Demonstration of Measurement Derived Model-Based Adaptive Wide-Area Damping Controller on Hardware Testbed USA. China USA

Transcription

1 2, rue d Artois, F-758 PARIS CIGRE US National Committee http : // 25 Grid of the Future Symposium Demonstration of Measurement Derived Model-Based Adaptive Wide-Area Damping Controller on Hardware Testbed L. ZHU, F. BAI, 2, Y. LIU, H. LIU, Y. MA, Y. LIU University of Tennessee Knoxville USA 2 Southwest Jiaotong University China E. FARANTATOS 3, M. PATEL 3, S. MCGUINNESS 3 3 Eletri Power Researh Institute USA SUMMARY One of the main drawbaks of the existing osillation damping ontrollers that are designed based on offline dynami models, is adaptivity to the power system operating ondition. With the inreasing availability of wide-area measurements and the rapid development of system identifiation tehniques, it is possible to identify a measurement-based transfer funtion model online that an be used to tune the osillation damping ontroller. Suh a model ould apture all dominant osillation modes for adaptive and oordinated osillation damping ontrol. This paper proposes a measurement derived model-based adaptive wide-area damping ontroller. Firstly, the system model in the form of multiple-input multiple-output AutoRegressive Moving Average with exogenous inputs (ARMAX is identified online by using ambient data or ringdown data. Seondly, the frequeny and residue angle of the target osillation mode are derived from the identified model to update the ontroller parameters. Thirdly, a time delay ompensator employing the lead-lag struture is proposed to ompensate for random time delay. Assuming onstant time delay in one ontrol yle, the parameters of the time delay ompensator are updated based on the measured time delay in eah ontrol yle. Finally, the proposed adaptive wide-area damping ontroller is implemented on a hardware testbed, whih emulates a two-area four-mahine system. The testing on the hardware testbed demonstrates the feasibility of pratial realization of a measurement derived model-based wide-area damping ontrol system for small and large disturbanes over a wide range of operation onditions. KEYWORDS Adaptive Wide-Area Damping Controller, Wide-Area Measurement System, System Identifiation, Hardware Testbed, Time Delay Compensation lzhu2@utk.edu

2 . Introdution In today's interonneted power grids, low-frequeny osillation is a signifiant issue limiting the power transfer apability and even deteriorating the power system seurity. Loal and wide-area power system stabilizers (PSSs are installed or proposed to provide supplementary damping ontrol through generator exitation systems [], flexible alternating urrent transmission systems (FACTS devies [2], and high-voltage diret urrent (HVDC links [3] to suppress these low-frequeny osillations. However, one of the main drawbaks of the existing osillation damping ontrollers that are designed based on offline dynami models, is adaptivity to the power system operating onditions. If the atual operating ondition is signifiantly different from the typial operating onditions onsidered in the offline design proedure, the ontroller's performanes may degrade. In some extreme ases, they even provide negative damping. A robust ontrol sheme an be utilized to improve adaptivity. In general, a robust osillation damping ontroller is designed based on a detailed system model under a seleted dominant operating ondition with bounded model unertainty [4], [5]. The variations of operating ondition are refleted in the additive and/or multipliative unertainty of the system model. Nevertheless, it is not easy to determine the unertainty boundary of the system model, and the ontroller performane may not be optimal when the atual operating ondition deviates from the dominant one. With the inreasing availability of wide-area measurements and the rapid development of system identifiation tehniques, it is possible to identify a measurement-based transfer funtion model in an online environment. Sine the system model an depit all the dominant osillation modes, it is feasible to optimize the ontrollers' parameters at the ontrol enter, and remotely onfigure the parameters of dispersed damping ontrollers. In this way, the ontroller parameters an be updated online to trak the ontinuous variations in operating onditions. For instane, a self-tuning adaptive PSS based on artifiial neural networks is proposed in [6]. In [7], the parameters of phase lead-lag ompensators are updated based on the online modal analysis. This paper presents a Wide-area Damping Controller (WADC based on a measurement derived model. The entire power system is represented by a linear transfer funtion in the form of multipleinput multiple-output (MIMO AutoRegressive Moving Average with exogenous inputs (ARMAX, and the ontroller parameters are updated based on the identified model to improve adaptivity. Additionally, an adaptive time delay ompensator employing a lead-lag struture is utilized to redue the impat of random time delay. The effetiveness of the proposed measurement-based adaptive WADC has been demonstrated in a two-area four-mahine system on the HTB under various disturbane senarios. 2. Design of adaptive measurement model-based WADC 2. Overview of the WADC Fig. shows the overall arhiteture of the adaptive WADC. The adaptive WADC is designed to damp a ritial inter-area osillation mode by providing supplementary damping ontrol signal through generator exitation system. The adaptive WADC onsists of two parts: onventional ontroller employing lead-lag struture, and a time delay ompensator. The entire power system is represented by a MIMO ARMAX model with generator bus frequeny as output signal. The identified system model will be used to alulate the eigenvalues of the osillation modes, whih an be used to update the parameters of the wide-area PSS. The time delay ompensator is used to redue the impat of random time delay due to ommuniation. 2.2 System identifiation In the disrete-time domain, the entire system an be represented by a MIMO ARMAX model as [8] = + ( where y is the vetor of outputs, is the exogenous part whih is the vetor ontaining the known exitations, and is the moving average part whih is the vetor with unknown noise., and are the autoregressive polynomial matrix, the exogenous polynomial matrix, and the moving average polynomial matrix, respetively. is the shift operator.the matries αz, βz and γz in ( an be expanded as (2-(4. The oeffiient matrix an be alulated by using two-stage least square algorithm [8].

3 =! '+( = ( * ( ** * ++( * *. /. / * + -. / (2. / ** *. / (3 3. * * * =! '+( ( (4 * 3. * * 9 5 *8 where n ;, n < and n = are the orders of the outputs, exogenous inputs, and noise, respetively. αz is an m m matrix, βz is an m p matrix, and γz is an m q matrix. Atuation signal Observation signal Identified system model Measured time delay V WADC V Ref + V PSS + Σ + Time delay ompensator Exiter G Wide-area PSS Rotor speed Wide-area measurement Loal PSS Fig.. onsidering time delay The online model identifiation is triggered by system events inluding generation trip, load shedding, and topology hanges due to line trip, et. and the model will be updated. In addition, the identifiation proedure an be triggered by a predefined timer (periodial trigger. If there is no system events, the model will be updated using olleted ambient data every 5 minutes. After identifying the ARMAX model, it is neessary to validate whether the model is good enough to desribe the system osillatory harateristis [9]. The identified ARMAX model is validated in both time domain and frequeny domain. In time domain, the response of the identified model is ompared with atual system response. In frequeny domain, the eigenvalues alulated by the denominator polynomial of the MIMO ARMAX model are ompared with results of Matrix Penil (MP analysis of the measurement data. MP is a modal extration tehnique (similar to Prony method, whih effetively estimates the dominant modes' information in a response []. 2.3 Parameter update of wide-area PSS As shown in Fig., the transfer funtion of a lassial wide-area PSS is H WADC T s + st + T s + st w 2 ( s = K ( WADC where T and T 2 are the lead and lag time onstants, respetively. T w is the washout onstant, K WADC is the gain of the WADC. φnmk T =, sin T2 = α T, α = 2 (6 ω α φnmk + sin 2 w 2 (5 2

4 where ω is the osillation frequeny of the mode λ NMk, φnm k is the residue angle of the mode λ NMk. The system model identified by using ambient data or ringdown data is used to alulate the eigenvalue of target osillation mode, and residue angle of the seleted ontrol loop. Then, the time onstants T and T 2 an be updated to trak ontinuous variations of operating ondition. 2.4 Parameter update of time delay ompensator In order to eliminate the effets of the time delay, the following transfer funtion will be used to ompensate the phase lead/lag and the gain drift: + st 2 H ( s = K ( (7 + st 2 ωτ sin where T =, T2 = α T, α 2 =, K = β, < β <. ω α ωτ γ + sin 2 where ω is the osillation frequeny of the mode λ NMk, τ is the measured time delay, and β an be adjusted aording to the performane of the ompensation. Assuming the delay is onstant in one ontrol yle, the parameters of the time delay ompensator ould be updated to redue the impat of random time delay. Aurate time from the global positioning system (GPS would be reeived loally in both the PMUs and the time delay ompensator. Wide-area phasors measured by the PMUs at time ta are olleted and resynhronized by the Phasor Data Conentrator (PDC and in turn proessed by the adaptive WADC to generate the wide-area damping ontrol signal with a time stamp [ta]. When the loal time delay ompensator, whih is installed lose to the generator exiter, reeives the ontrol signal [tb], the signal will be relabelled with a new time stamp [tb] also obtained from GPS and the exat time delay an be alulated aurately as τ = tb-ta beause of the high-resolution time servie provided by the GPS. 3. Demonstration on hardware testbed 3. Implementation of WADC on hardware testbed The hardware testbed in the National Siene Foundation and Department of Energy (NSF/DOE engineering researh enter-center for Ultra-wide-area Resilient Eletri Energy Transmission Networks (CURENT, is a platform built for power grid ontrol methodology test and demonstration. The two-area four-mahine system as shown in Fig. 2 is now emulated on the hardware testbed, whih provides a perfet environment for WADC implementation, testing, and demonstration. The onfiguration of the hardware testbed is shown in Fig. 3. A LabVIEW-based ontrol system has been developed to emulate some funtions of an atual power system ontrol enter. It gathers the measurement data from monitoring devies in the HTB and sends ontrol ommands to virtual generators through Ethernet. MATLAB ode of an adaptive WADC was integrated into the HTB ontrol software using standard LabVIEW interfae. The onfiguration and parameters of this twoarea four-mahine system are desribed in []. Fig. 2. Two-area four-mahine system. 3.2 Controller performane Fig. 3. HTB onfiguration. In order to test the effetiveness and robustness of the proposed adaptive WADC, the following four typial senarios and the aptured dominant mode during different operating onditions are shown in Table. The parameters of the adaptive WADC in eah ase are listed in Table 2. The time 3

5 delay in the ommuniation hannel of HTB is ms to 3ms. Table Case Study Details No. PSS Event Change Frequeny Damping Event Type Loation Loation (p.u. (Hz Ratio G,G3 Load Inrease L2.4 to % 2 G,G3 Generation Inrease G.3 to % 3 G Load Inrease L.4 to % 4 G Generation Trip G3.4 to % Table 2 Parameters of No. T w T T 2 K WADC The proposed adaptive WADC is ompared to the onventional WADC proposed in []. Under the operating ondition in Case shown in Fig. 4 (a, both the adaptive WADC and traditional WADC an suppress the osillation. However, when the operating ondition is hanged from Case to Case 2 or Case 3, the onventional WADC not only annot suppress the osillation but also triggers unstable osillation while the adaptive WADC still has good ontrol performane. From Fig. 4(d, it an be found that the onventional WADC even beomes an osillation soure. Note that the onventional WADC is tuned and tested in a similar operating point as Case, but the ontrol performane degrades when the operating ondition hanges, while the adaptive WADC updates the parameters with the identified SISO model that an trak the hanges of the operation ondition. For Cases and 2, the osillation an be damped even without ontrol ations, but the settle-down time was shortened after the adaptive WADC was implemented. In Cases 3 and Case 4 there is a poor damping without ontrol. Therefore, the WADC an adapt to a wide range of operating onditions and improve the stability of the power system. Frequeny Differene (Hz Fr equeny Di f f er ene( Hz Ti me( s Ti me( s (a Case (b Case Ti me( s ( Case 3 (d Case 4 Fig. 4 Controller performanes omparison in different ases Fr equeny Di f f er ene( Hz Frequeny Differene(Hz Time (s 4

6 4. Conlusion In this paper, an adaptive wide-area damping ontrol system fully based on measurement signals of the power system is proposed and verified. An adaptive WADC and loal time delay ompensation system was designed that an update the parameters online with the predition of the dominant mode and ompensate the time delay loally. The implementation on the hardware testbed demonstrates the feasibility of pratial implementation of a measurement derived model-based wide-area damping ontrol system for small and large disturbanes over a wide range of operating onditions. BIBLIOGRAPHY [] M. About-Ela, A. Salam, J. MCalley, and A. Fouad, Damping ontroller design for power system osillations using global signals, IEEE Trans. Power Syst., vol., no. 2, pp , 996. [2] Wei Yao, L. Jiang, Jinyu Wen, Q. H. Wu, and Shijie Chen, "Wide-area damping ontroller of FACTS devies for inter-area osillations onsidering ommuniation time delay," IEEE Trans. Power Syst., vol. 29, no., pp , Jan. 24. [3] Fuhs, A., Markus Imhof, Turhan Demiray, and Manfred Morari, "Stabilization of large power systems using VSC-HVDC and Model Preditive Control," IEEE Trans. Power Del., vol. 29, no., pp , Jan. 24. [4] Chuangjiang Zhu, Khammash M., Vittal, V., and Wenzheng Qiu, "Robust power system stabilizer design using H loop shaping approah," IEEE Trans. Power Syst., vol. 8, no. 2, pp. 8-88, May, 23. [5] Majumder, R., Pal, B.C., Dufour, C., and Korba, P., "Design and real-time implementation of robust FACTS ontroller for damping inter-area osillation," IEEE Trans. Power Syst., vol. 2, no. 2, pp , May, 26. [6] D. K. Chaturvedi and O. P.Malik, Generalized neuron-based adaptive PSS for multimahine environment, IEEE Trans. Power Syst., vol. 2, no., pp , Feb. 25. [7] J.B. Zhang, C. Y. Chung, C. Lu, K. Men, and L. Tu, A novel adaptive wide area PSS based on output-only modal analysis, IEEE Trans. Power Syst., to be published. [8] L. Ljung, System Identifiation: Theory for the User, 2nd ed. Englewood Cliffs, NJ: Prentie- Hall, 999. [9] H. Liu, L. Zhu, Z. Pan, F. Bai, Yong Liu, M. Patel, E. Farantatos, N. Bhatt, "ARMAX-based Transfer Funtion Model Identifiation Using Wide-area Measurement for Adaptive and Coordinated Damping Control," IEEE Transation on Smart Grid, under revision. [] M. L. Crow and A. Singh, The matrix penil for power system modal extration, IEEE Trans. Power Syst., vol. 2, no., pp. 5-52, Feb. 25. [] L. Yang, Y. W. Ma, J.X. Wang, J.Wang, X.H. Zhang, L. M. Tolbert, F. Wang, K. Tomsovi, "Development of onverter based reonfigurable power grid emulator," in Pro. of IEEE ECCE, Pittsburgh, USA, Sep. 3-8, 24. 5