ator rotor to de viate from the synchronous ref erence fr ame . This elect romechanical distur- bance propagates throug h the entire netw or k as an electromechanical w a v e with a cer tain speed of propagation [1] [2]. Sim ulation results [3] [4] and e xper imental obser v ations [5] emphasiz ed this phenomenon, where its speed of propagation depends on the gener ators and tr ansmission system par ameters [2]. The propagat ion of the electromechanical w a v e stresses diff erent equipments in po w er systems , e .g. gener ators and tr ansf or mers . The protection systems ma y be subjected to a tem- por ar y violation of their limits which ma y cause une xpected gener ator and/or tr ansmission line tr ipping, and consequently , the cascading f ailure occurs [6]. Pre v entiv e and emergency control str ategies w ere used to a v oid the dr a wbac ks of the electromechanical w a v e propa gation [7] [8]. The pre v entiv e control str ategies urge to oper ate the system with high secur ity margins to decrease the possibility of gener ator tr ipping. The emer- gency control str ategies tak e th e suitab le actions in case of sensing the in itiation of a disturbance to lessen the potential eff ect of its propagation [8]. The con v entional po w er system stabiliz er (PSS) has been used man y y ears ago to dampen electromechanical oscillations in po w er systems . It acts through the e xcitation system in such a w a y that the speed de viation gener ates a component of electr ical torque to assist the damping torque , where the lac k of sufficient damping torque results in o sc i llator y instability [9]. PSS is also used to mitigate the spatial propagation of the electromechanical disturbance in po w er sys- tems [10]. Based on wide area measurements (W AM), some impro v ements of the con v entional PSS w ere proposed through centr aliz ed and decentr aliz ed str ategies to m itigate the electromechanical Receiv ed F ebr uar y 19, 2015; Re vised Ma y 1, 2015; Accepted Ma y 14, 2015 Evaluation Warning : The document was created with Spire.PDF for Python.
364 ISSN: 2302-4046 disturbance propagation in po w er systems [11] [12]. The con v entional fuzzy logic po w er sys- tem stabiliz er (FLPSS) w as eff ectiv ely emplo y ed to e xt inguish the propagated electromechanical disturbance in po w er systems [12]. In [13], a z ero reflection controller str ategy to quench the propagation of electromechani- cal disturbance w as proposed. The str ategy in [13] w as analogous to the impedance matching to inhibit reflection of tr a v eling electromagnetic w a v es in tr ansmission lines . As the disturbance occurs at an y place of the system tends to propagate to the entire system and tends to be reflected upon reaching the system boundar ies , the proposed str ategy in this paper e xploits this nature of th e disturbance . As the speed de viation gener ates a propor- tional component of electr ical torque through the con v entional PSS or through the con v entional fuzzy logic PSS , the neighbor ing speed de viation signal cou ld be e xploited, through a FLPSS , to gener ate electr ical torque to suppor t damping of the coming and reflected disturbance . F rom the vie wpoint of disturbance location, the nearest to the first place of disturbance are the f oremost machines , and the late s t machines are the fur thest. Th rough the FLPSS , the speed de viation signals of the f oremost machines can assist the damping of the propagated disturbance in the later machines . With the order re v ersed, t he speed de viation signals of the latest machines also can assist the f oremost machines , through the FLPSS , to atten uate the reflected disturbance . The paper is organiz ed as f ollo ws . Section 2. introduces the system modeling and f or- malization of the prob lem of electromechanical w a v e propagation. Section 3. demonstr ates the proposed str ategy to impro v e the mitigation of electromechanical w a v e propagation. Section 4. presents a perf or mance e v aluation of the proposed str ategy . Finally , Section 5. concludes . The appendix presents the par ameters of the benchmar k po w er system used in the sim ulations . 2. System modeling and pr ob lem f ormalization This section introduces the detailed gener ator modeling and the fuzzy logic based po w er system stabiliz er control. Some dr a wbac ks of the the prob lem of electromechanical w a v e propa- gation are e xplained. 2.1. Generator modeling Diff erent models can represent synchronous machines depending on the deg ree of de- tails [9] [14]. The adopted detailed model ar e represented b y the f ollo wing dynamic equations [14]: _ E 0 q = 1 T 0 d 0 ( E 0 q + ( x d x d ) i d + E f d ) (1) _ E 0 d = 1 T 0 q 0 ( E 0 d ( x q x q ) i q ) (2) _ = ! 0 ! (3) _ ! = 1 2 H ( T m T e D ! ) (4) T e = E 0 d i d + E 0 q i q + ( x d x q ) i d i q (5) i q = < e f 1 r a + j x [( E 0 q + j E 0 d ) ( v q + j v d )] g (6) i d = = m f 1 r a + j x [( E 0 q + j E 0 d ) ( v q + j v d )] g (7) where E 0 q is the q-axis inter nal v oltage in pu , E d 0 is the d-axis inter nal v oltage in pu , v q is the q-axis ter minal v oltage in pu , v d is the d-axis ter minal v oltage in pu , i q is the q-axis current in pu , i d is the d-axis current in pu , T m is the mechanical torque in pu , T e is the electr ical torque in pu , E f d is the field v oltage in pu , is the rotor angle in r ad , ! is the speed de viation in pu , ! 0 is the rotor r ated angular speed in r ad=s , x d is the d-axis tr ansient reactance in pu , x = x d = x q , x d is TELK OMNIKA V ol. 14, No . 3, J une 2015 : 363 375 Evaluation Warning : The document was created with Spire.PDF for Python.
TELK OMNIKA ISSN: 2302-4046 365 the d-axis synchronous reactance in pu , x q is the q-axis tr ansient reactance in pu , x q is the q-axis synchronous reactance in pu , r a is the gener ator inter nal resistance in pu , T 0 d 0 is the ope n circuit d-axis time constant in s , T 0 q 0 is the open circuit q-axis time constant in s and D is the damping constant in pu . 2.2. Generator contr ol and fuzzy logic PSS T o counter act the rotor angle and speed de viations , induced from the gener ator po w er imbalance , tw o pr incipal controls ha v e been e v olv ed f or synchronous gener ato rs , which are the pr ime mo v er control and e xcitation control. The mathematical models of these controllers can be f ound in [9] [14]. Con v entionally , PSS is added to the e xcitation system to dampen po w er system oscilla- tions . A fine tuning of PSS par ameters giv es a satisf actor y atten uation of the electromechanical disturbance propagation in po w er systems [10] [12] . Fuzzy logic po w er system stabiliz er can replace the con v entional PSS and can pro vide more satisf actor y damping f or diff erent modes of electromechanical oscillations [15] [16] and f or the disturbance propagation in po w er systems [12]. The con v entional PSS emplo ys fix ed par ameter model based on system linear ization or optimization methods . Such a fix ed-par ameter PSS is widely used in po w er systems and has made a g reat contr ib ution in enhancing po w er system dynamics [17]. As po w er systems are dynamic systems and their oper ation is of a stochastic nature , the con v entional fix ed-par ameter PSS can be replaced b y a fuzzy logic based PSS [18]. Fuzzy logic is con s i dered as a po w erful tool in encounter ing challenging prob lems in po w er systems because of its capability to handle imprecise , v ague or ’fuzzy’ inf or mation [19]. Fuzzy logic implements human e xper iences and pref erences with the adjustment of membership functions and fuzzy r ules . Fuzzy membership functions can ha v e diff erent shapes , e .g., tr iangular , tr apez oidal or Gaussian, depending on the pref erence and/or the e xper ience of the designer . The fuzzy r ules , which descr ibe relationships in a linguistic sense , are typically wr itten as antecedent consequent pairs of (IF-THEN) statements . A b loc k diag r am of a fuzzy controller is sho wn in Figure 1 [20], in which the input and the output of the fuzzy controller are cr isp . The fuzzification process con v er ts each piece of input data to deg rees of membership . The r ule base introduces the designer’ s e xper ience in linguistic relationships . In the inf erence engine , the application of an implication method and the agg regation of all outputs , related to the fuzzy r ules , are perf or med to get the output fuzzy set. The resulting fuzzy set m ust be con v er ted to a n umber that can be sent to the process as a control signal, which is called the defuzzification oper at ion. All the processes in the dashed bo x (see Figure 1) are called fuzzy inf erence system [21] [22]. I nput   ( c r i s p)   F uz z i f i c a t i o n O ut pu t   ( c r i s p)   D e f uz z i f i c a t i o n     R ul e   ba s e   I n f e r e nc e     e ng i ne   Figure 1. Bloc k diag r am of a fuzzy controller . Empir ical kno wledge and engineer ing intuition pla y an impor tant role in choosing the lin- guistic v ar ia b les and their membership functions . Based on the pre vious e xper iences introduced in [23] [24] [18], the input signals are chosen as speed de viation ( ! ) and its der iv ativ e ( _ ! ) and the output signal ( V f l pss ) is added to the e xcitation system ref erence v oltage . The speed Evaluation Warning : The document was created with Spire.PDF for Python.
366 ISSN: 2302-4046 de viation der iv ativ e ( _ ! ) can be calculated from the f ollo wing relation [23]: _ ! = ! n +1 ! n t (8) where ! n is the speed de viation at step n , ! n +1 is the speed de viation at step n + 1 and t is time of step of integ r ation. 2.3. The electr omec hanical wa ve pr opa gation pr ob lems The phenomenon of electromechanical w a v e propagation has significant eff ects on the po w er systems and their protection systems . As the disturbance propagation imposes a de viation of the gener ator’ s rotor angle from its steady-state v alue , the po w er tr ansf ers in the netw or k is disturbed. The po w er flo w betw een b us i and b us j in a pure inductiv e line is giv en b y th e f ollo wing relation: P f ij = P max sin ij 0 (9) where P f ij is the po w er flo w betw een b us i and b us j , ij 0 is the steady state angle diff erence and P max is the maxim um po w er tr ansf er , which is giv en b y the f ollo wing relation: P max = j V g i jj V g j j X ij (10) where j V g i j and j V g j j are the magnitude of the inter nal v oltage of the tw o machines i and j and X ij is the reactance of the connecting line . Thus , the de viation in rotor angle results in a change in po w er tr ansf er , which could be significant and could impact the po w er system oper ation [12]. The eff ects of the electromechanical disturbance propagation on protection systems w ere presented in [25], where such a propagation often e xposes remote rela ying systems to f alse re- sponses . The de viation in the po w er flo w causes a de viation in the current, which aff ects the op- er ation of the o v er-current rela ys and also aff ects the apparent impedance which, consequently , aff ects the oper ation of the distance rela ys [25]. The propagation of the electromechanical dis- turbance ma y tr ip o v er-current rela ys and/or distance rela ys because of the tr ansient violation of their pic k-up settings . This unplanned tr ipping of protection rela ys could disconnect one or some components of the po w er system, which ma y lead to other disturbances [12]. 3. Pr oposed strategy of assisted FLPSS In the proposed str ategy , the phenomenon of propagation and reflection of electrome- chanical w a v e in po w er systems a re e xploited. The str ategy is based on the injection of the output signal of neighbor ing FLPSS as e xtr a inputs to the e xcitation system besides the FLPSS of the local machine to impro v e the dampin g of electromechanical w a v e propagation and the associated oscillations . More specifically , the e xcitation system of machine ( n ) has m ulti FLPSS signals , which are the local FLPSS signal ( V f l pss n ) and the FLPSS signals of all electr ically connected machines ( V f l pss m ) w eighted b y a gain K m as sho wn in Figure 2. The same str ategy is applied at all other machines . Basing on the phenomenon of disturbance propagation and i ts reflection upon reaching system boundar ies , this str ategy is de v eloped. Suppose that machine n is electr ically connected to machine m (which represents one or more machines) and it is firstly subjected to a disturbance . The FLPSS of machine n add e xtr a damping through the e xcitation system of machine m besides the local FLPSS of this machine to coun ter act the propagated disturbance . Also , the FLPSS of machine m add e xtr a damping besides the local FLPSS of machine n , through the e xcitation system, to counter act the reflected disturbance . This inter-assistance betw een the interconnected machines helps in damping the propagated and the reflected disturbance . Nor mally , a time dela y is implied f or the disturbance to reach diff erent machines , where , the f oremost aff ected machines are the nearest to the f ault location and the last aff ected machines are the most f ar . This can TELK OMNIKA V ol. 14, No . 3, J une 2015 : 363 375 Evaluation Warning : The document was created with Spire.PDF for Python.
TELK OMNIKA ISSN: 2302-4046 367 _ _ n m Figure 2. A schematic diag r am of the proposed str ategy of modifications of the fuzzy logic PSS . be disregarded f or neighbor ing machines with small electr ical distance , where the time dela y is small. T o adjust the gains K n and K m , an optimization toolbo x of MA TLAB (FMINCON) is used to minimiz e an objectiv e function, F obj , which is giv en as f ollo ws: F obj = n X i =1 Z t f 0 ( ! i ) 2 dt (11) where n is the total n umber of machines in oper ation and t f is the per iod of sim ulation. 4. P erf ormance e v aluation of the pr oposed strategy This section presents the perf or mance e v aluation of the proposed str ategy through the sim ulation of the unif or m tw o-dimensional (2-D) model [11]. The perf or mance is quantitativ ely assessed through perf or mance inde x es . 4.1. unif orm 2-D system A gener al regular tw o-dimension (2-D) g r id is sho wn in Figure 3. It consists of 8 8 nodes , where one gener ator and a shunt load ar e connected to each node . The model has identical tr ansmission lines connecting tw o adjacent nodes Z tl , and identical loads at each node , which is modeled as constant impedance Z l . The disturbance is initiated through a sequence of e v ents as f ollo ws: Bef ore t = 0 : 1 s , the system is in steady state . At t = 0 : 1 s , the gene r ator (1 ; 1) , which has coordinates in the g r id x = 1 , y = 1 , is lost. The initiated disturbance propagates to the entire netw or k in tw o dimensions x and y . The propagated disturbance in rotor angle and rotor speed of the 64 gener ators of the 2-D model are sho wn in Figures 4 and 5, where all the machines are equipped with the pr ime mo v er controller and the e xcitation system without FLPSS . These Figures sho w that the distur- bance occurs at the first location causes the electromechanical oscillation near this location. It propagates allo v er the netw or k causing oscillations at each gener ator and upon reaching bound- ar ies , it is reflected bac k to the first place causing a ne w disturbance propagating to the entire netw or k. 4.2. Application of the con ventional FLPSS to the unif orm 2-D system F or the studied test system, se v en linguistic v ar iab les are proposed f or each input and output v ar iab les , which are: LP (large positiv e), MP (medium positiv e), SP (small positiv e), VS (v er y small), SN (small negativ e), MN (medium negativ e) and LN (large negativ e) [23]. The typical membership functions ma y be chosen as tr iangular , tr apez oidal, bell shaped, etc. In the sim ulation of the application of the FLPSS to the 2-D test system, a bell shaped mem- Evaluation Warning : The document was created with Spire.PDF for Python.
368 ISSN: 2302-4046 Figure 3. P o w er system netw or k sim ulated arr anged in a g r id of tw o dimensions . bership functions are chosen f or all t he inputs and output v ar iab les . The membership functions of the tw o inputs ( ! and _ ! ) and the output ( V f l pss ) are sho wn in Figures 6 and 7, respectiv ely . A set of r ules , which defines the relation betw een the inputs and the output of the FLPSS is e xtr acted from the pre vious e xper ience of designing th e PSS [18] [23] [24], which are presented in T ab le 1. T ab le 1. Decision tab le f or the output of fuzzy logic PSS . P P P P P P P P ! _ ! LN MN SN VS SP MP LP LP VS SP MP LP LP LP LP MP SN VS SP MP MP LP LP SP MN SN VS SP SP MP LP VS MN MN SN VS SP MP MP SN LN MN SN SN VS SP MP MN LN LN MN MN SN VS SP LN LN LN LN LN MN SN VS There are diff erent methods f or finding the fuzzy output, e .g., minim um-maxim um and maxim um-product methods [22] [23]. In these sim ulations , the minim um-maxim um method is adopted. The fuzzy output of each r ule is obtained and all outputs are agg regated and defuzzified to get the final output of the FLPSS . There are diff erent techniques f or defuzzification of fuzzy quantities such as the maxim um method, the height method, and the centroid method, where the later is the most widespread one , and it is adopted in these sim ulations . 4.3. Application of the pr oposed strategy to the unif orm 2-D system T o e v aluate the application of the proposed str ategy , the rotor angle and rotor speed de viation responses are e xamined to e v aluate the eff ectiv eness of this str ategy . First, the responses of all rotor angles , ref erred to the angle of machine (8,8), and rotor TELK OMNIKA V ol. 14, No . 3, J une 2015 : 363 375 Evaluation Warning : The document was created with Spire.PDF for Python.
TELK OMNIKA ISSN: 2302-4046 369 Figure 4. Propagation of electromechanical w a v e in rotor angle , ref erred to the rotor angle of machine (8 ; 8) , throughout the 2-D test system without FLPSS . Figure 5. Propagation of electromechanical w a v e in rotor speed throughout the 2-D test po w er system without FLPSS . speed de viation, when applying the con v entional FLPSS , are sho wn in Figures 8 and 9 respec- tiv ely . When applying the proposed str ategy , introduced in section 3., the ne w responses of machines’ angles and speed de viations are sho wn in Figures 10 and 11 respectiv ely . Putting Figures 8, 9, 10 and 11 in perspectiv e , sho ws the impro v ement achie v ed when applying the proposed str ategy of the FLPSS . T o allo w a quantitativ e e v aluation of the the proposed str ategy , perf or mance inde x es ( t ) and ! ( t ) are proposed. The y are defined as f ollo ws: Evaluation Warning : The document was created with Spire.PDF for Python.
370 ISSN: 2302-4046 −0.03 −0.02 −0.01 0 0.01 0.02 0.03 0 0.2 0.4 0.6 0.8 1 ω Degree of membership LN MN SN VS SP MP LP −0.4 −0.3 −0.2 −0.1 0 0.1 0.2 0.3 0.4 0 0.2 0.4 0.6 0.8 1 d( ω )/dt Degree of membership LN MN SN VS SP MP LP Figure 6. Membership function of the fuzzy inp uts (the speed de viation ( ! ) and the der iv ativ e of speed de viation ( _ ! )). −0.2 −0.15 −0.1 −0.05 0 0.05 0.1 0.15 0.2 0 0.2 0.4 0.6 0.8 1  V flpss Degree of membership LN MN SN VS SP MP LP Figure 7. Membership function of the fuzzy output ( V f l pss ). TELK OMNIKA V ol. 14, No . 3, J une 2015 : 363 375 Evaluation Warning : The document was created with Spire.PDF for Python.
TELK OMNIKA ISSN: 2302-4046 371 Figure 8. Rotor angle response f or all machine (ref err ed to machine(8,8)) with applying the con- v entional FLPSS . Figure 9. Rotor speed de viation of all machines with applying the con v entional FLPSS . ( t ) = n X i =1 ( i ( t ) C O I ( t )) 2 (12) ! ( t ) = n X i =1 ( ! i ( t ) ! C O I ( t )) 2 (13) These perf or mance inde x es consider tw o char acter istics; First, the div ergence from the center of iner tia ( C O I ) of rotor angle and speed de viation. Second, the speed of the controller to mitigate the disturbance propagation i.e ., the time required f or the perf or mance inde x to reach z ero . Evaluation Warning : The document was created with Spire.PDF for Python.
372 ISSN: 2302-4046 Figure 10. Rotor angle of all machines (ref erred to machine(8,8)) when applying the proposed str ategy of FLPSS . Figure 11. Rotor speed de viation of all machines when applying the proposed str ategy of FLPSS . T o allo w quantitativ e compar ison betw een the con v entional PSS , the con v entional FLPSS and the prop osed str ategy of FLPSS , the perf or mance inde x es f or speed de viation and f or rotor angle are presen ted in Figures 12 and 13 respectiv ely . A lthough the con v entional FLPSS in- troduces some damping f or the propagation of electromechanical w a v e , the proposed str ategy of FLPSS add an additional damping f or this propagation and f or the associated oscillations . TELK OMNIKA V ol. 14, No . 3, J une 2015 : 363 375 Evaluation Warning : The document was created with Spire.PDF for Python.