You are on page 1of 9
TERE Transactions on Power Apparatus and Systems, Vol. PAS-99,No.2 March/Apeit 1980 10 EVALUATION AND DIPHOVINENT OP ELECTICNECHOITOAL OSCILLATION DAMPING BY MEANS OF EIGHVVALUB-ETGENVEOIOR ANALYSIS. PRACTLGAL RESULTS IN THE CEVDRAL FEAU POWSR SYSTEM Velroidiacono, B.Perrari, Rallarconato ‘BIEL, DSRCRA (Italian Electricity Board, Research and Developnent Dept.-futouatica Research Center) Nidan (Italy) Austrict - To eliminate oscillstory instability in electric power eystene by inserting stabilisers, it 4a inportant to have available analytical criteria for ‘the choice of the most euitable generators to which to put the stabilising feedhacke, and for defining the fnost convenient parameters for then, without resorting to aimulation, Applying sone properties of contrel oy etens it is possible to eet up digital prograns bas ef on calculations of eigenvalues and eigenvectors ‘that solve these problent satisfactorily. The report ‘then shows application of the procedures thus defined to a practical cage of instatility, and field test re sults that confim the elimination of instability. ‘persopuerron 141) Scope Oscillstory instability in multi-nachine systons as been the subject of mmerous reports an studies. The results of a recent enquiry by Clere, which i now being published [1] , confim that saprovenent in damp ing ie ueually obtained ty introducing adstional feod backs in the excitation control of suitable generators There aleo appeare to be a groving use of techniques tased on the calculation of eigenvalues for the analy- sis of systen etability. Much less popular is the u fof algoritmae thet make 1% possible to pick out the gp neratore nost suitable for the application of tho sta= bilizing feedbacks, and to calculate, with fair accurs cy; the increase in damping of the’ various critical cocillation nodes obtainable with auch feedbacks, vith out resorting to eimilatfon, except when it in necesea ry to check aysten behaviour in the face of major aie furtences. The basic Lines to be followed in possible determination of the sensitivity of the danping to the fftecte of speed goverrore and voltage regulators, With particular regard to low2requeney oscillations dni to particular structures of netrorks, have already been deveribed in recent studies [2,3,4] + Tis paper taking up these concepteygives a more general and crea rio description of then.the remulta are elven of the 779 649.5 __® paper recomended and approved by the TSEE Power System Engineering Comittee of the IEEE Power Engineering Society for presentation at the IEEE PES Sumer Meeting, Vancower, British Columbia, Canada, July 15-20, 1979,Manuscript submitted January 51, 1979; made available for printing April 4, 1979, 3s Dos Ghali Dscrandee Boole Polyteoinique Péd. de Lausanne Blectroperu Laborat. Mach. Electr. Gerencia de Bxplotacion Lausanne (Sxitzeriand) Lams (Peru) caloulation procedures applied to a practical case of oscillatory inetability observed in a power eystem in Peni, reporting aleo sone of the results of the field tests with which it was ascerteined that the stabilis ere inserted make it possible fully to elininate any op fermtional restrictions due to instability. 442) Digficalties in the Auslysie and Synthesis of Denping by Simulation Methods Ae ie now know, oscillatory instability (for enall varistione) {e almost entirely due to the effects of speed-govermore and voltage-regulators: to be more preciae, if, ideally, we were to eliminate speed gover hors ani voltage regilatore in an electric power not= ‘work, we should not, except in very rare cases, have Sscillation modes with negative damping} euch inctabild ‘ty occure especially in particular network structures (areas comected with woak tic-Lines; eystens with lon gitudinal structures generatore joined by long lines to the gest of the eysten), in which the stabilizing ef fects due to the structure of the generators (damping windings, solid rotor) becone less important, while the effect of the speed-govenore ant voltage-regulators,es pecially with ragami to loversfrequency oscillations, ‘beoomee mare 20. Consequently, the transfer function of the regulators has, in such cases, ® fundamental influm ence on danping, and investigations on the effects of nodifications in the calibration of regulators, or in the structure of excitation contrel (mainly, the intro~ duction of stabilizing feedbacks) represent the main ‘chapters in studies on the improvenent of damping. Tt is therefore at once evident that much tine can bo saved if we have at our dispogsl means of analysis ‘that indicate which regulators say havo the greatest effect on danping, beth when it cones to choosing the ones in which to put the stabilizing feedbacks, and in concentrating on sone of then greater attention in de fining the model. Bot apert fron thie obvious considerstion, Ainect calailation of the danping values, or of their in creases provided by suitable additional feedbacks, is much easier than calowlation by simulation, for the fol, lowing main reasons: i) Guantitetive detemination of damping from the faving curves supplied by simulation may be diffi= uit when, in the response that is examined, the conteibution of various ogoillation aodes is impor tent. ‘The damping values observed in simulation may be considerably influenced ty mumerical integration ‘teokniques. iii) Also bearing in sini the previous point, which may 4s) ‘0018-9510/80/0300-0769S00.75© 1980 IEEE. involve using a particularly small integration step, similations often call for larger calcule- Hon times than those required ty programe based on the nodal synthesis ond analysis procedures ‘that will be dealt with further on in thie papers 2) PROCEDURE FOR THE ANALYSTS AND IMPROVE OF aD ‘DAMPING OP ELBOTROWECHANICAL OSCILLATIONS 2.1) Stabinity Analyste 4 initiel ampect of the procedure in question cop cerns checking the characteristicn and damping condi= Hons of the electronechanieal oscillations in an elec trie poser aysten before any steps are taken to inprove stability ; it may te necessary to do this during the Planning stege, empectally if the stricture of the eye, ten ives reason to mumpect that there may be a ride of oecillatory instability, or else may be aivieabie if ip eeiticiently danped ostitiations have boon noted during operation, in order to make comparison with the expe Finental Teoults,and to extend the examination to other possible netwre situations. let us asmune we have tho model of an electric pow er systen, consisting of 1 machines, linesrized around the operation point conremponding’ to a condition of equilibrium, weitten in the fom [5] Ais Aacteau a) im which x ie the vector of the atate variables of the aysten, which are ~ the angular positions snd tho angular velocities of the rotors of the machines ~ the electromagnetic state variables of machines /1/ ~ the stato varisbles relating to the excitation com trol systens ~ the state variables relating to the spoodepvermor syetens (including supply aystene) wile y is the vector of the inputs, including the vol talowregulator and speed-governor reference: ‘The eigenvaluec A) of the matrix A supply the im formation regarding stability of the aysten for auall Varistions : in particular , the characteristice of the 44 electronechanical_ oscillation modes (damping &y and frequency Wn =Gpn joan be obtained fron the com Feoponding pairs of eigenvalues, written in the fore fo Sn 0n aioe B10, eter (2 Iéentifiontion of the eigenvalues Ananong all the Avmay be effected ty meas of the eigenvec. tore of matrix A by looking for those whose elenente have a relatively large nodule corremonding with the Clectronechanical state variables (angles and angular velocities). Very often, hovever, the low values of the danpings for the nodes corremponding to the electron chanical oscil1ations (especially for the "eritical™ ag den) are already sufficiently indicative for thie idee: tification. /h[ Moat is, for a 5th order model, the fluxes Linked, Tespectively, with the field circuit, the direct axis damper winding, snd the quadratureanie damp windings or the transient and eubtransiont e.m.f.4¢ of the quadrature axis, and the sub-tpansient 2.2) Inprovonont of Damping As ie known , if there exist cloctronechanical mou deo sith nogative ory in ary cage, incufficiont damping (en8+s6<0.03) , the’ best method of obtaining an ineret Yonent is that of introducing into the excitation tom ‘trol of suitably chosen generators sone local addition a feedbacks, connected with the speed of the generates and of ite derivatives. let us know how to select + = the generators to which applying the stebilisers = the transfer function of the additional feedbacks. Tho purpone of the stabilisation is to shift tho eigenvalues corresponding to the critical modes so ae to bring about sone increase in damping. A mall shire fs acoeptatle~ a a$ value of 0.1 in alzeady cation factory = and yet the gain. of the additional feedbacka stould not be too high, to avold excessive disturbances in voltage regulation, as well as for other reasons. Re course may therefore le made to. the teciniaie of seal shifte of the poles ty neana of additional feedtactss Ify in Pig. 1 G(s) representa the trancfer Anon tion between to quantities u(s) , y(e) , between \hioh it is desired to close a feedback’ i(s) = cm («) Ce being scalar and n(s) of given structure) aseuing that all the seroee and poles of G(s) and H(s) are ai tinct, any pole of (x) being indicated by Ay and the corresponting residu with c!then it can be show (ave e+8. [6] ) , based on the knoim properties of the ini Hal derivative of the roct loousy thaty for Mmrall""e Yalues , the clomurd of the feedback causes a shift in ‘the pole A (see ales [4] ) 1 Hea @) Taie property may te sppliod to the hy poles seg ciated with the electronechanscal oscillations with rer duced damping. HO, ae H hie Ret. Tt showla be lets at all points of the cirewnference , hose contre is at An and whose radius is]4y| , there corresponds the same gain value |¥(An)| of the’ additional feedback, For a given value of [H(An)) among all she Posatble shifts Afn y the one to which the maximus damping inorease oprremonds, is obviously the shift tn the direction jay that ip with Wan + comnts Insertion of a featback with a mall grin in a gysten noted thet, since Una S.m.f.0f tho dizect axie, Given the mall damping value of the electronechanieal oscillations, ve have Aya j@y » and thue # [oA9]eogn const, => OR AR The maxisun effect on the danping » assuming the sane harmonic transfer function gain of the alditional feed= back ie therefore obtained Wy selecting Sh ta) Lessons x ~ Loo) () land, in addition wo have ‘thet ie 0°) woven aon ©) This result may be applied to the case of the alditionm sl feedback derived fron a quantity connected with tho angular oscillations (velocity, electric power) and im serted in the excitation contre! of a suitable gonars tor 1 we will eaoure, for example, that + Yea) = BA), 0nd (8) = AvrKCS)0 in which Micp and Vek are the angular velocity ant the reference of the vala tage-regulator of the IO generatorsIf the resicu OU) te know, expressions (4), (5), (6) aake it possible to choose the nechines on witch The plamed feedback 18 ost effective, and to define the value of The feedback tranafer function corremonding #0 the coctliation fre- cuesay) in tenetin; ta pWinin aigiven dmg ing iecrenwe (or to evaluate this incresse for a given additional feotbacie exin). Assuming that equation (5) ie estiefied , the (6) makes St possible to define the sensitivities of the damp fing as compared sath sho gain [Hy (1y)] of the Lom cal feodbaoke on the various nackine y which, fer each machine ety, of the network "and for each mode Metyevey Dt are expressed ty ¢ (7 ‘he colentation of the resid cl? necessary for the nensitivity detereination can be done with Me rey One in which yy is the elenent of the eigenvector ry (a sociated with the eigenvalue Ay) relating to the state varisble 84, ond Qh, the elenont of the matrixas™” B relating to the mode hand to the input avjq + To krow ‘the residus we therefore have to calculate the eigen values of matrix A, the direct matrix 1 fant the ine /2[ After having chosen the generator where the eed Yack shall be applied, the residu corresponding to ‘the more sophisticsted model may be evaluated for every pole fron the pole ehift due to the feedback boy means of eigenvalue caleulations + see par. 3.3. Ai tastead of recording the velocity directly , it may ‘ve more cunvenient , experinentally, to observe the oscillation of the electric pow ah. ‘thie ie m verse matrix 171 of the eigenvectors of Ay and the ma trix 8. The cnount of calcalation depents to 2 great ex tent on the order of the aystent thie ie a first reason wy St nay be of interest, purely in order to determine ‘the sensitivities, to use'a simplified model of the gm neratorsy as explained in point 2.3 /2/ . Yoreover, it is very interesting to note thet determination of the residus cl) can be effected experinentally, Wy obser ing the retponse, in our case, of the velocity /3/ ay to an AVR reference step . Tals reponse (see Fig. 2) via detemination of the Fesidu ant of the oscilistion frequency supplies all the information necessary for determining toth the seneitivity of the damping to a ai vvon additional feedback applied in the rachine on which the experinent is carried out y ani the value required for the transfer function of the feedback Hx(s) core mpondinely to sein. Since the values {c)| are propor ‘ional to the oscfliation amplitudes, the most effects vo gonsratore for danping purposes are those where, ci- ven the sane step Av; 4 we observe the greatest speed oscitlation amplitudes /4/ « ro Figs 2 Reeidue evaluation fron a step remponse 2.3) Reprnmtation of uu tyeim with a 2ubortor te a The nnaitivitien 10), cateutetd with equation (1) et anterniont esurinenaliy wih the aforenntia be focatens = frtbiay cles buat ot abe neat iste or the voltngeegutee fo mctane™ nay tere™ ture of eneah affect ieee. tery Seomn as fo ‘Stgpcngution top has tn iow a sutoft Prequney Oy iM Toation fo te setlletion freweay ty nore particularly aivisable whon the feedback is de, rived, either wholly or in part., from the electric OWED Py /4] Uy in a power station, there are several unitein parallel , the stop onthe reference should be ap plied simultaneously to all the generator, also in relation to the fact that the stabilizing feedback should be fitted on all the unite st the power at Mion. m Given the relatively low cost of sodemizing the exoitation control eyeten to make 4% sufficiently fort, 4% say be of interest, sn determining the sensitivities YQ), to Leave aside the AVR characteristion; this is @ futther reason that justifies caloulating the eenaitivi ‘Hes ‘ith a simplified model of the machines, nanel; ‘he well Known 2ni-onder model in whitch they are repre~ sented ty an e.s.f. 0, , behind suitable reactance (esgey the direct transient reactance). Mith this model, 48 Tt ie asmimed that oi the control variable, by eB lying equations (4), (5), (6) and (7), we obtain the Senaitivity ond characterietice of the tranafer Aino “Hon of an additional feedback, on the assumption that the generator Je equipped with an "ideal" excitation With the aforementioned model, the equation (1) ‘takes on the following fom (see Pig. 3): bea] having puts cewek, ABM was {0}; Ke (Rus ig a BP Figs 3. Insertion of o stabilizing feedback in 2 power systen: matricial block diagran (gonoratore 1 presented with 2 order model). {5/ The notation ( )* indicates values calalated at ihe stecdj-nete point rou oeieh the aguten aw teen Limeastsede /6/ Bquation (10) may be written F © 9 ar raat He? ( sd the coatftctents 4? are oxpressed by 0) « Sant“ HT aR By indicating uitn Rak the additional feedback between the owtmut Band the control dey 1 can eatim ay be show, given the epecial etmuctare of the dymanio tmatrix A relating to equation (8), that application of squation (4) leeds to the purely algeoraic tranefex inetions 205 46, Bien Prk @) sn Which Gjq sre clanents veotore of € ani 1f the matrix G of the eigen Poy are cleonte of the satrix Korvover the frequency of the ht? electromechanical ogeiL story node 42 given by Ve im which Ly ie an eigenvalue of matrix ¢ Tt jo Obvious that there has been a drastic reduc tion in the volune of calculation, taking into account the fact that it ie eufficient to know the eigenvalues fant eigenvectors of the matrices C and the matrices Gt and A, that are matrices of an onler equal to the mun ber of machines present in tho network. Bron (9), we find that, if as uaually happens (see rote /6/) Cyn Prk <0 y 2 given increase in damping of node | with the minimin feedback gain value Rays 18 ob, tained by modulating the internal eam.f. ey in phase ith the respective velocity Oy. If, a2 happens in par tioular in lowfrequency oseilistions, the phase of 6} in the oecillation® can be confused with that of the ma chine voltage vk then this result corrwsponde to the own practical mle based on thet phase agresaaxt to "adjuet" the additional feedback transfer function dur ing field teste. In this onder of approxitation, it may be said ‘that the tranefer flmetion of the additional feodback must “compensate”, at the froquengy or frequencies to bbe danped, for the phase lag (a9 well aa, if necessary, for the attemiation) corresponding to tho closed-loop transfer function of the AVR: this compensation ie sore ‘casily effected for a lange frequency bantwidth of the oscillations W, if the cut-off frequency w, of the main voltage-contrel loop is sufficiently high’, as oocurs particularly in the cage of static exelters. Moreover, uch compensation in many cages influences the choice of the variable fzon which to derive the fecdback — by selecting, for example, @ feedback fron tho electric po wor at the place, oF in addition, of a feedback from ‘the velocity or local frequency. Fros equation (9), we deduce the sensitivities of ‘ihe danping ae regards additions] fesdbask applied in ‘the variow: sachines, assuzing the latter to be equip ped with "ideal" excitation controle in the sense shown here! Oe eo ec (10) fof Tees aise 12,3},)6 The values of Wf") are influmosd by the loanflow expecially dus 0 the effect of the factors (9Pakf ite shen howovor, generally renains positive. | For further discussion of the effect of the network structure, the load flow, and tho nature of the Toads, the reader is referred to [4] « Should the loads be Linesr, by adtittances, having resort to the adnittance matrix elated to a, the sensitivities 70) may be written [a] as ine ou'of tuo tema, the first of which corre Spode mainly to the effect” of the modulation in the active power absorbed y the loads, ani the seco! mais Ly by the effect of the nodilation of the power flowes Yery often, little is tow of the nature of the louie, s0 that it may be useful to separate euch contributions in order to ake a qualitative assesenent of how far the load characteristics nay affect the damping calcula ted; trom this point of view, obviously, the eesumption of Linear loads is generally optinietio (see alao [4] o that ie representable Bvalustion of the effect of speedngovemors may al so be made if a linear model /7/ is adopted for the go- vomore , cousidering each govemor al a local foedberie O,() =~" a (8/4 n\6) between the speed ny and the ne chanieal porer Png of cash generator, and spplying agus son (3) with Hyt}p)=—O4l) Op) » Generally, since the corresponding reeidu" is ofly slightly affectet ty the seneretor model, st 18 sufficient, fon purposes of thie tranination , 12 consider a scconfconier sedel for the generstore. With thie moder, yo find = eee also [4] that the dang ing variation A{ cue to the speod-godermore 40 expres sed Wy! ees. al pect? ache Zul Ded i) saving ot sed ene[0,s04 ]e[0,0.,] eos of 8 / oop) woite the ooettictentg yf) are defines im the rote Zo/~ Since, unaiiy 1 +0 /8/, we agin tind the com Sition t n'a't the contestation so danping by e spect Sgoremor ie negative if =909>04>~270" a8 nay hapoery wit vais of B(J@p) that are" sot eeally in the came Of hytraulic turtiner? due to the water inertia effects: on(T4)ekowe that knowledge of the sonestiviti t} way serve to provide indications on the inportanc c's nove or lose careful. modelling. of the. tramafer function D4(s) in onder to cblain a correct eval of the camping. 3) APPLICATION OF THE PROCEDURES ILLUSTRATED 10 4 REAL case OF DISTABILITY 31) A.brief description of the Central Peru systen and of the operations! difficulties encountered An example of the application of the analytical procedures described in the previous section is given dy the study , made in July 1977 y for the purpose of elininating the oscillatory instability observed in the Central Peru power systan , run by the Electzopem Coxe pany. Aif wor an assesenent of tho posible efvects of the nonlinearity due to spoed-governor tackleah, 268 ta. A wnifilar diagram of the eysten ie given in Pied, The generating power stations are all hydraulic ; the biggest is the ono at Mantare , where nowadays three 1 nite, each of 120 MYA, are installed . Theee are equip pel with rotating oxcitere, and the turbines are Pelton with clectrohytraulic speed-govermore . The power ata tion will be developed ty the inetallation of a further 4 units , omippad with etatic exciters. Tt is intended that 6 units shall be able to operate eimltaneouely at full load. The Mantaro power station te connected, across the Andes , with two 220-kV lines , which are (or will te) double cireuit, to the rest of the ayeten , distributed along the Pacific coast . The triangle formed by the Tn @ipendencia-Mantare-San Juan Lines oan be opened at Man ‘taro, where there are two bushar ayeten that can be separated. arian, © a190 INOCPENDENCIA 1220) Figs dy The Central Pema power ayetem Oscillatory instability was obeerved in the systen, with oscillations that were particularly vieible at Man, ‘taro « Tt vas possible to eliminate the instability by reducing the pover generated at Mantaro, and the inata bility phenomens could be lessened by reducing the gain in the Mantaro excitation control system . On the other hand, it was desired to keep up a high response veloci= ty in the Yantaro voltage-rogulators , particularly in consideration of possible ewitching-off at the far end of the Lines starting from Mantaro. The introduction in ‘the Mantaro excitation control systen of an aiditional feeiback fron the electric pover had brought about some Anprovenent; however , since there were certain Limitee ‘tions in the tranefer motion of that feedback, it had not been possible to obtain stable operation in ‘all the configuretions as planned , with the units operating at full. powers (/9/ Tis is certainly true if the matrix K is symmetry eal[2}; and also in the mot part of real cases, ei von the fact that matrix « is not far from eyanetzy, Tt had also seemed that the instability phenosena oteerved wore influonced by tho static excitation eye ten of the synchronous compensator at San Juan. Sone inprovonent in stability had been observed when the dynanic gain of the Mantaro spesd~govemors had been reduced. In addition to complete and reliable elimination of the operating restrictions under prevailing circu stancat, in all the configurations, it was also desined to make sure that stability could be quarenteed in the course of the future aforementioned development of the systen. In dune 1977, con duct a etudy ained at discovering the cause of the in stability and defining the characteristics of the aost economical devices for ite climination, bearing in sind various present or future configurations under various loading conditions. Por thie study, advantage was taken of the experience aoquized in this field at BIEL's Auto natica Research Center, which algo had digital prograns 4o perform the procedures described in Section 2 abovey which had alzesdy been weed with suecese in other appli cations, The results of the stuly, the subscquent ac ‘ion taken, and the results were diecuseet and defined bby @ technical connittee, on which experte representing ‘the user, the manufacturers, and their roepective com eultaitte served. The data on the network, machines, and regulators were supplied ty Electroperu. In checking on stability, it wae seouned that loads depondod on voltae g in the following mamer: ap fare ay /vi aa, /oi, nas ane Asruning thet the contritution to overall inertia ue to the loads were oqual to 20% of the inertia of ‘the generatore, approximate account wae taken of this by suitably increaring the inertia coefficients of the eenersitre nearest to the loads, At waa therefore decided to 342) Homulte of the analysie Six network situations were exanined (four with 2 generators, one with 3, aml ono with 6, at Mantaro: in the latter, the Mantaro buebars were comected , and in addition the San JuanMantaro Line consisted of '& dou ble circuit). ‘The analysis was conducted by repeated application of two digital prograne . The firet (SE0) , starting from losd-flow and assuning a second-order model for ‘the machines (in accomlance with has been said unter point 2.3) and a linear model for the loads, calculate for each oscillation mode , frequency ant amplitude of oscillation: at the various power stations , and the Genping sensitivity of the various modes in respect of stabiliaing feedtacks (inserted in "ideal" excitation controle) and applied one st a time at cach power stse tion , giving separately the contribution to damping mainly due to the loads (oo Table 1). The progran also supplies the sensitivity values [yy » aed therefore , /s/ The higher frequencies correspond to modes in which ainly the three synchronous compensators are invol, ved , ae shom by the approximate distribution of the anplitutes ise | ter on om LE sme tne’ oomrcnes spree we cease ‘TABLE I - Approxinete anplitude aistritutions and damp ing contitivities of the different power sta, ‘tions, for the oritical modes. sakes it possible to evaluate the influence of the speciegoverors by applying (14) + The second progran (Wit) provides sainly the opotLlation and damping Frequencies of the various nodes , by weing en adequate representation of generators (5th omer sodel), loads , speed and voltage-regulatore, stabilising feedbacks if nyt and therefore, in particular, ty scans of compari fon, nakes it possible to check the effect of the feed ‘mola, taking into acorunt the teal characteristics of the voltage-regulators. For the sake of Brevity, only © for of the quantitative rerults obtained are glven here Table 1 refere to two network eituations with to unite at Hantaro, with separate and comected busbars With separate burbare, the two lovest-froquency oscil lation nodes have negative or alight damping; in then y the two Hontaro generators oscillate in phase opposi= ‘Hon and in phase , respectively . With comected tuom tere , there de only one critical ode , in which the two generators obviously oscillate in phase with cach cther ~ tho axplitudes are slightly different, due to Gifterence in the respective operetion points ssomed for the generators thenselves, For the other oscillation modes , which are mostly lool in character , the frequencies foun were between 4,2, and 1.98 /9/,nith good or acceptable damping /10/, Table 1 shove that the Nentaro unite are far the most convenient ones for insertion of the stabilising Toodtacke j a2 can also be seen , fron the network dior Gran y the danping thus obtained ie little affected by the Loade. /H0/t showld be noted thet the spood ard voltage-regu- Latore of the unite located in the other power stan ‘ions have @ relatively low cut-off frequency. Similar conclusions have been reached in the othe cer network conditions exanined , with the observation ‘tha the strengthening of the network due to the dou Be cireuit betwen Naxtere snd San Juan , as well as ‘the interconnection of the busbars at Mantaro, produc 8 a much less uneven distribution of the oseillation amplitudes anong the various generatore . Moreover, it As reached that , in the G-generators situation — in Which, for the slowest mode, the values 1}=0.99 ace and [0.007 yyithout stabiisters, were found - the Noro panpa and Ketucana power stations would aleo be suite ble for dumping inprovenent; in thie case, the contri ution due to loads would be mich greater . Neverthe less , at these powr ctations , it wuld ke necessary to modify the excitation control eysten, All in allyal so taking into account the type of excitation control ‘tho Mantaro power station was found to be the most aul table in all the situations exanined , for the damping of the critical node The possible effect of the epeed-governare at Man taro ~ the effect of the others being entirely negligi ble = vas evaluated ty applying tho equstion (14); ai ve the osoillation frequency is beyond the resonance frequency of the penatock , and St ie therefore aiffi- cult to determine exactly the amplitude and phase of D (Jwp) as regards contribution madé ty the penstock, pestinistic asemptions have been made on this point } even sy a maxim negative contribution (mith separa te tusbars) was found of AC =~0.045 with initial eet ting of the regulators, and A = —0.015 with the exist ing setting; the conclusion drawn wae that, if conside red appropriate for other reasons , after introduction of the stabilizers the governor setting could be put ack to the initial values. As rogards the possible effect on stability of the AVR of the gynchronous compensator st San Juan, the analysis, which was confimed by subeequent experiment al teste, indicated that this wae very alight. 3.3) Laprovenents in Stability with the Stabilizing Foodbacke at Kararo In the study in question, regerence was made to a systen of stabilising feedbacks of a type that, in ax cordance vith a project once developed at the Autonati ca Research Centerjhas been used with muccese in other cases [3] yand that was aftervards used effectively for Stebilization of the power systen under consideration, /11/ Xn practice, the velocity ie detemined by meamr ing the signs) frequency of a tachinetric genera tor, and oan be replaced by the frequency of the AG nachine voltage;the particular type of frequen cy transducer used dos not introduce any apprecia Ble lag. /12/ Prom the point of view of the transfer function, ‘there are no substantial aifference between thie solution using the structure + set fold nc eB [rene] Tey (ia soceleration constant ; Tw = washout tine constart = 3.5 200 5 j= rated powr j Og> aynohra uous speed j Ke, Ras Me expressed in put) ms According to this design, the feedback consists of two channelsy the first being derived with gain Ke fron the electric! power (neasured according to the Aron method), land the second derived with gain Kq from the velocity 711/ 4 settable weighting in the mun of the tw sige gals generally makes it possible, with modern types of voltage regulators, to compensate for the delay due to the voltageregulating loop within a faisly wide range of oscillation frequencies , this achieving with fair accuracy the phase relationship required by eg. (5) /12/ «A rough ovalustion of the most suitable va Tues of Ke and Ka fer compensating for the lag in the voltage loop can be made on the basis of the characte Fistics of the dominant electronagnetic mode of thie, Joop. A closer approzination is obtained ty evaluating the sensitivities of the shifts in the electronechani= cal pole towards feedbacks from the velocity and fron the electric power. Fig. 5 indicates the shifts in the poles corre sponding to the critical modes in the situations in Ta Be 1, for growing feedback gain values , calculated With the MARSTA progran . Aesuning we have” = feedback of the type neredtiie ae [hot ser] ‘by applying eq. (3) , from the shifts of the poles due to mall values of the gnine Ke ant Kg (outs Ke= 0005; Miq v0 ) 4 can Ghdnce tin associated residue valane, and subsequently , ty applying eq.(4) we determine the most suitable value of the ratio Ke/Kqya value for Suen there ie ‘a tendency of the pote to ehife in the direction indicated in point 2.2). The damping valuen for the critical nodes provide a ty the RST progean without stabilizers aad with stabiliser sith a gain of Ke= O18 and Kg 10 are show in Fig. 5. "canbe seen that , vith the ratio chosen , the stabiliser dove not appreciably alter the oactlletion Frequencies , an desired « We can ales ave Yat s Miah asin values “that have been kept acceptable from the Point of view of the isturhances intreduoed into the weltage-cegilation ; dampiag of the aritiodl acdes i Trought wp to satisfactory valuce. Using these settings, sinilariy sstiatoctory results wore alan obtained in 1 the other netwrk coniitions euauinedy moreover;ea tegutive eftect wae Soeerved on the ether ovotllstion odes, soe damping ia Tittle affectea by the stabilia the slgnate et Mntare, fs conpared with that one using a feedback deriy ea fron the electric power snd a transfer function ‘that is usually of the following type + Bee ih ane Ja] gs Ot Ge TE [teone Meo ge Sew Nevertheless, fron the point of view of aisturban ces introduced in the voltage regulation , it ape peared to be better to use the first solution, en it is desired to achieve efficient operation even at very low oscillation frequencies (around 0.15 Hz), and it is therefore necessary to exploy high values of Ty and; ). 76 Figs 5+ Shift of critical poles due to additional foo decks - Generators represented with a 5! om der aodel = Worelinear loads =Other conditions ae in Table 1. fa) Separated Mantaro busbars ) Gomected Mantaro tusbare 3e4) Experimental reoulte Given the results of the study, <4 was decided, in Sept. 1977, to install, at the Mantaro power station , stabilizing devices that had none of the Limitations of ‘those previously inserted . The particularly wide range of effectiveness of devices constructed in accordance with the pattem described in the foregoing made it de- sirsble to use thon , bearing in mind the possible pre ‘aence of lowenfrequsncy oscillation modes in future ex, tensions of the network and of the intercomection in ‘the Peruvian electric power syston. In Pebruary 1978 , when the installation work was completed , a vast prognante of tests was carriod out conieting mainly ina series of transients , each of which wae repeated twice, with and without stabilizers, in order to evaluate the latter's effectiveness . This ‘effectivencss tumed out to be high in all the network situations tested. Final setting of the gains vas detemined in the field (depeming on the actual characteristics of the voltagenregulstor, and including network situations not examined in the etudy) at the values Ke= 0.25, th= 6 (the gain Wg refers to the feodback from the local fre quency actually used, so that the value of K,coPreepond ing to it is snaller). Of the many tent of ogcillograns reconted, just two examples are given here, in Pig.s 6 and 7. ‘The trensicnts in Fig. 6, roconted with one of the network structures exanined in the study (see Table 1); but with a different losi-flow, confimed the procence, with separate tustare at Mantaro , of two critical ose Cillation modes with opcillation periods close to those : eh ht es. nee sero stabaers we ss Figs 6. Ganezator trip st Mantaro without and with ets, vilizers. Field teat of 2.26.1978 st 6 pam Figs Ts Instab{lity caused by elimination of stabili~ ere, which aro reinserted at point A. Field test of 2.261978 at 2 pene indicated ty the etudy (200 Fig. 5). Bach of these oscillation modes has anplitude that predominates at one of the two Mantaro generstore , and the slower mode is Yess danped. These nodes are both effectively damped by ‘the stabilising devices. Pigs 7. refers to a particularly severe situation as regamis stability, that is nevertheless of practical operational interest! it should te noted the high nega- tive damping value in the abeence of stabilizers , the nnariked increase in damping caused by the latter, and al 80 the strong nodification in the phase of the voltage oscillation in respect to the frequency oscillation caused by insertion of the otabilizers covetusrons ‘Tho availability of algorithms for evaluating the stability of systons in which there may be poorly damp fed oscillating modes, and for the approximate determin ation of danping sensitivity to the stabilising actions set up at various points in tho mtwens anamoudy ‘omesers the Pan pidityettictiveness an extent with wich mich imectigations ome madey a= compared with procedures based on simulations ‘Use can be mato of the fact that determination of ‘the initial sensitivities of damping to stabilising ax ‘ions can be relatedjwithout any model approxination,to calculation of the residues associated with the trane- fer function betwen the two pointe between which it i Aeeired to close the feedback. This also askes it poses ble, ant often cacy, to arrive at on experinental aval- uation of the sensitivities. The use of a simplified mg Gol to determine the sansitivities ie convenient for re Gucing calculation times for avoiding the Limitations ‘that may poseibly be imposed by existing voltageregula, tors. Various criteria during analycis or teste in the field generally make it possible, and often aasy, to de fine the nost suitable paraneters of the ctabilizing feedbacks , taking into account the characteristic: of ‘the voltage-regulator in whion they must be inserted ‘Tae application of analytical procedures of the ™ aforenentioned type to a practical case of oscillatory instability has nade it posible to arrive quickly at a solution eliminating difficulties and restrictions in power eysten operation. After stabilizer inplenentation food results were obtained in field tests, even in com itions that were very severe for stability, fully oom fiming that operation difficulties had been removed. REFERSUCES, [1] c.Barnter, B.Ferrari, K.B.Johanson "Report on the severe #0 the Questionnaire on Hlectronechanienl oscillation danping in power systens" , CHINE WO 31/3203 (to be published on Electra ,1979). F.Saccomarno "Senaitivity analysis of the charac istic rovte of « Lina ‘time-invariant “aynanis aysten: application to the aynthesie of danping ag Hong in electric power systems" IEAG/TS Gmnereas canbridge, Mase., USA. YeArcidiacon, EeFerrart, P.Sacconanmo "Studies on dseping of eloctronechanical osctliations in mlti sackine aystene with longitudinal structure", IEEB Transactions on Power Apparsius anf Systane,’ Yole PAS-95, no. 2, March/April 1976. YeAroidiacono, BoPerrari, Rullarconato, P.Sacconan- 0 "Analyaie of factore affecting the’ danping’ of owfrequercy oscillations in mltinachine eystend” 1976 GIL acting, paper 32-19. Micastrovilli, Rllarconato "Oonstzction of Linear sad nodela of electric power sysions for stebili= ‘ty studies" (to be published on L’Bnengia Elettri~ @ 1979). EN. Grabbe, S. Rano, DiBs Wooldridge "anbock of dntenation, conputation, and control” Woe Ty Com rol Fundatontals, ed. John Wiley & Sona, ince, New Yor, 1958 [2] G) U4) (5) {6}

You might also like