September 214, Vol. 1, No. 1 Desgnng Intellgent Load-Frequency Controllers for Large-Scale Mult-Control- Interconnected Power Systems Nguyen Ngoc-Khoat 1,2,* 1 Faculty of Automaton Technology, Electrc Power Unversty, Hano, Vetnam 2 School of Energy Scence and Engneerng, Unversty of Electronc Scence and Technology of Chna, Chna * Correspondng author, Emal: khoatnn@epu.edu.vn, Phone: +8613 8851248 Dao Th-Ma-Phuong 3,4 3 Faculty of Electrcal Engneerng, Hano Unversty of Industry, Hano, Vetnam 4 College of Electrcal and Informaton Technology, Hunan Unversty, Changsha, Hunan, Chna Emal: daophuong@hau.edu.vn, Phone: +841684 439292 DOI: 1.67/IJARCSEE/v1-1/1382 URL: http://dx.do.org/1.67/ijarcsee/v1-1/1382 Abstract Ths work concentrates on desgnng an effectvely ntellgent control methodology to mantan the network frequency aganst load varatons n a mult-control-area nterconnected power system. Snce conventonal controllers (e.g., Integral, PI and PID) have only obtaned poor control performances, such as hgh overshoots and long settlng tmes, they should be replaced wth the ntellgent regulators usng modern control technques. Fuzzy logc-based control strategy, whch has been one of the most effectvely ntellgent methodologes, s able to perfectly substtute for such conventonal controllers when dealng wth the network frequency stablzaton problem. Ths paper proposes a type of PD-based fuzzy logc controllers wth the proper 49-rule base n order to solve thoroughly the load-frequency control ssue. The effectveness and outperformance of such ntellgent load-frequency controller over conventonal regulators wll be verfed through numercal smulaton processes. Such smulaton processes are performed n a fve-control-area nonreheat thermal electrc power grd model, a typcal case study of the large-scale nterconnected power networks, usng MATLAB/Smulnk package verson 213a. Keywords Load-frequency control, te-lne bas control strategy, conventonal regulator, ntellgent controller, PD-based FLC Symbols control-area order, = 1, 2, 3, 4, 5 f network frequency, Hz f n nomnal network frequency, f n = 5Hz P te, te-lne power, p.u. P D, load varaton, p.u. f change of the network frequency, p.u. T g, governor tme constant, s T t, non-reheat thermal turbne tme constant, s K p, generator-load unt coeffcent, Hz/p.u.MW T p, generator-load unt tme constant, s te-lne synchronzng factor, s T j 38
September 214, Vol. 1, No. 1 P te, B M D R te-lne power devaton, p.u. bas factor of frequency, MW/p.u.Hz generator nerta constant, p.u. load dampng factor, p.u. MW/Hz speed regulaton, Hz/MW I. Introducton Load n large-scale nterconnected electrc power grds, dependng upon customers, s changng contnually over tme (Had 21; Bevran & Hyama 211). Ths leads to the energy mbalance between the generaton and load, causng the devaton of system frequency from ts nomnal value (5Hz or 6Hz) (Wood et al. 213; Kundur 1994; Rchard 21). Due to the proportonal relatonshp between the workng frequency and actve power n an electrc power grd, such devaton usually causes the change of the generaton demand. As a result, t s hghly necessary to desgn robust and effectve Load- Frequency Control (LFC) strateges to automatcally control the electrc generaton for the grd. The control strategy, especally n a large-scale mult-area power system, s usually called as Automatc Generaton Control (AGC) (Chongxn et al. 213; Ibrabeem et al. 25). As an mportant part of the AGC scheme, the LFC therefore ams to protect both the workng frequency as well as the power nterchange of the te-lnes n accordance wth the scheduled dspatch, ensurng the stablty, relablty and economy of an electrc power grd. To obtan the objectves of the LFC strategy, the transent oscllatons of both the system frequency bas and te-lne power change, whch affect all power system devces, need to be damped effcently enough to recover quckly the steady-state of the network after load varatons (Shayegh et al. 29). As a result, desgnng an effectve LFC strategy plays an mportant role on the successful operaton, stablty and relablty of large-scale nterconnected power networks n realty. When conductng the LFC problem, based on the te-lne bas control strategy, two categores of controllers have been appled recently, ncludng conventonal and ntellgent regulators (Kundur 1994; Rchard 21; Shash et al. 213; Hamed et al. 213). Bascally, the conventonal control methodologes employng tradtonal regulators (.e., Integral, PI or PID), have been ntally adopted to extngush the transent oscllatons of both the network frequency and te-lne power devatons. Nevertheless, when applyng these controllers, a large-scale power network, whch can be consdered as a nonlnearcomplcated control system, has only obtaned hghly poor performances, such as large overshoots and long settlng tmes. These undesrable control ndces may strongly affect the operaton and stablty of an electrc power grd (Shayegh et al. 29; Hamed et al. 213; Tan 21). In order to overcome the above drawbacks, ntellgent controllers usng modern control technques, e.g., fuzzy logc (FL), have wdely been nvestgated recently. Fuzzy logc controllers (FLCs), whch have been appled effcently n many control systems, can be employed to carry out the LFC strategy due to the followng reasons (Shash et al. 213; Hassan et al. 1993; Chown et al. 1998): (a) FL s a thnkng process of users ncorporated n control strategy, and hence t s not necessary to know clearly and fully parameters of the control system, (b) FLCs can utlze effcently the ncomplete nformaton to make a good control decson whch only depends on the knowledge of experts, and (c) wth FL rules, t s able to set up successfully an HMI (Human Machne Interface) whch s hghly useful for the nteracton property of a modern control scheme. 39
September 214, Vol. 1, No. 1 Obvously, the most domnant advantage of the FLCs s that the control parameters can be changed fast enough to respond well to the dynamc varatons of the system. Ths s because none of parameters may be needed to estmate accordng to the workng prncple of the FL archtecture. As a result, by applyng the FL-based LFC controllers, the man characterstcs as mentoned earler can be sgnfcantly mproved n order to acheve effcently the objectves of the control strategy (Santos et al. 24; Demroren & Yesl 24; Subbaraj et al. 27). Ths paper wll nvestgate four categores of controllers, namely Integral (I), PI, PID, and PD-type FLC based on the te-lne bas control strategy to conduct the LFC problem. The frst three loadfrequency regulators are canddates of conventonal methods. The last control archtecture based on FL technque wll be proposed n an effort to further mprove the control qualty of the LFC soluton. In ths study, a fve-area non-reheat electrc power grd model based on the te-lne bas control strategy s bult frst to mplement such four LFC controllers. Subsequently, a numercal smulaton process applyng dfferent LFC strateges to such nterconnected electrc power grd model wll be realzed to valdate the robustness as well as the effectveness of the proposed control archtecture. The results obtaned from the smulaton processes demonstrate the outperformance of the ntellgent PD-type FL control strateges over the conventonal controllers when dealng wth the LFC ssue. The present paper s arranged as follows. Secton II wll focus on desgnng a model of fve-control-are nterconnected thermal power system, whch s chosen as a typcal case study of a large-scale practcal electrc grd. Secton III then descrbes the prncple and desgn of two LFC archtectures, namely conventonal and ntellgent methodologes. Secton IV wll realze the numercal smulaton processes to evaluate the effectveness of such two categores of LFC controllers. Fnally, conclusons and future work wll be deduced n Secton V. II. Modelng a Fve-Control- Interconnected Electrc Power Grd It s the fact that the mult-control-area nterconnected power systems are hghly complcated, dependng upon the constructon plans of each country. Despte the complexty and dversty, each control-area of an nterconnected power grd always composes of three basc unts, namely governor, turbne and generator, to generate the electrc energy from the other sources (e.g., hydro and thermal energy). In ths paper, a fve-area nterconnected thermal network wll be selected as a canddate of the large-scale power systems to conduct the LFC problem. Fg. 1 shows two smple archtectures of such a fve-area electrc power grd. For the frst archtecture, only the 5 th area s nterconnected wth each other area to exchange the power. In the second case, an area s nterconnected wth each other one, so that the load varaton can appear randomly at any area, affectng both the system frequency and the te-lne power flows of the net. Therefore, the frequency and te-lne power devatons resultng from ths phenomenon need to be reduced by applyng effectve controllers n each area. In ths work, te-lne bas control based controllers are used for each control-area to solve the LFC problem (Had 21). For the typcal case study, the second archtecture ndcated n Fg. 1 (b) s chosen. Thereafter, a smulaton model, whch s bult n MATLAB/Smulnk envronment, can be shown n Fg. 2. In order to buld ths model, the transfer functons of a governor, a non-reheat turbne and a generator-load unt are formulated respectvely as follows (Kundur 1994; Tan 21): G g, 1 () s st. 1 g, (1) 4
September 214, Vol. 1, No. 1 G G t, P, 1 () s st. 1 t, K p, ( s). st. 1 P, (2) (3) #1 #1 #2 Te-lne P 52 Te-lne #5 P51 Te-lne P 54 #4 #2 P 23 P 13 Te-lne P 12 P 24 P 25 P 35 Te-lne P15 P 14 #5 P 45 Te-lne #3 P53 #3 Te-lne P 34 (a) (b) Fg. 1. Fve-area nterconnected power system models. (a) #5 s nterconnected wth each other area (b) An area s nterconnected wth each other area #4 Fg. 2. A fve-control-area non-reheat nterconnected power system model n MATLAB/Smulnk envronment 41
September 214, Vol. 1, No. 1 The smulaton parameters used n the above expressons can be found clearly n Appendx of ths paper. The state-space model for such power system model can be descrbed below X AX BU FD (4) T T where, X [ f PG PV Pte, ] s the statc varable vector, U [ PC 1 PC 2 PC 3 PC 4 PC 5] s the T control varaton, and vector D [ PD 1 PD 2 PD 3 PD 4 PD 5 ] denotes the load dsturbance varable for all areas. Here, the te-lne power flow bas can be calculated from the system frequency changes of the nterconnected control areas. A typcal expresson to compute the te-lne power devaton can be gven as follows (Kundur 1994): 5 2 P ( s) T F ( s) F ( s) (5) te, j j s j1 j where, T j and F() sare the synchronzng factor of the te-lne and the devaton of the frequency of the th control-area n the Laplace doman, respectvely. Applyng the te-lne bas control technque, the nput sgnal of the correspondng LFC controller used for the control-area # s computed relyng upon the defnton of Control Error (ACE) as shown below: ACE ( s) P ( s) B. F ( s) (6) te, where, P te, (s) and F (s) are the te-lne power change and the net frequency devaton n the Laplace doman, respectvely. Usng the defnton expressed n (6), t s the fact that only one controller s needed to mnmze the devatons of the net frequency as well as the te-lne power nterchange n accordance wth the prncple of the te-lne bas control strategy. Ths enables to smplfy the desgn and reduce the calculatng process of the control archtecture. Ths s also the most mportant reason why such a method s appled n the present study n order to mantan the power system frequency. The followng secton wll descrbe two LFC methodologes based on the bas control strategy. III. Te-lne Bas Control Strategy Based Load-Frequency Controllers A. Conventonal LFC Controllers Let us now consder an Integral controller wth the gan K I, whch can be nserted n the th control-area as one of the conventonal controllers. Based on the te-lne bas control strategy, ts nput s ACE and the correspondng output s the control sgnal U (s) as shown below: KI KI U ( s) ACE ( s) Pte, ( s) B. F ( s). s s (7) The gan constant of the above controller, K I, must be defned to satsfy both condtons of the systematcally dynamc response: the fast transent restoraton and the low overshoot. Accordng to some researches (Had 21; Kundur 1994; Shayegh et al. 29), the mplementaton of ths controller s too slow to stablze mult-control-area nterconnected power networks whch comprse non-lnear elements. Therefore, t s necessary to mprove ths controller to acheve the better control performances (Rchard 21). The second type of conventonal controllers appled to deal wth the LFC problem s PID controller. It s well known that PID controllers have been wdely and effectvely used n control systems. Also, they are more useful to be appled n the te-lne bas control strategy for the load-frequency stablzaton of the 42
September 214, Vol. 1, No. 1 power network (Shash et al. 213). Bascally, ths controller has the smlar prncple to the ntegral controller as follows: 1 U ( s) KP 1 std ACE ( s). (8) sti In (8), K p denotes the proportonal coeffcent. Meanwhle, T I and T D are the tme constant values of the ntegral and dervatve, respectvely. Such three factors strongly affect the qualty of a control system (Hamed et al. 213). Hence, t s hghly necessary to consder the tunng methods of these factors n control systems applyng PID controllers. In ths work, we employ the Zegler-Nchols method (Hamed et al. 213) to tune these coeffcents by ts domnant advantages. Applyng ths method, frst, we set the ntegral and dervatve gans to the zero values, then, the proportonal gan s tuned to reach a value at whch the control system output wll fluctuate. In the second step, the dervatve gan wll be defned wth the tuned proportonal gan above to make sure the transent performance. In the last step, the ntegral gan wll be fnally fxed wth the other factors chosen above to ensure the steady state characterstc of the control system. Because the dervatve acton s too senstve to reach the steady state, a PI (proportonal-ntegral) controller can be used as a substtuton of the PID controller n the control system. The obtaned results by applyng the PI load-frequency controllers wll also be mentoned n ths work. B. Intellgent LFC Controller Based on the aforementoned analyses, conventonal regulators can be replaced wth the fuzzy logc controllers due to ther outstandng advantages. In prncple, a FL nference conssts of three processes as descrbed below (Tmothy 21; Bmal 22): (a) the sutable membershp functons (MFs) are desgned to convert a set of crsp values nto fuzzy logc doman, (b) a fuzzy logc rule base should be determned to process and evaluate control rules, and (c) a defuzzfcaton process s mplemented to convert a set of fuzzy logc values nto the correspondng crsp set that can be used to make the control sgnal for the system. Followng three above processes, a PD-type FL archtecture appled to the control-area # s llustrated n Fg. 3. As shown, each FL archtecture uses two nputs: ace (t) and dace (t) relatng to the ACE (t) sgnal and ts dervatve, dace (t), as follows: e, e, te, ace ( t) K. ACE ( t) K. P ( t) B f ( t) (9) d dace ( t) Kde,. ACE ( t) dt (1) d Kde,. Pte, ( t) B f ( t) dt where, K e, and K de, denote the scalng factors correspondng to ACE and ts dervatve dace. The output of the proposed controller s u (t), relatng to the control sgnal of the th control-area, by the proportonal factor Ku. It s found that such an FL archtecture can be consdered as an nput/output statc nonlnear mappng, and thus the prncple of such FL controller can be wrtten as follows: d u ( t) K1. ace ( t) K2. ace ( t) (11) dt where, K 1 and K 2 are nternal-nonlnear coeffcents of the FL nference. From (9), (1) and (11), t s clear to nfer the followng equaton: 43
A (x) Internatonal Journal of Academc Research n Computer Scences and Electrcal Engneerng September 214, Vol. 1, No. 1 hence, ΔP te, U ( t) K. u ( t) u, PD-TYPEFUZZYLOGIC CONTROLLER d Ku, K1. Ke,. ACE ( t) K2. Kde,. ACE ( t) dt Rule base (12) ACE ace Calculate and evaluate ACE, dace K e, K de, dace fcaton Fuzzfcaton Evaluaton of control rules Defuzz- u K u, U # Model Δf dace Database Fg. 3. PD-type FL controller archtecture for the th area 1.8.6.4.2 c 1 = -.7; 1 =.2 c 2 = -.35; 2 =.15 c 3 = ; 3 =.1 FL1 FL1 where K K. K. K and K K. K. K P, 1 u, e, -1 -.8 -.6 -.4 -.2.2.4.6.8 1 x Fg. 4. Gaussan membershp functon shapes FL1 FL d 1 U ( t) KP,. ACE ( t) KD,. ACE ( t) (13) dt D, 2 u, de, denote respectvely two factors whch are hghly smlar to the proportonal and dervatve coeffcents of a PD regulator. Therefore, t can be sad that the type of such FL controller s dependent on the PD prncple (PD-based FL controller). Bascally, there have been plenty of shapes of MFs can be employed. Also, many methods of defuzzfcaton process are able to be appled n control practce employng FL controllers. In the context of ths paper, Gaussan MFs are used for all of two nputs and one output of the proposed PD-type FL controller. In prncple, each Gaussan MF, μ A (x), s mathematcally formulated as follows: A ( x) exp x c 2 where c denotes the MF center, whle σ s the wdth of the MF #. Fg. 4 shows several cases of the Gaussan MFs wth dfferent values of such two parameters. In ths study, seven logc levels are used for each Gaussan MF of two nputs and one output of the proposed PD-type FL controller. Table 1 shows the meanngs and values of these logc levels appled to ths work. In addton, Table 2 descrbes an approprate rule base appled for the proposed PD-type FL controllers usng the Mamdan archtecture. There are a total of 49 rules used for such controller. Each of these rules s able to be wrtten as: IF the frst nput ace (t) s e and the second nput dace (t) s de THEN the output u (t) s u. For 2 2 (14) 44
September 214, Vol. 1, No. 1 example, the last rule (correspondng to the last row and the last column of Table 2) s: IF ace (t) s BP and the second nput dace (t) s BP THEN the output u (t) s BP. Accordng to the composton rule theory of an FL reasonng, each gven rule base can be used to perform a meanngful control acton n accordance wth a specfc condton of the varables. Such a composton rule, employed for the FL nference to generate the output control sgnal, should be chosen properly enough to acheve the desred control qualty. For ths study, the MAX-MIN (maxmum-mnmum) composton s selected snce t s the most common and effcent composton for the FL nference. Accordng to such composton rule, the output MF s calculated by usng a MIN mechansm. In contrast, a MAX mechansm wll be used to calculate the output of the fuzzy model. In the followng secton, the effectveness of the proposed LFC methodology usng the PD-type FLCs wll be demonstrated through smulaton processes usng MATLAB/Smulnk package. Table 1. Lngustc terms for two nputs as well as one output of the proposed PD-type FL controller Lngustc varable Meanng c σ BN Bg Negatve -1 MN Medum Negatve -2/3 SN Small Negatve -1/3 ZO Zero.1414 SP Small Postve 1/3 MP Medum Postve 2/3 BP Bg Postve 1 Table 2. The rule base of the proposed FL nference dace(t) ace(t) BN MN SN ZO SP MP BP BN BN BN BN MN SN SN ZO MN BN MN MN MN SN ZO SP SN BN MN SN SN ZO SP MP ZO BN MN SN ZO SP MP BP SP MN SN ZO SP SP MP BP MP SN ZO SP MP MP MP BP BP ZO SP SP MP BP BP BP IV. Numercal Smulaton In ths secton, to mplement the numercal smulaton processes based on the te-lne bas control strategy correspondng to the Fg. 2 llustrated earler, four smulaton cases of load-frequency controllers are consdered, ncludng I, PI, PID, and PD-type FL regulator. In order to evaluate and compare the effectveness of such four controllers when dealng wth the LFC problem, a typcal condton of load T changes, whch s gven as a vector D [2(%) 1(%) 1.2(%) 1.5(%) 1(%)], wll be fed to all smulaton cases. By usng MATLAB software verson 213a, smulaton results have been obtaned as plotted n Fgs. 5-1 as well as ndcated n Tables 3-5. Frst, the frequency devatons of the 1 st and 5 th areas are descrbed n Fg. 5, correspondng to the applcaton of three conventonal regulators (.e., I, PI, and PID). 45
f f f 1,5 f 1,5 f 1,5 Internatonal Journal of Academc Research n Computer Scences and Electrcal Engneerng September 214, Vol. 1, No. 1 Meanwhle, Fg. 6 shows the transent oscllatons of the frequency devatons for all control areas usng the proposed ntellgent PD-type FLCs. Next, Fg. 7 plots the te-lne power devatons of the 2 nd and 3 rd areas resultng from four categores of the LFC controllers. In order to compare the dynamc responses of dfferent controllers, Fg. 8 (a) llustrates the frequency devatons for only the frst area. Also, after calculatng the frequency devaton errors of the above controllers, the correspondng error curves can be obtaned as represented n Fg. 8 (b). As shown, the frequency change error between I and PID controllers s the smallest, whereas the bas of I and PD-type FL controllers s the largest. Furthermore, to demonstrate numercally the obtaned results, Tables 3-5 represent the comparson for all cases. An acceptable frequency tolerance of.1% s gven to calculate the settlng tmes of the transent oscllatons. From these tables, both the overshoot (maxmum peak) and settlng tme of the proposed FL controllers are the best control performances. Fnally, Fgs. 9 and 1 present the comparson of the frequency devaton, n percentage, correspondng to Tables 3 and 4. It s found that both comparson ndces of the proposed ntellgent PD-type FLCs wth the conventonal controllers are much smaller than 1%. As a result, t s able to valdate clearly the outperformance and robustness of the PD-type FL control archtecture when solvng the LFC problem..2.1.1 -.2 -.4 -.6 A#1-I A#5-I -.8 5 (a) -.1 -.2 -.3 -.4 -.5 A#1-PI A#5-PI -.6 5 (b) -.1 -.2 -.3 -.4 A#1-PID A#5-PID -.5 5 (c) Fg. 5. Transent oscllatons of the frequency devatons n the area #1 and area #5 usng three conventonal LFC regulators.2 #1 - FLC -.2 #2 - FLC #3 - FLC -.4 1 2 3 4 5 (a).1 -.1 #4 - FLC #5 - FLC -.2 1 2 3 4 5 (b) 46
Errors of f 1 f 1 P te P te P te P te Internatonal Journal of Academc Research n Computer Scences and Electrcal Engneerng September 214, Vol. 1, No. 1 Fg. 6. Frequency devatons n all fve control areas usng ntellgent FL controllers.1.1 A#2-I A#3-I -.1 5 (a).1 A#2-PI A#3-PI -.1 5 (b) 5 x 1-3 A#2-PID A#3-PID -.1 5 (c) -5 A#2-FLC A#3-FLC -1 5 (d) Fg. 7. Transent oscllatons of the te-lne power bases n the second and thrd area usng dfferent controllers.5 FLC #1 - I -.5 PID #1 - PI #1 - PID Integral #1 - FLC -.1 1 2 3 4 5 (a).2 -.2 I vs PID -.4 I vs FLC PID vs FLC I vs FLC -.6 1 2 3 4 5 (b) Fg. 8. A comparson of the frequency devatons n the frst area usng dfferent LFC controllers 47
Maxmum peaks (%) Internatonal Journal of Academc Research n Computer Scences and Electrcal Engneerng September 214, Vol. 1, No. 1 Type of controller Table 3. Maxmum peaks, n p.u., for all control-areas usng dfferent LFC controllers #1 #2 #3 #4 #5 I -.66 -.612 -.632 -.64 -.632 PI -.526 -.529 -.557 -.562 -.557 PID -.447 -.472 -.469 -.472 -.465 FLC -.183 -.25 -.13 -.149 -.199 Table 4. Settlng tmes, n second, for all control-areas usng dfferent LFC controllers wth ɛ f =.1% Type of controller #1 #2 #3 #4 #5 I 33.4343 27.862 34.926 41.4941 34.926 PI 25.258 22.7616 24.5867 3.3569 24.754 PID 11.83 17.8729 18.7391 13.2268 1.9479 FLC 1.1979 14.2259 15.9634 1.7732 8.645 Table 5. Absolute values of maxmum peaks of the te-lne power flow devatons, n p.u., for all control-areas usng dfferent LFC controllers Type of controller #1 #2 #3 #4 #5 I.157.97.76.8.76 PI.142.81.68.72.68 PID.121.76.58.65.65 FLC.83.49.26.21.61 6 4 2 FLC vs PID #5 #4 #3 #2 #1 FLC vs I FLC vs PI Fg. 9. A comparson (%) of the maxmum peaks of the frequency devatons for all areas usng dfferent LFC controllers 48
Settlng tmes (%) Internatonal Journal of Academc Research n Computer Scences and Electrcal Engneerng September 214, Vol. 1, No. 1 1 5 #5 #4 #3 #2 #1 FLC vs I FLC vs PI FLC vs PID Fg. 1. A comparson (%) of the settlng tmes of the frequency devatons for all areas usng dfferent LFC controllers wth an acceptable tolerance of.1% V. Conclusons In ths paper, the nvestgaton of dfferent load-frequency controllers, focusng on the ntellgent FL control methodology, has been conducted to deal wth the LFC problem of a mult-control-area nterconnected power network. Frst, a typcal type of fve-control-area nterconnected non-reheat thermal power systems has been mathematcally modeled. Subsequently, the prncple and desgn of two categores of load-frequency controllers ncludng the conventonal and ntellgent regulators are dscussed. Thereafter, varous smulatons have also been performed to verfy the qualty of such two types of controllers. Gven the desred tolerance of the frequency devatons, ntellgent controllers usng PD-based FL archtecture have acheved the better control characterstcs, e.g., smaller overshoots and shorter settlng tmes, n comparson wth the conventonal controllers. These promsng results wll further promote the applcaton of the ntellgent control methodologes based on the FL technque n order to address effcently the LFC problem n a practcal nterconnected power system. For future work, the other modern technque such as Artfcal Neural Network (ANN) should be nvestgated to ntegrate wth the proposed FL control archtecture n order to further mprove ts control qualfcaton. Such an ANN archtecture can be used to optmze the parameters of an FL nference, such as the rule base and scalng factors. Moreover, the practcal-complcated power systems, e.g., Schuan s power network, should be consdered n order to further enhance the real applcaton of the proposed control methodology. Ths means that such a practcal electrc power grd needs to be modeled more exactly, consderng adequate nonlneartes and uncertantes. These characterstcs wll make the proposed FLbased control archtecture more completely and practcally. 49
September 214, Vol. 1, No. 1 Appendx Parameters for fve-area nterconnected thermal power system model T g,1 =.8, T g,2 =.12, T g,3 = T g,4 = T g,5 =.1 T t,1 =.28, T t,2 =.32, T t,3 = T t,4 = T t,5 =.3 K p,1 = 12, K p,2 = 1, K p,3 = K p,4 = K p,5 = 11 T p,1 = 22, T p,2 = 2, T p,3 = T p,4 = T p,5 = 18 T j =.71 Acknowledgment The authors wsh to thank Mr. Nguyen Thanh-Mnh, emal-ngockhoa21@yahoo.com.vn, for hs meanngful dscusson of numercal smulaton processes usng MATLAB/Smulnk package. References [1] Had, S. (21). Power System Analyss. New York: PSA Publshng LLC. [2] Bevran, H., & Hyama, T. (211). Intellgent Automatc Generaton Control. New York: CRC Press. [3] Wood, A. J., Wollenberg, B. F., & Sheble, G. B. (213). Power Generaton, Operaton, and Control, New York: Wlley-Interscence. [4] Kundur, P. (1994). Power System Stablty and Control, New York: McGraw-Hll. [5] Rchard, G. F. (21). Power system dynamcs and stablty. Boca Raton: CRC Press LLC. [6] Chongxn, H., Kafeng, Z., Xanzhong, D., & Qang, Z. (213). Robust Load Frequency Controller Desgn Based on a New Strct Model. Electrc Power Components and Systems, 41(11), 175-199. [7] Ibrabeem, P. K., & Kothar, D. P. (25). Recent Phlosophes of Automatc Generaton Control Strateges n Power Systems. IEEE Trans. Power Systems, 2(1), 346-357. [8] Shayegh, H., Shayanfar, H. A., & Jall, A. (29). Load Frequency Control Strateges: A State-of-the-Art Survey for the Researcher. Energy Converson and Management, 5(2), 344-353. [9] Shash, K. P., Soumya, R. M., & Nand, K. (213). A Lterature Survey on Load-Frequency Control for Conventonal and Dstrbuton Generaton Power Systems. Renewable and Sustanable Energy Revews, 25, 318-334. [1] Hamed, S., Behrooz, V., & Majd, E. (213). A robust PID controller based on mperalst compettve algorthm for load-frequency control of power systems. ISA Transactons, 52, 88-95. [11] Tan, W. (21). Unfed Tunng of PID Load Frequency Controller for Power System va IMC. IEEE Trans. Power Systems, 25(1), 341-35. [12] Hassan, M.A.M., & Malk, O.P. (1993). Implementaton and Laboratory Test Results for a Fuzzy Logc Selftuned Power System Stablzer. IEEE Trans. Energy Converson, 8(2), 221-228. [13] Chown, G. A., & Hartman, R. C. (1998). Desgn and Experences wth a Fuzzy Logc Controller for Automatc Generaton Control (ACC). IEEE Trans. Power Systems, 13(3), 965-97. [14] Santos, D. M. J., Perera, J. L. R., Flho, J.A.P., De Olvera, E.J., & Slva, C. D. (24). A New Approach for Interchange Control Modelng. IEEE Trans. Power Systems, 19(3), 1217-1276. [15] Demroren, A., & Yesl, E. (24). Automatc generaton control wth fuzzy logc controllers n the power system ncludng SMES unts [J]. Electrcal Power and Energy System, 26, 291 35. [16] Subbaraj P., & Manckavasagam, K. (27). Generaton Control of Interconnected Power Systems Usng Computatonal Intellgence Technques. IET Generaton, Transmsson and Dstrbuton, 1(4), 557-563. [17] Tmothy, J. R. (21). Fuzzy Logc wth Engneerng Applcaton. New York: Wlley. [18] Bmal, K. B. (22). Modern power electroncs and AC drves. New York: Prentce Hall PTR. 5
September 214, Vol. 1, No. 1 Nguyen Ngoc-Khoat (S 7-M 9) receved the B.S. and M.S. degrees n Automaton and Control at Hano Unversty of Scence and Technology, Hano, Vetnam, n 27 and 29, respectvely. Currently, he s workng as a Lecturer at Faculty of Automaton Technology, Electrc Power Unversty n Hano, Vetnam. He s now pursung the Ph.D. degree n Electrcal Engneerng at the Unversty of Electronc Scence and Technology of Chna, Chengdu, Schuan, Chna. Hs research nterests nclude large-scale nterconnected power systems montorng, control and protecton. Mr. Ngoc-Khoat has publshed several academc papers followng hs research feld. Dao Th-Ma-Phuong (S 7-M 1) receved the B.S. and M.S. degrees n Automaton and Control at Hano Unversty of Scence and Technology, Hano, Vetnam, n 27 and 21, respectvely. Currently, she s workng as a Lecturer at Faculty of Electrcal Engneerng, Hano Unversty of Industry n Hano, Vetnam. She s now pursung the Ph.D. degree n the feld of Automaton Engneerng at Hunan Unversty, Changsha, Hunan, Chna. Her research nterests nclude large-scale power systems, control theory and applcaton. Mrs. Ma-Phuong has publshed several academc papers followng her research feld. 51