455 IEEE Transactions on Energy Conversion, Vol. 11, No. 2, June 1996 DESIGN AND ANALYSIS POWER OF AN ADAPTIVE SYSTEM FUZZY STABILIZER K. Tomsovic P. Hoang School of Electrical Engineering and Computer Washington State University 99164-2752 Pullman. WA system stabilizer Abstract Science (PSS) provides a positive damping torque [1,2] in phase with the speed signal to cancel the effect of Power system stabilizers (PSS) must be capable of providing a broad appropriate stabilization signals over operating conditions and disturbances. Traditional range of PSS rely on robust linear design methods. In an attempt to cover a wider range of operating conditions, expert or rule-based controllers have also been proposed. Recently, fuzzy logic as a novel robust control design method has shown promising results. The around emphasis uncertainties in fizzy control design in system parameters centers and operating conditions. Such an emphasis is of particular relevance as the difficulty of accurately modelling the connected generation is expected to increase under power controllers are based on empirical rules. In this paper, a systematic approach control design is proposed. Implementation reqtures specification performance parameters real-time criteria which of performance translates into can be calculated control to fuzzy logic for a specific criteria. three off-line This controller or computed in in response to system changes. The robustness of the controller is emphasized. Small signal analysis methods are discussed. This work developing robust appropriate when fuzzy logic is applied. stabilizer design and transient is directed at and analysis methods Power system stabilizers operating stabilization conditions 039-8 EC Control implemented has shown promising (PSS) must be capable of providing signals over and disturbances. results [6,7]. algorithms based on fuzzy logic have been in many processes [8,9]. The application of such control techniques has been motivated that obtained (2) simplified by the desire for (1) improved robustness over using conventional linear control algorithms. control design for difficult to model systems. e.g., the truck backer-upper problem [10], and (3) simplified implementation. In power systems, several controllers have been developed for PSS. One such controller hdyro unit has been undergoing field [8] for a small test in Japan. Other systems have been developed for voltage regulators the control of FACTS devices have been tuned to a specific numerical solutions generally [12]. Most [11] and of these designs system. Unfortunately, require such a large computational general design and analysis The proposed method also ptirsues small signal stability analysis which provides the opportunity to design a system with adjustable controller a broad range A traditional of power paperrecommendedandapprovedby the IEEE Energy Development and Power Generation Committee of the IEEE Power 96 W rule-based controllers effort. In this paper, a methodology is proposed. 1. Introduction appropriate condition, they may not be valid for a wide range of operating conditions. Considerable efforts have been directed towards developing adaptive PSS. e.g. [3,4]. In an attempt to cover a wide range of operating conditions, expert or rule-based controllers have been proposed for PSS [5]. Recently, the introduction of fuzzy logic into these one or more of the following: deregulation. Fuzzy logic machine industry the system negative damping torque. Because the gains of this controller are determined for a particular operating A parameters to obtain suitable root location Although [13]. fuzzy logic methods have both a well-founded theoretical basis and numerous successful implementations, controversy has surrounded the developed systems. This is due in part to the lack of satisfying performance measures. Recently, there have been efforts directed at appropriate Engineering Society for presentation at the 1996 IEEE/PES Winter Meeting, January 21-25, 1996, Baltimore, MD. Manuscript submitted July 27, 1995; stability measures for fuzzy logic controllers [10, 13]. In the power system, performance concerns are particularly acute made available with the high reliability for printing January 4, 1996. of instabilities. models may requirements and the costly effects Yet, analysis using precise mathematical be infeasible due to the power system complexity (i.e., large dimension, non-linearities, uncertainties in load fluctuations, disturbances and generator dynamics, and so on). The viewpoint offered here is that 0885-8969/96/$05.00 @ 1996 IEEE 456 fuzzy Ioglc has been introduced because of the above ditllculties and thus, the approach should be better equipped than conventional methods concerns. In this work, systematic analysis. to address these performance a first For simulation step is taken towards studies. the non-linear power system and controller are linearized and small signal stability analysis is performed, It is proposed to rethink the traditional methods greater uncertainties. of stability this section. in of the MODELS generator and system models presented (details can be found in the Appendix). diagram terms Fig. 2. Excitation 2. SYSTEM In assessment of the generator plant model are The block is shown in Fig. 1. Specifically, the plant is modeled based on a generator model incorporating single-axis field flux variation and the simple excitation system shown in Fig. 2. The PSS used for comparison studies is a lead compensator The following H=.(s) models the excitation continuous transfer function system and regulator: occur in an overload or islanding condition. The second term of (2) is a lead compensator to account for the phase lag through the electrical system [14]. In many practical cases. the phase lead required from a single The development of the fuzzy logic approach here is limited to the controller structure and design. More detailed with a fuzzy signal is introduced in conjunction In this study, both a traditional logic based stabilizer PSS is modeled by the following with the exciter respectively. stabilizing () using the filtered of the machine. That is. for the controller. The control output, u. is the signal Vs. Each control rule R, is of the form: The traditional transfer function: IFe is A, AND e is B, THEN . rule-base PSS and a fuzzy (FPSS) are analyzed. I+sTl — 1+sT2 logic and acceleration the deviation from synchronous speed and acceleration of the machine are the error, e, and error change. &, signals, reference voltage to obtain feedback for the regulator- .~T case. cascaded lead discussions on tizzy logic controllers are widely available, e.g., [9, 12]. For the proposed FPSS. the second term of (2) 2.1 Power system stabilization ~(s) In this 2.2 Fuzzy stabilizer speed deviation system. is greater than that obtainable lead network. stages are used where k is the number of lead stages. is replaced The stabilizing System u is C, k (2) where .-t,. B I membership and C, functions are fuzzy sets with as shown normalized triangular between -1 and The first term in (2) is a reset term that 1s used to “wash 1 in Fig. 3. These same fuzzy sets are used for each variable out” the compensation effect after a time lag T. The use of reset control will assure no permanent offset in the terminal of interest: only the constant of proportionality is changed. These constants are h-,, K, and k- for the error. error voltage change and control output, respectively. The error and error change are classified according to these fuzzy membership due to a prolonged error in frequency, which may functions modtiled by an appropriate signal may have non-zero Similarly. a specific contribution membership control of more than constant. A specific in more than one set. signal one rule. may represent the Rule conditions are joined by using the minimum intersection operator the resulting membership fimction for a rule is: UR, \ —-[ Ffiter Fig. 1 Generator Plant Model (e, e) = min(h(e)> !-h,(e)) so that (3) The suggested control output from rule I is the center of the membership function C,. Rules are then combined using the center of gravity control output U method to determme a normalized 457 @ MN LN SN SP ZE MP LP I -1 -.65 -.3 0 Fig. 3. Membership (LP=lqge posltwe: positwe; negatwe; MP= ZE=zero: SN= 1 LP LP LP AfP IMP SP I ZE MN LP MP A@ IUP SP ZE Slv SN LP MP SP SP ZE NV AL%i ~ ZE MP IUP SP ZE SN ‘AIN lMAT SF’ MP SP IZE SN SN MN LN I scaled from -1 to 1 medium SN=snlall small .65 .3 functions LN SP=small posltwe: negatwe; MN=medlum e negative) I (4) I 2.3 Proposed LP FPSS design steps The fhzzy logic controller development so far is general. particular control design requires specification of all control The control rules are rules and membership functions. designed from an understanding controller, K, to be symmetric control and many manual functions needed. rules is shown in Table 1. Each under the assumption tuning to establish of control that have proposed methods for tuning the controller). rules artificial and e.g. [15], neural net Such manual tuning may be very time consuming and perhaps more importantly sheds some doubt on the claims for robustness of the fuzzy logic approach. In this work. a systematic tuning methodology is proposed. It is assumed that the fimdamental control laws change quantitatively not qualitatively with the operating In this vein. control rules and membership condition. functions functions are designed once as above. are modified by scaling through error is described below: control output for K based on The membership the constants K,. a significant disturbance e and e. respectively, until system oscillations exceeds the observed values for during the simulation penocl. If damping appears inadequate then: 5. Linearize if the desired performance system. (There are some exceptions, authors are italicued) either begin to settle or the stability limit. 4. Set A-. and h-,to the maximum point will be is no longer necessary any asymmetries could be best handled through scaling. In addition, adjacent regions in the rule table allow only nearest neighbor changes in the control ouput (LN to MN. MN to SN and so on). This ensures that small changes in e and e result in small changes in u. Many of the fuzzy logic controllers proposed in the membership 1 I system and range of operating 1. Select the maximum of the 49 control rules represents a desired controller response to a particular situation. The control rules were for a specific outputs ~Lhr ./X of the desired effect of the that the desired operating set of control on MN The methodolo~ 3. Simulate This rule anticipates rely MN I I the physical limitations of the controller. 2. Replace the FPSS with a constant gain K. reached soon and stabilization literature I SIV and A“ for a particular conditions. IF e is SN AND e is SP THEN /.{ k ZE designed I (Control For example, consider the rule: The complete I IL.V , Table. 1. Rules table ZE A I I the system and FPSS around the nominal operating 6. Using K. point (see section 4). traditional eigen value and K. together (i.e. maintaining magnitude) As an objective adjust to obtain desired damping. of the fuzzy controller range of operating analysis, the same relative conditions is to manage a wide and modelling uncertainties, the simulation in step 3 for method 1 may need to be repeated under a set of parameter variations. The controller is adaptive in the sense of the varying of these gains but not in terms of varying the control rules. Further discussion on these design steps can be found in [16]. 3. NUMERICAL In this section, simulations to several disturbances. exercise the controller SIMULATION illustrate The the controller scenarios are response intended to rather than to represent any specific system scenario. Two simple systems are presented but higher order systems have been simulated with similar results. The FPSS constants are found by simulations of a 458 angle response to this disturbance for systems with the PSS and the FPSS design (see Fig. 6) w Case 2: Three phase to ground fault at A.. (Fault is 30% of the distance along line). Line is removed with a fault clearing time tc=0.2 The plots show the sec response of the systems with the PSS and the FPSS design (see Fig. 7). -1- 3.2 Multimacbine Fig. 4 Single machine connected inthite bus A system with two machines 2 ,1 connected to an infknite bus is shown in Fig. 5. The system has the following parameters (all values in per unit): Network: line 1-2 R=.O 18, X=.11, B=.226; R=.008, X=.05, B=.098: both lines 2-3 line 1-3 R=.007, X= 04. B=.082; 1~1=0.6-j0.3, Y,2=0.4-J0.2 Mechanical Dower: Pml=l .2, P~2=l Generator parameters are the same as in the single machine case. One scenario is presented here: + Fig, 5. Two machine step change in mechanical model of Appendix A. - three bus system + power input using the non-linear The PSS is designed using a Case 3: Three phase to ground intlnite bus. adapted from [17], was used in the design phase studies of a connected Fig. 4 has the following Generator: M=9.26s, .1”;=. 190 p.u. external In all cases, the FPSS shows superior to an infhite D=.O1 The T., = 05s. negatwe R arises =9.25. K= from modelhng V=l .05: T,=O.6851s, on the following to P.= robustness effective despite significant in response to disturbances. the For the more severe has significantly simulations that the better demonstrates controller remains changes in the system dynamics. ANALYSIS 4.1 Small signal stability For small disturbances, linearized of this stability about the can be characterized operating point. by the If system lie in the left hand plane, the the system is small disturbance stable. In this study. the delay caused by computation is neglected so that the tizzy logic and are shown in the figures page. 1.3 p.u. the FPSS controller elgenvalues 1.0 @ Case 1: Step change of mechanical is not as noticeable. The multi-machine controller or similar smaller disturbances, system input power’; P~=l Two cases were simulated For 4. PERFORMANCE Y= O 6 +JO.3. T=3 .0s. controller. damping lE~~l S 10 both lines R=-O.68, .Y=l.994; KC= 7.09, Power system stabilizer: Mechanical traditional performance. the p.u, T>O=7.76 s. A~=.973 P.U. of generation.) K, bus shown in system parameters: : KA=50. equwalent T,=O.1s ~: & =40,0, 3.3 Discussion improved The single machine Voltage remdator (iVote: Network: FPSS design. the 3.1 Single machine is The line returns to service with clearing time tC=O. 15 sec. (Fig. 8). Plots show response for systems with the PSS and the conventional phase lead technique to precisely compensate for the phase lag of the electrical loop. Two systems are used for the simulations. A single machine connected to an and then this controller was used for multi-machine system. adapted from [18]. at B. (Fault fault 30’% of the distance along line 1-3). power from P.= 1 The plots show frequency and rotor controller can be modeled as a zero memory non-linearity. This is a reasonable assumption as the rotor oscillations of interest are orders of magnitude slower than the time required for the FPSS computations. The FPSS does not introduce new poles but acts to shift the eigen values of the uncompensated system. 459 A difficulty of the small signal analysis lies in the fact that the FPSS is not differentiable. This problem is managed fuzzy logic by numerically critical operating calculating point. Table a linear 2 shows approximation the near the eigenvalues for the system with the traditional PSS and the FPSS design. As the FPSS should provide proper stabilization control over a wide range of operating conditions, the eigenvalues are found at The system is designed for the two operating points. nominal operating at second the controller point and the eigenvalues operating parameters. point While controllers preliminary but our results in this area are stall A numerical clearing operating time approach (CCT) is pursued is calculated here. The for a number points for the PSS and FPSS systems. of The results are shown in Table 3. In all cases, the FPSS designs improve the margin of stability as indicated by the CCT. 5. CONCLUSIONS AND DISCUSSION are recalculated without changing the uncompensated the system has This paper proposes a general stabilizer. maximum structure for a fuzzy logic Controller design requires calculation of the ranges for frequency and frequency deviation two eigenvalues in the right hand plane, both the traditional PSS and FPSS act to move the eigenvalues into left hand plane and establish small disturbance stability. It is interesting to note that the I?PSS shows good small signal during some specified disturbance. The advantage of th~s design approach is that the controller is insensitive to the precise dynamics of the system. Simulation of the response performance with relative insensitivity point Similar results were found for system. to disturbances has demonstrated the effectiveness of this design technique. Small signal and transient stability analysis give some evidence of the robustness of the to the operating the multimachine controller. 4.2 Transient stability This research methods For large disturbances, considered. the system non-linearities It is possible to apply Lyapunov Operating Point Pss Nominal -1.361 + j4.452 -4.290 ~ j8.199 ~lm=] must be functions -2.072 ~ j8.703 -19.59 -0.33 - 8.027~ j13.312 under [1] F. -1.790 P. -8.309 .4pparatus ~ j14.412 316-329. for the single machine systems FPSS machine P,”=l.30 0.08 sec. 0.84 sec. 0.14 sec. System 0.25 sec. Pm,=l.2, Pm,=l.2, Pm,=l.o Pm,=l.2 Table 3: Critical 0.23 sec. 0.21 sec. clearing O. 31 sec. 0.28 sec. systematic logic based to design controllers which of dynarnlc model sec. times for example systems C. Concordia, Stability ,“ IEEE and S’stems, Vol. “Concept as Transactions PAS-88. of Affected by on Po~er April 1969. pp. Approach for Adaptation Proceedings Svstern of .4ppltcation of Second Power System Sympostum to Power S’stems, of Seattle. July 1989. pp. 465-568. Power System Stabilizers Transactions “Design of Self-tuning for Synchronous on Energv Conversion. PID Generators.” Vol. EC-2, No.3. 1987, pp. 343-348 [5] D.Xia and G, T. Heydt, “Self-tuning Controller for the Transactions on Generator Excitation Control, ” IE~E Vol. PAS-102, 1983. Power .4pparatus and Systems, pp. 1877-1885. [6] T. 0.24 and [4] ‘Y. Y. Hsu and K. L. Lious, IEEE Two machme Nominal tizzy [2] S. E. M. de Oliveira, “Effect of Excitation Systems and Power Systems Stabilizers on Synchronous Generator Damping and Synchronizing Torques. ” Ii7E Vol. 136, September 1989, pp. 70 Proceedings, [3] G. Honderad, M.R. Chetty and J. Heydman. “An Expert Expert 0.53 sec. developing for uncertainty Machine Control Stabilizer,” Nominal extreme deMello ~ j2.634 J Point The ability Synchronous -0.33 Table 2: Eigenvalues Single are effective Excitation t j3.727 -4.814 Operating control. at analysis 6. REFERENCES -0.33 Pm=l.30 stabilization directed and ~ j3.235 -0.33 -0.818 is design parameters is felt to be of growing relevance as the number of energy suppliers connected to the network increases. FPSS -18.88 to of Hiyama. “Real Time Control of Micro-machine System using Micro-computer based Fuzzy Logic Power 11’mter .Ileelmg. 93 System Stabilizer,” 1993 IEEIOPES W 128-EC, Columbus, OH. January 1993. 460 ““’~ ,. ,, ,, . . . . o 6 4 2 0 8 4 2 6 $ Tune (SN ) Time (sec.) Fig 6 SteD chm?e of me-ctical lxnver from P-=l to P.=1 3,--- F’SS, . FPSS 100 ‘“0081~ I I /-, J I 1 ,, ,, II 0.998 0 4 2 6 I 1 20 0 8 2 4 Tu-ne (see ) 6 8 Tune (see ) Fig. 7 Faull at A with t, =0 2 sec , -- PSS, - FPSS Machine 1 1.02: ,,, I -1,01 3 & I ,,, ,,. ,,, ,,’! > ‘,>,, c w $ ,,, Madune 2 101[ ;’ ,, ,. ‘, ,(’ ‘,’’., ,, ? . - 1 , 0 1 2 3 4 5 6 Time (SW ) Excitation \olt.age 1 20- ‘, ‘, , 1 099— o 1 I 2 3 4 5 6 5 6 Tune (sez ) Excitation voltage 2 20 -1 -lo~ 0 1“ I I 1 I 2 3 Tme (SW) 4 Fig 5 6 $ Fault at B mTth :,=0 0 1 15sec, -- Pss, - FPss 2 3 Tune (S= ) 4 461 [7] M. A. Hassan and 0. P. Malik, Laboratory Test Self-Tuned Results for “Implementation a Fuzzy (B-7) and Logic Based (B-%) Power IEEE/PES System Stabilizer.” Summer .Weeting, 92 SM 474-9 EC, Seattle, July 1992 [8] S.T Wierzchon, “Mathematical Tools for Knowledge Representation,” Reasorung in Set Thecvy Zimmerman, Fuzzy Application, Mlwer-Nijhoff, First Ed., 1985. [10] J K. Tanaka Problem to (B-IO) Expert (B-11) 1985, pp. 61-69. Svstem. [9] H. Approximate (B-9) and M. Sane, of Fuzzy Control Backing up TransactIons Control on “A Robust a17d Stabilization Generator rotor angle Systems and Its Application of a Truck-Trailer,” Systems, Vol. 2, No. Fuzzy (B-12) Its Synchronous IEEE 2. May 1994. [11] A. Hasan and A.H.M. Sadrul Ula, “Design and Implementation of a Fuzzy Controller Based Automatic Voltage Regulator for a Synchronous Generator,” 1994 IEEE/ PES W’inter Meeting, 94 WM 025-7, New York, Feb. 1994. [12] M. Noroozian, “Robust, G. Andersson Near Time-Optimal Oscillations and Control with Fuzzy Logic,” K. Tomsovic. of Power System 1995 IEEE@ES winter Meeting, 95 WM 237-8 PWRD. New York, Feb. 1995. [13] K. Tomsovic and P. Hoang. “Approaches for Evaluating Fuzzy Logic Based Power System Stabilization 1994 Control,” P. Kundur, McGraw-Hill, [15] J. Lee, “On PI-Type on Sets and Soft Power Fuzzy Logic Fuzzy System 1994. Methods and for Improving on Fuzzy Controllers,” J.J. Grainger flna[vsis, and McGraw-Hill, APPENDIX 1, No. IEEE 4, &stenz E&, = (–Ej, Generator real power pL Load real power QG Generator reactive power QL Load reactive power Id Direct axis Moment of Direct axis Quadrature – (.Y~, –.~j, reactance axis transient Number of generators Y Y bus matrix Y V;ef Reference voltage pp. reactance Field voltage 1993. Angle of Y bus matrix K. Stabilizing input signal Regulator amplifier gain T. Regulator F, amplifier time constant Academic Stevenson, Power Systems 1994. DETAIL differential (B-1) pm, – pG, )/Af, (B-2) ) Id, + Efd, )/~dOz (B-3) PG, = ~’>,Id, i- l’~,Iql (B-4) QC, = t ‘qId, – ~’d,I~, (B-5) Id, = (E&– (B-6) ~“~,)/.\:, current inertia reactance axis reactance BIOGRAPHIES W.D. A: MODELING (~, – (l&j]+ power input n Dynanucs, co, – WYef & = (–D, o voltage factor Transactions Nov. The study systems are described by the following and algebraic equations: & axis voltage Mechanical of ~vstems. Yu, Electrlc Power Press, 1983, pp.8(J-81 Y.N. [18] Quadrature Direct axis transient K. Tomsovic and P. Hoang. “Design and Analysis Methodology for Fuzzy Logic Stabilization Control of Power System Disturbances.” to be submitted to LEEE Transactions [17] Direct axis voltage PG Control, Performance axis voltage Pm .Yd 298-301. [16] direct axis voltage Quadrature Vol. $vstems. Stablhty Internal D ltf Computing quadrature Terminal Damping F San Jose. CA, Nov. 1994, pp. 262-269. Svmposlum. [14] Rough ~; frequency Internal Patrick Hoang Received the B.S. and M. S. degrees from Washington State University. Pullman. in 1992 and 1994. respectively, all in Electrical Engineering. He is currently employed at DISC Inc. His research interests include expert systems and fuzzy set application to power system and plant controls. the B. S.E.E. from Kevin Tomsovic (M’87) received Michigan Tech. University. Houghton, in 1982, and the M. S.E.E. and Ph.D. degrees from University of Washington, Seattle, in 1984 and 1987. respectively. He has held visiting professorships at National Cheng Kung University. National Sun Yat-Sen University and the Royal Institute of Technology in Stockholm. His research interests include expert systems and fuzzy set applications to power systems