svc control

8/2/2019 SVC Control

1/12

524 IEEE ransactionsonPower Systems,Vd.6,No.2,May 1991

ADVAN CED SVC CONTROL FOR DAMPING POW ER SYSTEM OSCILLATIONSE. Lerch D. Povh, Sen ior Member, IEEE

SiemensAG ErlangenFed. Rep. of Germany

L. x uZhejiang UniversityP. R. China

Keywords:Static Var Compensation, power oscillations, damping,local phase angl e, estimation.

AbstractDynamic reactive power compensation is used to an in-creasing extent to improve voltage and reactive powerconditions in ac s stems. Additional tasks can also beperformed by the itatic Var Compensators (SVC) to in-crease the transmission capacity as result of employment ofSVCs for power oscillation damping. This is of particula rimportance in the case of weakly coupled power systems.A new SVC control for damping of power system oscilla-tions ha s been developed. To increase system damping SVCuses phase angle signal estimated from the measurementof voltage and power at th e SVC location. By means of a noptimization and identification procedure optimized designof the dampi ng control with variou s control concepts can bedetermined taki ng into account non-linear power systems.As a result of this method it is possible to increase powersystem damping considerably, in particular in critical situ-ati ons close to the st abili ty limit, usi ng only locally meas-ured state variables at the SVC thus leading to increasedtransmissi on capability of the power system.

1. IntroductionIn recent years SVC ha s been employed to an increasingext ent since dynamic reactive-power control gives con-siderable advantages for power system operation. Besidesto the voltage control as a main task SVC may also beemployed for additional tasks resulting in improvement ofthe transmission capability.90 3 1 L60-6 PWRSby the IEEE Power System Engineering Committee of theIEEE Power Engineering Society f o r p r e s e n ta t i o n a t t h eIEEE/PES 1990 Summer Mee ting , Min nea pol is, Min nes ota,July 15-19, 1990.7990; made available f o r pr in ti ng June 21, 1990.

A paper recommended and approved

Xanuscript submitted January 29,

An important a spect when using SVCs is damp ing of poweroscillations. Damping of power syst em oscillations plays a nimportant role not only in increasing the transmissioncapability b ut also for stabilization of power system condi-tions after critical faults, particularly in weakly coupledsystems. In this paper the use of SVCs for damping crit icalpower system oscillations on the basis of local stat e vari-ables is described.To achieve t his objective it is necessary t o improve t he SVCcontrol concept by introduci ng signals which reflect powersystem oscillations. The normally used SVC voltage controlis not in a position to effectively d amp t hese oscil lations. Insome critical cases the voltage control can even amplifyoscillations. The optimum variable would be the phaseangle difference of systems which oscillate w ith respect t oeach other. In absence of a telecommunication link thisvariable is not available. However, an estimation of phaseangle difference can be carried out at the point of instal-lation of SVC which is adequate for improvement of powersystem damping.2. Estimation of Machine Phase Angles UsingVoltage and frequency and in addition active and reactiveg w e r and currents in t he incoming lines to the node of th eVC are locally available as measured variables. The useof these local state variables only is an objective to calcu-late the phase angle difference and to use this signal fo rreduction of power system oscillations. The realization ofthis new method became easie r through the development ofdigital SVC control which is able to calculate th e estimatedphase angle difference.In a number of publications various concepts for dampingof oscill ations b means of frequency correct ion signalshave been descriged [l , 21. This method can a lso be derivedfrom equation (A14) in the appendix. The frequency is anadequate control variable for example in a relatively smallsystem, oscillating against a relatively lar ge system whosefrequency is hard1 affected b the power oscillat ions I 7 .The use of a localry measuref frequency is suitable onlywhen the power system oscillation frequency can be clearlyfiltered. In th e case of loosely coupled power systems thisrequirement is not always satisfied so that filtering of th eoscillation signal from the influenced frequency is compli-cated [3].Another often used control signal is the measuredvalue dPldt [41.For large phase angle values close to the synchronizationlimit erroneous signals can be generated since the phaseangle deviation and the active power flow may be of op -posite phase .

Local Stat e Variables

0 8 8 5 - 8 9 5 0 / 9 1 / 0 5 0 0 4 5 2 1 . ~ 1 9 9 1EEE


2/12

525A new method will be described using local measurementof voltage, active power and reactive power flow to derive asignal for the phase angle of th e generators with respect tothe SVC (reference node). A good estimation is, however,only possible if the desired phas e angle i s observable in t hepower flow at th e location of th e SVC.If the SVC is considered a t th e node j in a power systemwith a generator or subsystem k the complex voltage dropE& can be derived from Fig. 1.

--I

where5 Vj is taken a s reference.The voltage angl e difference between Vj and & s then

(2 )

If E_k is defined as t he voltage behind the transient ma-chine reactance x'dk which can be taken into account inY'k, 6k j gives an estimate for the phase angle of the gener-d o r a t he node k with respect to the load angle at node j .As a result of power system reduction to a two generatorsystem taking t he SVC node into account, the coupling ad-mittance Y .k can be calculated for an actua l power systemcondition.$he phase angl e difference between the systemsis obtained by eliminat ion of th e reference angle of the SVCnode. By means of this difference signal the oscillation ofthe two power systems with respect to each other can beestimated. Fig. 7depicts a comparison between estimatedand ex actly calculated load an gle difference of both gener-ator systems. The rough estimation is sufficient t o approxi-mate the phase angle difference. N o attempt was made toimprove matching since the phase angle of both signals isin close agreement and this is essential to give the propersignal for damping control. In a multi-genera tor system thephase ang le differences of individual systems with respectto each other can be selectively calculated if these a recoupled via th e SVC node (measur able and observable ).In order to check the sensitivit y of feedback with respect tothe equivalent reactance load switching was performed inboth power systems by opening and closing one of thedouble circuit lines of the studied system (Fig. 2 ) . Theequivalent reactance of the unfaulty power system wasadequate for estimation of the phase angle within thispower system configuration.

Fig.1 Definitions for estimated phase angle 6 (reference node )kiIn the case of more complex systems i t is possible to provideadaptive matching of the equivalent reactance; deter-mination and matching of the equivalent reactances is alsopossible on-line by using extended Kalman filter [51.

2.1 Effect of SVC on Oscillation Behaviour ofOn the basis of a single generator system connected to afixed frequency power system described in the appendix itcan be shown that there is a direct correlation betweenalteration of th e voltage a t the SVC and alteration of th ephase angle of the generator. Therefore, damping of thesystem oscillations cannot be directly influenced by thevoltage control of the SVC. However, i t is possible to in-crease damping if the voltage of the SVC is controlledlinearly as a function of the rate of change of the phaseangle (change in generator speed) [ 8 ] .The effect of SVC onth e improvemen t of dampi ng conditions, however, de-creases with the increased power system short-circuitcapacity. However, in case of high sh ort circuit capacity th eSVC location is also not su itabl e for voltage control.The phase angle of gener ators seen from th e location of theSVC is estimated on the basis of these theoretical con-siderations in o rder to control the SVC.

Generators

2 .2 Control Concept Employing a Local PhaseThe configuration depicted in Fig. 2 was investigated inorder to demonstrate the basic effectiveness and robus tnessof the new local damping signal.The 600 km long 500 kV double circuit line connects twopower systems with a total capacity of 6600 MW.Approximate1 40 % of the charg ing capacity of th e line iscompensated gy means of shunt reactors. Under steadystat e conditions 815 MW are fed to power system 1. Acontrol range of + 20 0 Mvar was selected for the SVC. thesteady state condxions are given in Fig. 2 .

Angle Signal

fault locationvoltage control

6000 MVA 500 kVH=7s

815 MW15 MW

f200 Mvar 1300 MW2000 Mvar 500 Mvar4441 MW2449 Mvar

5200 MWline datar=0.028 aKm, x=0.26 alKm, c=14 nF/kmFig. 2 Single line diagram and pre-fault conditions for two area system

The effectiveness of SVC for damp ing oscillations is limitedby the ma ximum rating of the SVC. Maximum damping isthus achieved employing bang-bang control with correctphase angle of the signal thus utilizing the maximum SV Crating [41. Fig. 3 depicts a SVC control employing adamping signal. Additional filters are required in order t ofilter out interference signals from the relevant frequencyrange of oscillation from 0. 3 to approximately 2 Hz . Thetransfer function to filter out harmonic content in theestimated phase angle signal from ( 2 1

was employed for bang-bang control. Parameters can bedetermined by a optimization procedure in the NETOMAC


3/12

526

SVC-sig.

V-control sig. TCR and TSCvoltage control firing circuit

bang-banglinear control 8 -control Sig.filtering - r /and

- -$. - -easured atSVC-node -= SVCl6ode - -

Fig. 3 Concept of SVC damping controlprogram [6],whereby th e ma ximum damping is calculatedtak ing into account the whole non-linear system. The mea-suring equipment for Pjk, &, v . i n (1 ) was modelled as afirst order time delay with 2d ms helay time.If the target function of oscillation damping for optimiza-tion procedure is minimized

z = p P - z d r +m i n (4 )where a bang-bang system i s designed according to (3 )andrepresents t he change i n active-power flow between system1 and system 2 at the SVC node, the parameters of thetransfer function Gl(s)can be obtained.The para meters of Gl(s)are, however, not optimal with re-gard to higher frequency content of the bang-bang oscil-lation. Parameters, determined by means of an identi-fication procedure, however, minimize th e overall time be-haviour of AP1.2. Consequently the higher frequency sig-nals of the SVC are additionally evaluated in the proce-dures and suppressed by matching the parameters.The optimizatiodidentification rocedure is a special modeof NETOMAC program which calcula te a la rge number ofalternatives using various parameters and determinesautomatically the optimum parameter set according to atarget function.The system dynamics were simulated with the NETOMACprogram including the control concept depicted in Fig. 3whereby the SVC was modelled a s variable susceptance.Fig. 4 depicts optimized employment of the SVC for os-cillation damping with the SVC capacity of f200 Mvar.Parame ters of Gl(s)are shown later in Fig. 8.System reac-tion without SVC is also shown. A three-phase fault in thevicinity of system 2 of 70 ms durat ion was assumed to bethe cause of power system oscillation.Power oscillations of approximately 0.5 Hz and amplitudesin excess of 500 MW (AP1-2 occur at th e transmi tted powerof approximately 815 MW (P2-1)after the fault in case SVCis not in operation. The system is operating at its limits.Oscillations are weakly damped. At transmit ted power inexcess of 905MW the power system would become u n -stable. If damping is defined over the area under AP1-2 inaccordance with equat ion ( 4 )without and with SVC, amp-ing due to th e damping control is increased by 788 Fig. 5 )

As shown from Fig. 4, the change of local frequency charac-teristic (Af at SVC node) is not suitable to be taken asinpu t signal for th e damping controller because of difficultyto filter out the low frequency signal of generator oscilla-tion.without SVC in operatlon

' f chan e in frequencyat ~ 3 c - n d e8 - controlledSVC

AV - h h r -at SVC-node

0. 1.5 3.0 4.5 6.0 7.5sec~9.0Fig. 4 System oscillation without SVC and with

6 - controlled SVC in operationThe influence of the SVC ra ting on the reduction of oscilla-tions can be seen from Fig. 5 where the damping near theinstabilit y of the s yste m in absence of an SVC (Fig. 4) isdefined as 1pu. SVC ratin gs of k 00 to k 500Mvar havebeen taken into account.Voltage control of SVC is not able t o damp power s stemoscillations. The transmitted active power is sole& de-pendent on the phase angl e difference of the two power sys-tems and the SVC used to maintain a constant voltage in-creases the synchronizing torque and SVC in voltage con-trol mode acts to increase stability limit. This influencegoes, however, hand in h and with a reduct ion of power sys-tem damping Fig. 6shows a s a n example th e unfavourablephase angle ( A a) of the voltage sign al for power system os-cillation damping compared wi th the optimum phas e anglesignal.

7 .

1 0 t200 t500Fig. 5 Damping of system oscillations as function ofSVC-rating based on the case without SVC ( Z,system conditions according to Fig. 2

QSvcWvar) +


4/12

527-SVC-sig. -6 -controlled V-controlled - -

~-

- _ _I I

AV V-controlled- - - - -V V V V kat SVC-node

0. 1.5 3.0 4.5 6.0 7.5sec,9.0Fig. 6 SVC operation with voltage and phase angle control

( system condition according to Fig.2 )Under the selected load conditions system damping mayalso be increased by the classical correcting signal dP/dt;however, if the transmi tted active power is increased closeto the stability limit, e. g. to 92 5 MW, the robustness of thenew signals becomes apparent (Fig. 7 ) especially in theregion of initial oscillation (phase angle difference in ex-cess of 90"). In the case of classical dP/ dt control the re couldbe a faulty control order in the time interval A t dependingon the saddle in the power flow. Using the same para-meters for filtering and control a dP/ dt controlled dampingresults in up to 10 % less system damping in this casedepending on the prefault transmitted power. Comparisonof the estimated phase angle 8estim with the actual value82.1 (Fig. 7 ) demonstrates the robustness of the selectedcorrection signal at this critical load situat ion close to t hestability limit. The phase of actual value and estimatedvalue are practically equivalent.As shown here the estimated signal contains the suddenchanges in the SVC-voltage. The assum tion of constantSVC-voltage simplifies equations (1)a n 8 (2 ) and reducesthe calculation accuracy, but with the advantage t hat thefiltering transfe r function can be neglected. Com arison ofthe damping effects of both signals with and wit1out SVCin this way shows that the filtering transfer function re-duces the damping up to 15 % depending on the washouttime constant Tw from(3).In many cases linear control is very effective to solve dyna-mic stability problem caused by small disturbances. But forthe transient stability problem there is much larger controlarea needed. Damping of power system oscillations bymeans of bang-bang feedback results therefore in optimumutilization of the available SVC rating. To get the samedamping effect, bang-bang control needs less investment.Meanwhile bang-bang control ha s some disadvantages:--

Optimization is mathematically complicated. Reali-sation is more difficult tha n for l inear controller.High non-linearity of the signal introduces largeamount of harmonic components in system voltageand current and may result in difficulties for correctSVC control ac tion.

AV A AP.V v t SVC-node

Sestim0. 1.5 3.0 4.5 6.0 7590Fig. 7 Transmitted power increased to 925 MW( system condition according to Fig. 2)

In later stage of transient process when oscillatioh isnot anymore severe, bang-bang control may haveadverse effect (hig h frequency oscillation).

To tackle the above mentioned problem, control modifica-tion ha s been introduced. In the first stage of the trans ientprocess bang-bang control alone is effective. Later, whenoscillation has been already damped a normal linearcontroller including a PD-block and a differentiation blockis introduced.

K=600.


5/12

528

SVC-sig. :

.25 pu

blockedA /

vdta e controlin oseration\v

Fig. 9 Combined bang-bang and linear SVC controller( system condition according to Fig. 2 )Damping of power system oscillations improves pro ortionally with the large SVC rating. However, this lea& tiincreasing reactive power oscillations and thus to voltagedeviations during the phenomenon. Optimized designbetween the damping effect and the limitation of voltagedeviations is possible. The results for various targetfunctions and parameters are collected in Table 1,whichtake active and reactive power oscillations and voltagedeviations into account. Low value for K v an d K Q meanspreference for damping effect and hi gh val ue preference forthe limitation of voltage deviation and reactive poweroscillation during damping control action.

0 KQ in(pu)I(a QsvCWar) ZlZg

5 Z2 Qsvcadditiona! parameter 2 266 068for optimization 5 2188 0.38Tab. 1 Optimized system parametersZ : ref. for evaluation of damping , expressed in equ. (4 )AP1-2, A Q ~ - ~n pu, based on 200 MVA , AQsVc in kV / MVA

In the investi gated cases the used turbi ne governor show afast reaction, sensitive to the SVC-control. Slow turbinegovernors will be less affected by SVC bang-bang control.When turbine governor has fast control behaviour one canobserve the effect, that a linear control concept mightresult in a better system damping.

3. 3-Generator SystemsTo verify the effectiveness of t he proposed controller, fur-the r digital simulations ar e performed for various 3-gener-ator systems of different network structures and differentoperation conditions.All 4 systems, as shown in Fig. 10, are developed from th eabove used two are a system by connecting a new generat orG3 with th e rati ngs of generators G1. G2, G3 being

4000 MVA, 3000 MVA an d 2000 MVA and the inerti a con-stan ts being 5 s, 5 s and 3.5 s, respectively. Each rtiachineha s a voltage controller with PSS. All the other machineparamete rs are ust the s ame as those used above.A SVC with rating of + 400 MVar is installed at themiddle bus between G1 an d G2. The same SVC-control hasbeen used as in the two area example because simulationhas shown that larger changes in controller structure orcontroller parameters a re not necessary.For comparison, the fault of same type and same durationis also applied close to bus 2 in all studied network con-figurations. The damping control signal is the same phaseangle difference between G1 and G2 as in the two gener-ator system. Simulation results show, that the proposedcontroller is also effective in damping the oscillations inthe enlarged system.System I differs not very much from the original two areasystem, if one considers G1 and G3 as a n equivalent gener-ator. The introduction of G3 produces additiona l oscillationmodes between the generators, but it's influence is not sosignificant because of its location and relatively smallerrating. Therefore we get similar result as for the two areasystem.In system I11 the distance between kl and G3 is eitendeato 300 km. The connection between G1 and SVC is changedfrom double line to a single line and new single line isadded between SVC and G3.Considering G1 and G3 as a subsystem, conditions are as insystem I. So the results a re also similar to tha t of system Iwit h a cert ain difference in t he tran sien t process of G3.In the radial system 11,the coupling between G1 and G3 be-comes weaker by deleting the direct line between them.Consequently, the swing mode between both gene rators be-comes less important, while t he other 2 swing modes, i. e.,the swing between G1 and G2 are effectively controlled bySVC. As seen from the reduced time to dam p out the oscil-lation G3 increases the synchronizing torque in the system.In the system IV, there is a direct connection between G2and G3. G3 has to the SVC location the same position as G1and G2. It is obvious tha t the power flow through thi s linecan hardly be controlled by the SVC. Only small dampingby the PSS of G3 is seen i n t he oscillogram. Therefore theoscillation in thi s system lasts much longer th at i n othertest systems. But the considerable damping effect can stillbe observed in this case.A fault on the line 2-3, cleared by disconnecting this line(changing the system structure), was also calculated anddamped out without problems.4 . ConclusionA new method is described which defines the phase angle ofgenerators on the basis of voltage and power measurementa t the location of SVC. This sta te variables are employedfor improvement of damping of power system oscillationsby SVC. The new SVC control can optimally damp activepower oscillations. The new damping signal is shown to berobust in particular in vicinity of the stability limit andhas the advantage that no error signals occur at largedifferences of the phase angle.The estimat ed phase an gle can be utilized for various con-trol concepts. On the basis of a bang-bang control combinedwith a linear controller the optimized control loop designwas demonstrated in a fully non-linear system with differ-en t objectives. By means of paramet er identification proce-dure optimum feedback can be determined. As a result offreedom with regard to the optimization criteria varyingtechnical optimization factors can be take n into account.Tests on a simulator ar e planned, but not yet realized.


6/12

529

System I System I1

System 111 System IV800 M W, 800 MW ,

P21 734 MW

300 km

12 ec

Fig. 10 3 areas-system with bang-bang and linear SVC control

Fl = 549 MW '- 300 km

I00 kmA V at SVC

12 ec800 MW APG3

~ 8 -contolled SVC operationSVC not in operation_ _ _ _


7/12

Definitionsinert ia of machinemechanical machine constantmachine phase angledamping coefficientactive electrical powergenerated active powerturbine powervoltage behind transient reactancevoltage dropadmitt ance node j-kshunt admittance node jactive power flow node. kreactive power flow nodi j-kcomplex numberweightin g factors for voltage-and reactive power influenceSVC damping controller gain

operating pointlinearized or change inReferencesE. V. Larsen, H. H. Chow:"SVC control design concepts for systemdynamic performance"Sympos. on Appl. of stati c VAR systemsIEEE PES, San Francisco, July 1987,pp. 36-53J. R. Smith, D.A. Pierre, I. Sadighi, M. H.Nehrir"A supplementary adaptive var unit controllerfor power system damping"IEEE TRANS on Power Systems, Vol. 4, No. 3,August 1989, pp. 1017-1023A. G. Phadke, M. G. Adamiak, J.S.Thorp:"A new measurement technique for trackingvoltage phasors, local system frequency, andrate of change of frequency"IEEE TRANS on Power App. Syst., Vol. PAS-102,No. 5, May 1983, pp. 1025-1039A. Olwegard, K. Walve, G. Waglund, H. Frank ,S. Torseng:"Improvement of transmission capacity bythyristor controlled reactive power"IEEE TRANS on Power App. Syst., Vol. PAS-

S. Yokokawa, Y Ueki, H. Tanaka , e. a.:"Multivariable adaptive control for thermal100,NO. , August 1981, pp. 3930-3939

generator"IEEE TRANS on Energy Conversion, Vol.3,No. 3, September 1988, pp. 479-486B. Kulicke, H.-J. Hinrichs:"Parameteridentifikation und Ordnungsreduk-tion mit Hilfe des SimulationsprogrammesNETOMAC"etz-Archiv, Bd. 10 (1988) H. 7, pp. 207-213W. Bayer, N. Sudja:"Use of Static Compensators to provide faststand-by reactive power for a 500 kV remotetransmission line"6. Cepsi Conference, Jakarta Indonesia, 1986W. Bayer, P. Sihombing:"Aspects of damping of power oscillations bypower modulation"7. Cepsi Conference, Brisbane Aus tralia, 1988

Appendix: S ingle Generator Sytem with SVCThe classical machine equation for a single generator sys-tem with a SVC and a fixed frequency power system asdepicted in Fig. A1is taken as a starting point. IM - + D - + P = P m26 d 6

dt2 *If UG = Unet = 1 pu, UG an be written a s& = cos6 + j s i n 6For gsvc it follows tha tlJsvc= Us vc (cos 6/2 + sin 8/21The power flow in the system is given by

The active power transm itted between generatorand power system is found from (3) o beP = (UsvJX) sin 6/ 2From th e difference of,the reactive power flowsQSVC-net QG-svc it can be found for SVC tha t

as per figure A1Qsvc= (usvc-o - US "C) ysvcIf equations (A7) and (A8) are linearized following rela-tionships are found for the steady state operating point(Usvc-0= u o , 60 ) .

an dA us, ysvc (A10)

KSVC'y-,xvc "svc -0Fig. A1 Basic diagram of one generator - SVC - system


8/12

53For A Usvc it follows tha t

U,sin 6d 2Ysvc X + 2 ( 2 U , - COS^J2 UsVC = - A6

Linearization of equation (1) esults ind2A6 dA 6dt2 dtM - + D - + A P = A P m

( A l l )

(A121

Alte ration of act ive power is found from (5)A P = [ C l l X ) ~ i n 6 ~ / 2 ]lJ ,,+ [ (UJ2X)cos6 ,12] A6 (A131Equation (A131 shows that it is possible to increasedamping of the system if Usvc is altered linearly as afunction of the rate of change of the machine phase angle,that is

d A6[ ( l ~ X ) s i n 6 0 1 2 1 A U s v c = K-t (A141

Edwin N . Lerch was born in Germ anyin 1953. He received Dip1.-Ing. degreefrom the universi ty of Wuppertal 19 79and complete his Ph. D. in 1984 inelectrical engineering. Since 1985 he isa member of the high-voltage trans-mission engineering and system plan-ning department a t Siemens, Erlangenin the industrial power system andmachines group. He is working onpower system stability, dynamics ofmultimachin e systems, control, optimization an d identifi-cation problems i n electrical power systems. He is memberof th e ETGIGMA group 3 network control since this year.Dugan Povh was born in BeogradlYugoslavia in 1935. He receivedDip1.-Ing. degree from UniversityLjubljana/Yugoslavia in 1959, Dr.-Ing.degree from TH Darmstadt/Germanyin 1972 and is also Professor at theUniversity of Ljubljana. He is active ina number of committees and workinggroups of CIGRE and IEEE.His areas of interest are systemanalysis, network planning, insulationmission systems and development of HVDC a i d static va rcompensator te chnique. Prof. Dr. DuEan Povh is the head ofthe department on system planning in the Siemens PowerTransmission and Distribution Group.Liwen Xu was born in HangzhouP. R. China on May 25 , 1956. He re-ceived his B Sc. from HuazhongUniversity of Science and Technologyan d M. Sc. from Zhejiang University,both in electrical engineering. Cur-rently he is working toward the Ph D.degree at Zhejiang University underthe supervision of Prof. ZhengxiangHan. At present he is visiting Siemensand works on the SVC control. Hismain interest is in the field of power system stabilitycontrol.


9/12

532DISCUSSION

A. H a m m a d a n d M. Ha eus ler ( AB B Po wer Sy s t em s ,B a d e n , S w i t z e r l a n d ) : The authors are to be commendedfor an interesting paper.Static VAr compensators - in the form of thyristorcontrolled reactors and/or thyristor switched capacitorshave been in use in transmission networks since the early1970's [A] . Controls of a typical multi-task SVC comprise thefollowing [B]:1. dynamic voltage control2. system stabilization3. power oscillation damping4. small signal damping5 reactive power control (slow)Local measurements at the location of SVC are sufficientfor estimating the phase angle across the transmissionsystem. The deviation of the estimated phase angle acrossthe transmission system is a robust control technique fordamping power oscillations. Combining a bang-bangcontrol strategy with a linear (smooth) control proved tobe an efficient method for rapid damping [C]. In fact,several SVC installations around the world since 1985employ such controls [D].In practical SVC controls, however, large system voltageexcursions like those shown in Fig. 7 in the paper, onaccount of the SVC control action, are not allowed.Intelligent coordination among the various SVC controls,rather than a simple summation, is usually employed.The authors' comments are appreciated.

REFERENCES[A] CIGRE Working Group 31-01, "Static Shunt Devicesfor Reactive Power Control", CERE paper 31-08, 1974.[B] A. Hammad, "Applications Jf Static VAr Compensatorsin Utility Power Systems", IEEE PES Special Publication onApplication of Static VAr Systems for System DynamicPerformance, No. 87 TH0187-5-PVJR. pp. 28-35.[C] A. Hammad, "Analysis of Power System StabilityEnhancement by Static VAr Compensators", IEEE Trans. onPower Systems, Vol. PWRS-1. No. 4, Nov. 1986, pp. 222-227.[Dl A. Hammad, M. Haeusler, P. Enstedt, B. Roesle, "StaticVAr Compensators for Maximizing Power Transfer andStabilization of HV ac Transmission", IEEE/CSEE JointConference on HV Transmission Systems in China, Oct. 1987,paper No. 87 JC-17, pp. 108-114.Manuscript received August 4 , 1990.

JOHN F. HAUER (Bonneville Power Administration,Portland, Oregon): The authors have presented veryinteresting results and insights concerning the damping oflarge-scale oscillations. I have some reservations as to howbroadly these apply to more complex power systems, however.Similar work at BPA, dealing with an SV C on the Pacific ACIntertie damping transient oscillations in the western NorthAmerican power system, lead to somewhat differentconclusions [1,2]. The accompanying figures illus trate basicconcepts for sizing and tuning transient dampers that differmarkedly from those for ambient dampers. These represent alimiting condition underlying the authors' method, which isapproached as filter bandwidth tends to zero.

0

4 , , , , , , , , , , , , , I I I 'I d I I I I I I I I I I I I I I 4 .J qoo0TIME IN SECONDS

FigureA. Response to pulse disturbance200% U00

4 L

PULSE OlSlURBlNtE U: :- - z o o nc

+ , , , , I I I 1 I I I I I I I I I I 2d I I I I ! - U 0 00TIME IN SECONDS

Figure B. Transient dampingvia 39 Mvar pulse trainTransient oscillations commonly involve just one criticalmode. This suggests a time-optimal "output cancelation"strategy, in which the damper produces opposing oscillationsuntil the oscillatory term in the system output y(t) is reducedto an acceptable level (a t time t=T). The figures show this fora 5-mode linear model having a negligably damped modenear fc=0.714 Hz. The bang-bang control u(t) has a level ofU=39.27 Mvar, for which the 1s t harmonic isul( t)=Ul~in(2xf~t+0~)ith U1=4U/n=50.00 Mvar. Systemresponse using ul(t) as the control signal is not graphicallydistinguishable from tha t of Figure B.The amplitude of the canceling response will, for zerodamping, increase linearly with time. This provides initialguidelines for trades between the SVC modulation limit Uand the "quenching time" T, for assumed oscillation levels.Exact control timing requires knowledge of the oscillationfrequency fc, the phase BC of the associated modal componentyc(t) in y(t), and the phase component $(fc) of the (suitablydefined) frequency response function Ur?=I L$(f )relating yc(t) to ul(t ). Then 8u=ec-$(fc)+1800. It appearspossible to determine these on-line, using a phase-locked l00p[l] or a moving-horizon strategy drawing upon the signaldecomposition and identification methods of [31. Extensionswould be needed for multi-modal oscillations. The authorsdevelop this information implicitly, using narrow filtersdeveloped through off-line studies.The mode of concern on the western system ranges infrequency from roughly 0.75 Hz to 0.69 Hz or lower. I tinteracts strongly with another mode, at a frequency ranging


10/12

533from 0.63 Hz to at least 0.65 Hz . System dynamics changesignificantly with operating conditions, and with the criticalresource loss that triggers oscillations. This is especially trueof system response to control action, which often exhibitspathological phase characteristics. In addition to these andother complications, accuracy of the models that predicttransient oscillations is suspect.I share the authors' view tha t the key to transi ent dampinglies in bang-bang control, controller supervision, and reliablecontroller input(s). The supervision should also coordinatedamper action with open-loop remedial action schemes, andit may require some degree of self-tuning o r parametersheduling. While locally estimated angle difference has beenexamined at BPA as a controller inpu t, AC Intert iecomplexity and system variablity argue that the controllershould be provided with an ample reserve of directlymeasured dynamic information.My chief concern is with the filters. Their tuning must besha rp enough to focus contro!ler action up m the criticalmode(s) and to avoid adverse interactions with nearby modes,but broad enough to accomodate uncertainties in systemdynamics. Very sharp filters may require supervision toavoid oscillatory response to step inputs, and may produceadverse interactions. Would the authors please compare thefrequency response of G l ( s ) against t hat of the power system,for systems I11 and IV? How would they deal with thevariable, bi-modal situa tion described above? Finally, howwould they validate or refine their controller settings in thefield?

J.F. Hauer,"Reactive Power Control as a Means for Enhancedlnterarea Damping in t he Western U S . Power System--AFrequency-Domain Perspective Considering RobustnessNeeds," Appllcatlon of Static Var Systems for SystemDynamic Performance, IEEE Publication 87TH0187-5-J.F. Hauer,"Robust Damping Controls for Large PowerSystems," IEEE Control Sys tems Magazine, pp. 12-19,January 1989.J.F. Hauer,"The Use of Prony Analysis to Determine ModalContent and Equivalent Models for Measured Power SystemResponse," Elgenanalysls and Frequency DomainMethods for System Dynamlc Performance. IEEEPublication 90TH0292-3-PWR. pp. 105-115.Manuscript receive d August 4 , 1990.

PWR, pp. 79-92.

BAKER and T. KAKAR, Washington State University,Pullman WA: We agree with the use of localmeasurements to estimate the angle (and angle rate)between areas to control a SVC on an intertieconnecting the areas. We have successfully used thissame idea in the control algorithm for a phaseshifting transformer [l]. From local voltage andcurrent magnitudes, phase angles between the currentsand the voltage, and the Thevenin's impedances betweenthe node where the SVC is located and the connectedareas, the angle between the Thevenin Voltages of thetwo connected areas can be found. The authors arecertainly correct in saying there is more usefulinformation in this angle difference (and angle rate)than in local frequency. Their point is also wellmade that for small angles, the power (and power rate)comes close to conveying the same information. Forthis reason, the use of the rate of change of power asa feedback signal in SVC's has proven satisfactory formany systems where distances are very short withrespect to a wavelength.

John Hauer [2] has pointed out that there can be aproblem in multi-mode systems. We appreciate the factthat the authors have used a three machine systemwhich does have two modes. We would suggest anadditional three machine topology be considered(System V) and the control algorithm be tested for twoequilibrium conditions. We suggest that the algorithmwill enhance stability of both modes for one of thecases, and detract from stability of one mode for theother case. System V is shown in figure 1. LetX I = 0.5 xz = 1.0x4 = 1.03 = 0.5Hi = Hz H3 = 00In the fi rst example (System V-a), let the injectedpowers be such that the equilibrium values of themachine angles are

and in the second example (System V-b)Machine 3 is coupled tighter to machine 1 than it isto machine 2. Therefore the angle of the Theveninvoltage of machines 1 an 2 combined will be closer tothe angle of machine 1 than it is to machine 2. Thenin case V-a, if machines 1 and 2 swing apart, theangle between machine 3 and the Thevenin voltage ofthe other two machines will increase, causing the SVCto raise the voltage. This will enhance the powerflow between machines 1 and 2, which is exactly whatwe want it to do. However, in System V-b, if machines1 and 2 separate, the angle between machine 3 and theThevenin voltage decreases, causing the SVC to dropthe voltage which will decrease the power flow betweenmachines 1 and 2, letting them swing apart faster.

61 = - 30" 8 2 = -10" 83 = 081 = - 10" 82 = -30" 83 = 0

G3; G 2

I System VFigure 1The authors should be commended for their work indeveloping this algorithm. We feel that the problemcited can be mitigated.

REFERENCES[l ] R. Baker, G. Guth, W. Egli and P. Eglin, "ControlAlgorithm for a Static Phase Shifting Transformer toEnhance Transient and Dynamic Stability of Large PowerSystems," IEEE Transactions on Power ApuaratusSystems, Vol. PAS-101, No . 9, pp. 3532-3542, Sept.1982[2] J . F. Hauer, "Reactive Power Control as a Meansfor Enhanced Inter-area Damping in the the WesternU.S. Power System--A Frequency Domain PerspectiveConsidering Robustness Needs," in Applications ofStatic Var S stems for Svstem D namic Performance,IEEE ub 87TYHO187-5-PWR. pp. 79-;2, 1987Manuscript received August 4 , 1990.

G . ANDERSON and T. SMED, Dep. of Electric Power Systems,Royal Institute of Technology, Stockholm, Sweden.The authors are congratulated o n presenting acomprehensive analysis of an important issue in a wellwritten paper. Their physical insight in the discussedphenomena ha s guided their analysis which we believe isthe correct and most powerful approach.


11/12

534We appreciate the authors' comment on the use of bus-voltage as an input for damping purposes based on thefol lowing reasoning:The damping act ion is achieved by modulat ing thereactive power output from the SVC in an appropriatephase with regard to the power oscillations, and the tasko f t h e co nt ro l ler i s t here f o re t o de t erm ine t h i s(appropriate) phase from the chosen input signal. Ourexperience indicates that the derivative of the bus voltagemagnitude is a reliable indicator of the appropriate phaseof the reactive power modulation for damping purposes.This observation will explain the instable behavior shownwith pure V-control in figure 6 of the paper, when a PI-controller is implemented.Manuscript received August 17, 1990.

We would like to thank a ll the discussers for their interest nthis paper and their comments and questions.A. Hammad and M. HaeuslerIn Fi . 5 of our paper we have shown the increase of systemoscilfat ion dam ing by increasing the SVC-rating. However,the increase or the SVC-rating leads to a larger reactivepower oscil lation and voltage oscillation. In Tab. 1 we haveshown how to combine the damping of active power oscilla-tion and the limitation of voltage deviation during poweroscillation by using different target functions to imit the re-active power oscillation.We agree with the discussers that an intelligent coordina-tion of various SVC control demands is necessary. We wantedto show the principle reaction dependin on new localmeasured phase angle control. Using intejigent control itwill be possible to use different phase angle signals to adaptthe SVCdamping to the different modes of oscillation in amulti-area system. But the main target should be, to makethe control as robust as possible to increase system security.In our example the system is operating close to (and behind)the system stabi lity limits (calculated wi thou t SVC). Thetransient overvoltage appears for some seconds. The ques-tion is t o access whether the system instabili ty or the systemovervoltage is more dangerous for the system.R. Baker and T. KakarThe discussers presented a challenging question on dampingof oscillations in multi-mode system, which we have alsorealized a t the beginning of our work. It has to be admittedthat in principle, the damping effect of single input TimeOptimal Control to multi-mode oscillation s limited, regard-less of whether i t s being SVC control or being any other con-trol like braking resistor, excitation and governor control.Sometimes it may even be uncontrollable regarding o speci-fic systems. To improve his new control strategy, multi -inputTime Optimal Control hasst i l l to be developed, which i s justour next work. In terms of the example presented by Mr. R.Baker and T. Kakar there do exist the problem of getting thesignal correctly reflecting the oscillat ion between G1 and G2 .But in case of that network structure we would rather locatethe SVC to the connecting bus of XI, X2 and X3 to havebetter observabilityof the whole system oscillation modes.

J. F. HauerThe discusser describes a complicated bi-modal oscillationsituation where the swing frequencies depend on systemconditions. In such a situation the control concept can onlybe adaptive. In addition, tunin of the filters to the criticalmode is complicated because oathe modal interaction. Digi-t a l filters in combination with intelligent supervision to re-

duce the risk of adverse interaction can help to overcomethe problem.In Fig. 10of our paper we have given the results for four sys-tem configurations. We control the phase an le differencedepending on the optimization procedure. On?y in system IVwe find need for larger adaptation of parameters to the sys-tem configuration. The system is operating well also, whenopening line 2 3 after the fault, as shown in Fi . 1. Thisshows, that the design of the parameters S robust ayso in thecase of system parameter changes, but in our example thereis no strong interaction between the swing modes.

PZ-1 = 550 MWII 300 km 1'

____...____.___300" -.--3 phase short circuit at line 2-3cleared by opening line 2-3

SVC in operation

S2., exactly calculated -2., estimated at SVC-node - - - - - - -"' V U " without SVC

Fig. 1Change in topology after the three phase short circuit

To answer the question of controller setting in the field: Thefirst step of system investigation s calculation of eigenvaluesto find the modes of oscillation. In addition a digital modelof the system is created to identif a (reduced) equivalentd namic system structure by matiematical identification.T i i s equivalent model will be used in a hardware system s i-mulator also to test the control of the real SVC-controllerand adapt, if necessary, the control behaviour. Finally, in thefield we make tests limited by the requirements of thesystem operation. It will be also necessary to observe thecontrol behaviour over a longer period to prove the correctcontrol performance.

G.Anderson and T. SmedAsshown in Fig. 6 f the paper the change of voltage ampli-tude cannot be used for power osci llation damping. Thedamping information can be found using the derivative ofthe voltage phase angle. This signal can be interpreted aslocal frequency change. Coupling a relative small systemwith a large one shows transient requency changes withoutoffset from the basic frequency in case of power systemoscillation.A coupling of two systems as in the paper shown in Fi . 2shows a local frequency change depicted in Fig. 4. This %e-quency change was calculated by using the derivative of the


12/12

535voltage phase angle at the SVC-node. The mode of inter-system oscillation must be filtered out o f this signal, becauseboth systems are dri ft ing slowly from the basic frequency(zero-line). Therefore at least a 6th order fi lte r has to beused to find the swing mode of the system.For these reasons the use of derivative o f voltage phaseangle can be complicated.The derivative of the bus voltage magnitude can also beused as damping signal. But in Fig. 2 we have shown thechange i n voltage at SVC-node, the change in power flowbetween the two areas and the dVsvcIdt-values for a powertransportation of 785 MW and 850 MW to area two. It canbe depicted that increasing the flow between the areasresults in saddle-points i n the voltage characteristic. There-fore the dVsvc/dt-measurements can produce error-signals.Designing a bang-bang control will increase the problemsbecause of additional harmonics especially near the vicinityof stability limits.

PI-, = 785 Mw

,

Fig.2 Derivativeof SVC-node voltage for di fferent load flowsituations Manuscript received October 2 6 , 1990.

svc control

Documents