Trends in Power System Protection and Control

Transcription

1 Proceedngs of the 32nd Hawa Internatonal Conference on System Scences 1999 Trends n Power System Protecton and Control Mroslav Begovc 1, Damr Novosel 2, and Mle Mlsavljevc 1 1 Georga Insttute of Technology, Atlanta GA, mroslav@ee.gatech.edu 2 ETS Insttute, ABB Power T&D, Ralegh, NC, Damr.Novosel@chtet.mal.abb.com Abstract As a consequence of deregulaton, competton, and problems n securng captal outlays for expanson of the nfrastructure, modern power systems are operatng at ever-smaller capacty and stablty margns. Tradtonal enttes nvolved n securng adequate protecton and control for the system may soon become nadequate, and the emergence of the new partcpants (non-utlty generaton, transmsson, and dstrbuton companes) requres coordnated approach and careful coordnaton of the new operatng condtons. The paper revews the key ssues and desgn consderatons for the present and new generaton of SPS and emergency control schemes, and evaluates the strateges for ther mplementaton. 1. Introducton System-wde dsturbances n power systems are a challengng problem for the utlty ndustry because of the large scale and the complexty of the power system. When a major power system dsturbance occurs, protecton and control actons are requred to stop the power system degradaton, restore the system to a normal state, and mnmze the mpact of the dsturbance. The present control actons are not desgned for a fast-developng dsturbance and may be too slow. Further, dynamc smulaton software s applcable only for off-lne analyss. The operator must therefore deal wth a very complex stuaton and rely on heurstc solutons and polces. Today, local automatc actons protect the system from the propagaton of the fast-developng emergences. However, local protecton systems are not able to consder the overall system, whch may be affected by the dsturbance. The trend n power system plannng utlzes tght operatng margns, wth less redundancy, because of new constrants placed by economcal and envronmental factors. At the same tme, addton of non-utlty generators and ndependent power producers, an nterchange ncrease, an ncreasngly compettve envronment, and ntroducton of FACTS devces make the power system more complex to operate and to control, and, thus, more vulnerable to a dsturbance. On the other hand, the advanced measurement and communcaton technology n wde area montorng and control, FACTS devces (better tools to control the dsturbance), and new paradgms (fuzzy logc and neural networks) may provde better ways to detect and control an emergency. Better detecton and control strateges through the concept of wde area dsturbance protecton offer a better management of the dsturbances and sgnfcant opportunty for hgher power transfers and operatng economes. Wde area dsturbance protecton s a concept of usng system-wde nformaton and sendng selected local nformaton to a remote locaton to counteract propagaton of the major dsturbances n the power system. Wth the ncreased avalablty of sophstcated computer, communcaton and measurement technologes, more "ntellgent" equpment can be used at the local level to mprove the overall emergency response. Decentralzed subsystems, that can make local decsons based on local measurements and remote nformaton (system-wde data and emergency control polces) and/or send pre-processed nformaton to hgher herarchcal levels are an economcal soluton to the problem. A major component of the system-wde dsturbance protecton s the ablty to receve systemwde nformaton and commands va the data communcaton system and to send selected local nformaton to the SCADA centre. Ths nformaton should reflect the prevalng state of the power system. 2. Types of Dsturbances and Remedal Measures Phenomena whch create the power system dsturbance are dvded nto the followng categores: angular stablty, voltage stablty, overload and power system cascadng. 1

2 Proceedngs of the 32nd Hawa Internatonal Conference on System Scences Angular stablty The objectve of out-of-step protecton as t s appled to generators and systems, s to elmnate the possblty of damage to generators as a result of an out-of-step condton. In the case of the power system separaton s mmnent, t should take place along boundares whch wll form slands wth matchng load and generaton. Dstance relays are often used to provde an out-of-step protecton functon, whereby they are called upon to provde blockng or trppng sgnals upon detectng an out-of-step condton. The most common predctve scheme to combat loss of synchronsm s the Equal-Area Crteron and ts varatons. Ths method assumes that the power system behaves lke a two-machne model where one area oscllates aganst the rest of the system. Whenever the underlyng assumpton holds true, the method has potental for fast detecton Voltage stablty Voltage stablty s defned by the System Dynamc Performance Subcommttee of the IEEE Power System Engneerng Commttee [3] as beng the ablty of a system to mantan voltage such that when load admttance s ncreased, load power wll ncrease, and so that both power and voltage are controllable. Also, voltage collapse s defned as beng the process by whch voltage nstablty leads to a very low voltage profle n a sgnfcant part of the system. It s accepted that ths nstablty s caused by the load characterstcs, as opposed to the angular nstablty whch s caused by the rotor dynamcs of generators. The rsk of voltage nstablty ncreases as the transmsson system becomes more heavly loaded. The typcal scenaro of these nstabltes starts wth a hgh system loadng, followed by a relay acton due to ether a fault, a lne overload or httng an exctaton lmt. Voltage nstablty can be allevated by a combnaton of the followng remedal measures means: addng reactve compensaton near load centers, strengthenng the transmsson lnes, varyng the operatng condtons such as voltage profle and generaton dspatch, coordnatng relays and controls, and load sheddng. Most utltes rely on plannng and operaton studes to guard aganst voltage nstablty. Many utltes utlze localzed voltage measurements n order to acheve load sheddng as a measure aganst ncpent voltage nstablty [4] Overload and Power System Cascadng Outage of one or more power system elements due to the overload may result n overload of other elements n the system. If the overload s not allevated n tme, the process of power system cascadng may start, leadng to power system separaton. When a power system separates, slands wth an mbalance between generaton and load are formed wth a consequence of frequency devaton from the nomnal value. If the mbalance cannot be handled by the generators, load or generaton sheddng s necessary. The separaton can also be started by a specal protecton system or out-of-step relayng. A quck, smple, and relable way to re-establsh actve power balance s to shed load by underfrequency relays. There are a large varety of practces n desgnng load sheddng schemes based on the characterstcs of a partcular system and the utlty practces [5-6]. Whle the system frequency s a fnal result of the power defcency, the rate of change of frequency s an nstantaneous ndcator of power defcency and can enable ncpent recognton of the power mbalance. However, change of the machne speed s oscllatory by nature, due to the nteracton among generators. These oscllatons depend on locaton of the sensors n the sland and the response of the generators. The problems regardng the rate-of-change of frequency functon are [7]: A smaller system nerta causes a larger peak-to-peak value for oscllatons. For the larger peak-to-peak values, enough tme must be allowed for the relay to calculate the actual rate-of-change of frequency relably. Measurements at load buses close to the electrcal center of the system are less susceptble to oscllatons (smaller peak-to-peak values) and can be used n practcal applcatons. A smaller system nerta causes a hgher frequency of oscllatons, whch enables faster calculaton of the actual rate-of-change of frequency. However, t causes faster rate-of-change of frequency, and, consequently, a larger frequency drop. Even f rate-of-change of frequency relays measure the average value throughout the network, t s dffcult to set them properly, unless typcal system boundares and mbalance can be predcted. If ths s the case (eg. ndustral and urban systems), the rate of change of frequency relays may mprove a load sheddng scheme (scheme can be more selectve and/or faster). 2

3 Proceedngs of the 32nd Hawa Internatonal Conference on System Scences 1999 Adaptve settngs of frequency and frequency dervatve relays may enable mplementaton of a frequency dervatve functon more effectvely and relably. Ths wll be dscused later. 3. Possble Improvements n Control and Protecton Exstng protecton/control systems may be mproved and new protecton/control systems may be developed to better adapt to prevalng system condtons durng system-wde dsturbance. Whle mprovements n the exstng systems are mostly acheved through advancement n local measurements and development of better algorthms, mprovements n new systems are based on remote communcatons. However, even f communcaton lnks exst, systems wth only local nformaton may stll need mprovement snce they are envsoned as fallback postons. The modern energy management system (EMS) can provde system-wde nformaton for the network control and protecton. The EMS s supported by supervsory control and data acquston (SCADA) software and varous power system analyss tools. The ncreased functons and communcaton ablty n today's SCADA systems provde the opportunty for an ntellgent and adaptve control and protecton system for system-wde dsturbance. Ths n turn can make possble full utlzaton of the network, whch wll be less vulnerable to a major dsturbance. 3.1 Angular stablty Out-of-step relays have to be fast and relable. The ncreased utlzaton of transmsson and generaton capacty as well as the ncreased dstance of power transmsson are some of the factors that cause an out-ofstep stuaton to develop rapdly. The nterconnected nature of power systems cause large geographc areas to be affected by an out-of-step condton. The present technology of out-of-step trppng or blockng dstance relays s not capable of fully dealng wth the control and protecton requrements of power systems. Central to the development effort of an out-of-step protecton system s the nvestgaton of the mult-area out-of-step stuaton. The new generaton of out-of-step relays has to utlze more measurements, both local and remote, and has to produce more outputs. The structure of the overall relayng system has to be dstrbuted and coordnated through a central control. In order for the relayng system to manage complexty, most of the decsons have to be taken locally. The relay system s preferred to be adaptve, n order to cope wth system changes. To deal wth out-of-step predcton, t s necessary to start wth a system-wde approach, fnd out what sets of nformaton are crucal, how to process nformaton wth acceptable speed and accuracy Voltage Instablty The protecton aganst voltage nstablty should also be addressed as a part of herarchcal structure. Decentralzed actons are performed at substatons wth local sgnals and sgnals obtaned from slow communcaton wth other substatons and/or central level (e.g. usng SCADA data). The hgher herarchcal level requres more sophstcated communcaton of relevent system sgnals and a coordnaton between the actons of the varous substatons. The recommended approach for desgnng the new generaton of voltage nstablty protecton s to frst desgn a voltage nstablty relay wth only local sgnals. The lmtatons of local sgnals should be dentfed n order to be n a poston to select approprate communcated sgnals. However, a mnmum set of communcated sgnals should always be known n order to desgn a relable protecton, and t requres the followng: (a) determnng the algorthm for gradual reducton of the number of necessary measurement stes wth mnmum loss of nformaton necessary for voltage stablty montorng, analyss, and control; (b) development of methods (.e. senstvty analyss of reactve powers), whch should operate concurrent wth any exstng local protecton technques, and possessng superor performance, both n terms of securty and dependablty Power System Cascadng and Load Sheddng Strateges Conventonal load sheddng schemes wthout communcatons are not adaptve to system condtons whch are dfferent from the one used n the load sheddng desgn. For the relays to adapt to the prevalng system condtons, ther settngs should change based on the avalable spnnng reserve, total system nerta, and load characterstcs. These values may be perodcally determned at the central ste from SCADA data and provded to the relays usng low speed communcatons. In addton, the actual load, whch would represent an assgned percentage for sheddng at each step, may be perodcally calculated at a central ste based on the actual load dstrbuton. However, the system characterstcs may change dependng on the separaton ponts. If the 3

4 Proceedngs of the 32nd Hawa Internatonal Conference on System Scences 1999 separaton s controlled from a central ste or can be predcted, an algorthm may calculate the settngs and assgn the approprate load n coordnaton wth swtchng actons. However, hgh speed communcaton may be requred to and from the central locaton for fast-developng dsturbances, such as mult-machne angular nstablty. Another aspect, may be addng a correcton element to a scheme. If only slow speed communcatons are avalable, a fast load sheddng scheme may be mplemented to stop system degradaton. When adequate nformaton s avalable, correctve measures may be appled. If the composte system nerta constant s known, the actual power mbalance may be calculated drectly from the frequency. Ths detecton should be fast (to avod a large frequency drop) and done at the locaton close to the center of nerta. Hgh speed communcatons are requred to ntate load sheddng at dfferent power system locatons. Further, changes of load and generaton, wth frequency and n partcular voltage, mpact the power mbalance and calculaton of the average of the frequency dervatve. In addton, the power system mbalance changes after the ntal dsturbance due to dynamc system changes. Thus, relay settngs should be based on the spnnng reserve, total system nerta, and load characterstcs and dstrbuton. In concluson, sophstcated models and/or hgh-speed communcaton may be requred for accurate estmaton of the amount and dstrbuton of the load to be shed. If communcatons are avalable, t s easer and more relable to calculate the amount of load to shed from the swtchng nformaton and the msmatch (based on data on load and generaton before the separaton) n the sland. To avod dsadvantages of the underfrequency load sheddng and dffcultes wth mplementng the rate-ofchange of frequency functon, the automated load sheddng that wll reduce overloadng or prevent system nstablty before the system s solated s proposed as an advantageous strategy. 4. Example: Angular Stablty An algorthm for predctng the locaton at whch an outof-step can take place followng a dsturbance n a largescale system s shown as an example of the herachcal protecton and control strateges usng communcatons. To mplement ths scheme, one needs a central computer that receves nformaton from across the system. The sets of crucal nformaton that requre fast communcatons consst of: generator speeds, and changes n lne status. Other nformaton needed by the algorthm are generaton and load levels. Usng these sets of nformaton, a smple and quck processng method s able to tell, wth a hgh degree of accuracy, (1) whether an out-of-step s mmnent, and (2) the boundary across whch ths out-ofstep wll take place. Ths algorthm has been tested thoroughly usng a Monte-Carlo-type approach. At each test, random values are assgned to lne mpedances, load, generator nertas, as well as dsturbance locaton. It s found that the algorthm s capable of makng accurate predcton. To llustrate how the algorthm works, consder the power system shown n Fgure 1. Ths system s a modfed verson of the IEEE 39-bus test system. In Fgure 1, each generator s marked wth a crcle, and each load wth a square; the sze of each symbol ndcates relatvely the power generated or consumed at the node. For example, Generator 34 supples more MW to the grd than does Generator 38. gens; loads x outage x x PREDICTED O.O.S. LOCATION 36 Fgure 1. Graph of the 39-bus test system rotor angles, 0 deg tme smulaton tme, sec. #37 #38 #34 #36 #33 #35 #31 #32 #39 Fgure 2. Tme doman smulaton of the system. #30 4

5 Proceedngs of the 32nd Hawa Internatonal Conference on System Scences 1999 A dsturbance s ntroduced to the system where two lnes are smultaneously removed (each lne s marked by an 'x' n Fgure 1). Ths nformaton s fed to the algorthm, whch predcts that an out-of-step wll occur across the lne Fgure 2 reveals that the two generators 37 and 38 eventually separate from the other generators. All lne angles have been checked and none but lne 2-25 ndcate the boundary of the out-of-step. The angle of crtcal lne s shown n Fgure 3. Ths confrms the result of the algorthm. deg angle of crtcal lnes 0 1 tme Fgure 3. Angle of crtcal lne Such an algorthm requres a centralzed scheme and hghspeed communcaton lnks across the wde system. Decentralzed scheme requres communcatons wth a central locaton. Accordng to ths herarchcal scheme, each regonal computer ssues control actons to allevate problems that are mmnent wthn ts jursdcton; the coordnaton among regons s left to the central computer. 5. Example: TCSC Control In ths secton we brefly present a method for onlne detecton of nterarea oscllatons. The method does not requre the topologcal knowledge of the system and s based on the spectral characterstcs of the locally observable data. Moreover, we ntroduce a dfferent approach to the power system stablzaton. We apply the globally avalable nformaton about the system to the adaptaton of TCSC (thyrstor-controlled seres compensator). The control we propose attempts to optmally adjust the lne admttance on sample to sample bass n order to reduce the nterarea oscllatons n the power systems. We explan how such approach may combne the global system nformaton wth locally observable quanttes and eventually result n the dampng of nter-area modes Power System Modelng and Spectral Characterstcs We consder a generc N-machne, p-bus power system descrbed by the followng swng equaton: M ω = D ω + P ( t) δ = ω ω B where δ and ω represent the devaton of the rotor angle and ts speed relatve to the synchronous reference frame, P a (t) s the acceleratng power,.e. mbalance between the mechancal and mechancal power, M s the nerta constant for machne, ω s the synchronous B angular frequency, and D s the dampng factor correspondng to the machne. Swng equaton can be lnearzed to obtan x = A x + B u a x s the vector obtaned by concatenatng where vector rotor angle devaton vector and rotor speed devaton vector, and u s the vector of power mbalances for each machne. Informaton about possble modes of the system can now be obtaned by egenanalyss of matrx A, and ts partcpaton factors. Ths however, s not the only way that the nformaton on the oscllatory modes can be obtaned. In a system subject to the mpulse exctaton, the oscllatory modes can be determned from the power spectral densty of the system response. In other words, the modes would appear as peaks of the power spectral densty functon of state varable x calculated as: where F represents the Fourer transformaton. Peaks of the power spectral densty correspond to the modes of the nterarea oscllatons, and can be therefore determned from measurements of x over a tme wndow, f we consder x as a tme-frequency dstrbuton. Descrptve (measurement) quantty we use s the machne speed. We thnk of nterarea oscllatons as energy exchanges between the groups of the machnes, ths energy exchange should be vsble from the machne speed data. Moreover, the machne speed s easly measurable quantty, thus makng a perfect canddate for 5

6 Proceedngs of the 32nd Hawa Internatonal Conference on System Scences 1999 descrptve quantty for a pattern recognton task. Algorthm we propose s followng: Collect machne speed data. Perform spectrum estmaton on machne speed data. Identfy the frequences of possble modes of nterarea oscllatons. Perform clusterng on the phases of machne spectrum data at nter-area frequences. machne belongs. The phase data s summarzed n Table 1. Indeed, machnes 1 through 8 have the phase angle of approxmately 11 degrees, machnes 11,12 and 13 have phase angles of approxmately 3 degrees, whle phase anlges of speed spectra of other machnes do not belong Identfy nter-area groups wth clusters of phases of machne spectrum data. to any of the groups. Table 1. Phase angles of the power spectral densty at the nterarea oscllaton mode at 0.4 Hz. Fgure 4. Transent dsturbance of the test system as descrbed n the text below. We llustrate the performance of the algorthm on the followng example: Fgure 1 shows the machne speeds of the 16 machne 64-bus NPCC system subject to the fault on lne between busses 28 and 29. The fault s cleared by trppng the lne after three cycles but stll dvdes machnes n three groups thus producng the nterarea oscllatons. In ths smulaton we used the transent model for synchronous machnes. It s obvous that the system s dvded nto three groups that oscllate aganst each other. In one group we have machnes 1 through 8, n second machnes 11 through 13 and all other machnes oscllate by themselves. Modal analyss of the lnearzed system reveals fve oscllatory modes under 1Hz: at Hz, at Hz, at Hz, at Hz, and one at Hz. Partcpaton factor analyss ndcates that machnes most lkely to oscllate at Hz are machnes 1 through 9 and 13 through 16. Machne 15 seems very prone to the mode at.5hz, machnes 10 through 13 are susceptble to the mode at Hz, and machnes 14 through 16 are prone to oscllatory mode at Hz. We use the phase angles of the power spectral densty at the gven mode to determne the group to whch We propose to use the nformaton about groups of machnes belongng to nterarea modes to formulate an emergency control by modulatng the admttance of the thyrstor-controlled seres capactor (TCSC). The control we propose s based on the premse that we only want to enhance the stablty of the system by applyng the admttance adjustment control. The man control stll reles on power system stablzers, whch usually perform ths task very well. We would lke to use the admttance control to mnmze the nterarea oscllatons whch are exhbted as the flow of energy between machnes or groups thereof. We use the rate of change of machne speeds as an ndcaton of the energy change at partcular machne. In that sense, we would keep the undesrable varatons of the machne veloctes to ther mnmum, reducng the energy transfers between dfferent groups of machnes. In turn, reduced energy exchange reduces the nterarea oscllatons, whch s the goal of our control. Whle ths type of control does not guarantee the stablty, t does guarantee that the Lyapunov exponents of a system wth such control wll be smaller than the ones of the system wthout the control. Ths s a result of the facts that machne states are the state varables of lnearzed system, and that the logarthm s a monotonous functon. In that sense, the system s more stable than the system wthout the control. 6

7 Proceedngs of the 32nd Hawa Internatonal Conference on System Scences 1999 Moreover, proposed type of control can be computed wth a cost of QR decomposton for the matrx of senstvty coeffcents and one matrx multplcaton. In addton, ths type of the control guarantees the mnmum energy soluton,.e. mnmum per-sample change of lne admttances. The stablty of the control computaton procedure, low computatonal cost, and easy parallelzaton of the process ths algorthm very appealng for the real tme applcatons on multprocessor archtectures Mnmum Energy Control It was stpulated that the changes n machne speeds are an ndcator of nter-area oscllatons. Moreover, the change of machne speeds ndcates the drecton of the energy fluctuatons. The reducton of machne speed changes does therefore reduce the nterarea oscllatons. Per-sample machne speed change can be wrtten n followng form: ω dt = ω + ω k k 1 k δ k = A' + P y ωk Above two tasks have to be performed wth mnmum change of the lne admttances snce the TCSC devces have range typcally lmted to a certan percentage of the nomnal lne admttance, for a varety of reasons. Above consderatons exclude the apror knowledge of the fault locaton. Takng nto account the above consderaton, selecton process may be based on an extensve senstvty analyss spannng a large number of stuatons, ether by usng some form of a composte objectve functon, or by applyng some effectve means of combnatoral optmzaton, such as genetc algorthms. The end result s the allocaton of a lmted number of TCSC controllers, whch perform reasonably well over a fnte set of dfferent dsturbance scenaros. The results of some comparatve tests, obtaned by tme doman smulaton on the system and fault used n Fgure 4 are shown below. where P s the senstvty of the machne speed changes wth respect to the lne admttance. By settng ω k dt = 0, the control s computed as the soluton of followng equaton: δ k P y = A' ωk In contrast to ths optmzaton, control of system modes would mply the mnmzaton of ω. Snce TCSC devces present a permanent change n the topology of a power network, ther locaton needs to be determned before the fault has occurred. Havng ths n mnd, TCSC selecton process has to consder several crtera: Varous and possbly multple faults may happen. All machnes need to be controlled at the same tme to nsure the control of all nterarea modes (ths condton may be relaxed by adaptve trackng of the nterarea groups). Snce all machnes partcpatng n a certan nterarea mode wll have smlar angles and rates of change of angles (speeds), t s techncally possble to use just a representatve machne from a group. Ths may be useful n speedng up the control processng to enable real-tme control. Fgure 5. Comparson of the machne #16 speeds (fault as per Fgure 4): Orgnal speed n center of nerta reference frame. No control s appled.. Machne #16 speed wth TCSC modulaton control Machne #16 speed wth power system stablzer control. sold lne Machne #16 speed wth both PSS and TCSC controls appled. The presented scenaros nvolve stuatons when no control was appled to the system under dsturbance, and 7

8 Proceedngs of the 32nd Hawa Internatonal Conference on System Scences 1999 when group of 5 optmally selected and tuned PSS and 5 TCSC was used, both ndvdually and together. The opportunty to speed up processng n power system stablzng devces to enable real-tme control s n the ncluson of the dstrbuted archtectures, whch combne low per-unt cost wth hgh effcency. The control algorthm we have descrbed n prevous secton can be easly mplemented n the parallel archtecture wth great reducton of complexty and large parallel effcency. 6. Concluson A large dsturbance such as a sudden outage of a transmsson lne may trgger a sequence of events leadng to machne swngs, voltage problem, and eventually power outage n a large area of the system. The role of a protecton and control system s to tmely predct the system nstablty, to perform actons to restore the system to a normal state and mnmze the mpact of the dsturbance. As communcaton and computer technology contnue to mprove, and protecton and control becomes more ntegrated, an applcaton of adaptve system-wde protecton s becomng more feasble. Snce any mprovement n system-wde protecton and control products provdes sgnfcant savngs to utlty, the decentralzed systems that provde mproved and economcal soluton for the system-wde dsturbance problems are very attractve. Automated load sheddng that wll reduce overloadng before the system s solated s an mproved soluton n comparson to underfrequency load sheddng. Although local measurements may suffce f tasks are smple (eg. protecton aganst few contngences only), nformaton communcated ether from central locaton or from remote substaton seems necessary for more sophstcated requrements. Mcroprocessor-based coordnated protecton, montorng and control systems are the key to nnovatons n power system operatng phlosophy. The coordnated system s clearly the future of relayng technology. As communcaton and computer technology contnue to mprove, and protecton and control become more ntegrated, the applcaton of the adaptve wde area dsturbance protecton concept s becomng more feasble. Snce any mprovement n the wde area protecton and control strategy provdes sgnfcant savngs to the utlty, the ntellgent systems that provde mproved and economcal soluton for the wde area dsturbance problems are very attractve. Intellgent emergency control systems (.e. systems descrbed n the paper) provde more secure operaton and better emergency responses, allowng utltes to operate at closer transmsson and generaton margns. 7. Acknowledgement Part of the work presented n ths paper was funded by the Natonal Scence Foundaton under Grant #ECS We gratefully acknowledge ths assstance. 8. References [1] L. Cederblad and T. Cegrell, "A New Approach to Securty Control of Power Systems -Local Protecton Coordnated wth System-Wde Operaton," IFAC Symposum, Brussels, September [2] A. P. J. Malt, et al, "Computer Based Supervsory Control and Energy Management System for the Cty of Cape Town," IEE Proceedngs, Vol. 135, pp , January [3] Voltage Stablty of Power Systems: Concepts, Analytcal Tools, and Industry Experence, IEEE Publcaton, 90TH PWR, [4] System Protecton and Voltage Stablty, IEEE Power System Relayng Commttee, IEEE Publcaton, 93THO596-7 PWR, [5], L. H. Fnk, et al., Emergency Control Practces, IEEE Transactons on PAS, Vol. 104, pp , September [6] System Dsturbances: North Amercan Electrc Relablty Councl, NERC Reports. [7] A. Apostolov, D. Novosel, and D.G. Hart, Intellgent Protecton and Control Durng Power System Dsturbance, 56th annual APC, Chcago, Aprl [8] NERC Plannng Standards and Gudes, "System Protecton and Control, Draft, North Amercan Electrc Relablty Councl, Interconnecton Dynamcs Workng Group of the NERC Engneerng Commttee, [9] K. Vu, M. Begovc, D. Novosel and M. Saha, Use of Local Measurements to Estmate Voltage-Stablty Margn, Proceedngs of the PICA 1997 Conference, May 11-16, [10] Proceedngs of Bulk Power System Voltage Phenomena-III: Voltage Stablty, Securty and Control. Davos, Swtzerland, Aug