Real-T eal-Time ime Computer ompu ter Con tr o l o f Power Systems TOMAS E. D Y LIACCO,
SENIOR MEMBER
IEEE
Invited P a@
Abstract A
dramatic trmsfommtion in s y s t e m monitoring and
is taking place in the electric utility industry. Hew control control centers are being equipp& with multiprocessor real-time computers
s y m i n and controlling thegenerationand *nsystemvia 1 hqh-qwd data-acquisition subsystems and nteracting with the human operator via dynamic, dynamic,color, color,graphic graphic displays. Within the computer aa many .as a.hrmdred or more programs are available to m n n.a multiprogrammirrg environment in response to changing poner-system conditions and to operator‘s demands. The main objective of thie new development m on-line control is the enhancement of the security of the power system in order to matrtrin a high reliability of elec electric tric powerservice. serv ice. The concept of security control is d i m u m d and the noteworthy featurea of its pres present ent slate of development in recently designed and inscalled control controlcenters centers are described.
INTRODUCTION
T
HE EVOLUTION of automation nelectric power systems has been marked by outstanding developments
in he pplication of state-of-the -art tec technol hnology ogy in both instrumentation and control. Manyof these applications are required for local control to carry out such functions as regulation, -sw -switc itchin hing, g, protective relaying relaying,, generator load ing, etc., using measurementsobtained loca locally. lly. In certa in cases, cases, such as in protective elaying, he 11 control logic also requires information from oneor more remote jocations, such information being transmittedvia elep hone lines, powerpowerline carrier, or microwave facilities. Local controlhas controlhas evol evolved ved to he poi nt wher where eminicomputersare puters are being u used sed fo forr spec specific ific purpose ssuch as boilerturbinecontrols, turbine controls, or to consolidate a t a substation hitherto separate instrumentation instrumentation and control functions. There have been some emarkabledevelopment emarkabledevelopmentss in the use of digital comp uters for ocal control. However, in he discussion discussion o off real-time of the power syste ms, the main from emphasis of computer this papercontrol will be on systemwide control central location. A mor more e compreh ensive summary o off the developments in both local local and central controls may be found in [ 4 ] . In an over overall all centralized control system, all local local controls become functionally residenta residenta t t h eowest eowest level of a multilevel control hierarchy [l] the high est level level being a t the central control. In becoming part of a hierarch y, certain local local controls require informat ion links with igher level level c controls. ontrols. The implem entat ion of central control in pow power er systems originated rom wo ndependentcontrol ndependentcontrol equirementsand equirementsand developed intowo intowo eparate entralized ontrol ystems. One central control sys tem is fo forr the superv isory control and indication of of t ransmission and/or distribution equipment . In addition to equipment status indication and simple alarming functions hesupervisory hesupervisorycontrol controlsystem systemprovided provided he disto remotely patcher at hecenter hecenterwith with hemeans he means remotelyactuate actuate Manuscript received December 31 1973. The author is w i t h the Cleveland Electric I l l d ~ t i n g ompany, Cleveland, Ohio44101.
variousstation various stationequipment, equipment,primarily primarilycircuit circuitbreakers. breakers.The The centraliz ation of superviso ry control could be carri ed out a t one lo cation or in several district ffic ffices, es, dicta ted more or less by the geography of the territory served, and by the organizational structure of th e e elect lectrica ricall operation divi division sion of of th e utility company. Theother The othercentral centralcontrol controlsystem system is for heautomatic heautomatic control of the ou tpu ts f generating units in orde orderr to meeth e continuous changes in load demands. The central control of of generation which started out as a direct regulatory type control contr ol to make ystemgenerationmatch heminute-tominute changes in load was improved in the 1950’swith the addition of an optimizing-control lev level. el. Th e optimizing controlautomatically trol automaticallyallocated allocated hegeneration hegeneration equirement n such a way that he otal operati ng cost was a minimum. Thi s two-level control system for gene ration soon became a universal standard.The standard.Thedirect-contro direct-contro l. level, level, known n he industry as ‘automatic generation control” (or by the more of ‘load-frequency control”), became commondesignation mandatory for power systems which are interconnected; the optimizing-control level, known as ‘economic-dispatch ‘economic-dispatchconcontrol,” became a necessity necessityor or systemswith w ith a significant amount of thermal generating units. The supervisory-control and the generation-control generation-control functions each had its data-gatheri ng system and its own central and emotehardware. emotehardware.Generation Generationcontrol control evolved evolved rom an analog system to a digitally directed analog and finally, finally, in the 1960’s o a direct digital-control systeni. Similarly, supervisory-controlsystems visory-control systems evolved from one hardwire dmaster per remote to one hardwired master for several remotes and finally to a digital-comput er master. Thus by the end of th e 1960’s here were in service two ypes of digital-c omputer control install installati ationsons-the the dispatch computer and the supervisory-control visory-c ontrol computer computer-usin -using g smal l compu ters and requ iring no more data than what was esse essenti ntial al eith er for generagenerationdispatch tion dispatch or for supervisory supervisorycontrol. control.Digital Digital elemetry was also coming into use to replace analog telemetry. Up t o this time the man-mac man-machine hine interface cons consiste isted d o off st ri pc ha rt recordings, recordin gs, logg loggers, ers, indicat ing ight s,annunciator s,annunciator windows, console switchesor orpush push butt on panels, thumbwh eels, and4 oth er spec specialial-purpurhardwar e.More M ore ecently,black b lackand white and color CRT’s are being used for either the dispatch computer or the supervisory-control computer. Over he years he mprovements n generation control and nsupervisory n supervisory control controlhad had been n the hardw are used and in the effici efficienc ency y of of the control techniq ues. The basic basic objectives had remained the same,Le., Le., to control generation and to control remotely located devices devices or equipment. Near the end of the 1960’9 owever,power-system power-system engineersbegan engineersbegan a sysanalyzing he entire system operation problem from tems viewpoint, motivated by he evident needfor a more comprehensive nd more ffective ffective oper ating ontro l han had been been conventionally available o he power-syste m dis-
885
DY LIACCO: CO M PU T E R CO N T ROL ROL OF POWER SYSTEMS
patcher. We are now wi witnes tnessing sing in the deca de of the 1970’s the beginnings of a new wave of of power-syste m con trol sys tems, much broad er in scope of system monito ring and control due o he nteg ratio n of operating operatingfunctions, functions,and and he add itio n of a new dimension-“ dimension-“system system secu rity .” The ddition of system-security onsiderationso onsiderationsohe strictly generation-dispatch and supervisory-control requirements has cau caused sed a quantum jump in the evolu tion of realtime central control of power power systems by digital computer. This ransform ation is gro growin wing g in numbers hrougho ut he world. I n many cases utility compan ies are in th e proc process ess of of replacing o r augmenting existing digital-dispatch or supervisory-control computers less than 10 years old with the new security-controlsystems security-control systems of the 1970’s. This Thispaper paper discusses the concept of security control and the current state of i ts implementa tion in power power systems.
SECURITYONTROL Provision for system security has alwaysbeen alwaysbeen an inherent pa rt of sound system design. Howev However, er, due to econ economic omic and other considerations, only so much security can be built into a system. I t has generally generally be been en assumed th at as ong as a system is built builtaccordin accordin g o prevailin prevailing gdesign standards,any standards,any abnormality in operation requiring control bey beyond ond th at pro pro-vided vide d for by conventi onal, autom atic d devic evices es would would be take n care of of by the human operator. Historical ly, this assumption had been been blithely made without prov providin iding g the opqrator with adequate tools toaid toaid him in he comp complex lex decision-making proces pro cesses ses concern ing security. In some companies wh where ere t he securityproblem problemhad had become become quiteacu quiteacu te, procedures had b e e n ins tit ute d to .aid the operator by way o off securi ty pro pro-grams run off off-li -line ne on general-purpose computers, in cert ain casess using rem ote erminals or, in a few rare nstances,a case nstances,a dedicatednetwork dedicated networkanalyzer analyzer or analog computer. These applications represent a transitional tep rom onventional control to security control. “Security control” or a “security “security control system” may be defined as a system of of integrat ed automatic and manua l controls for the main tena nce of electric power servic service e und er all condit ions of operatio n [2]. Note rom his definiti definition on th at security control is a signif significan icantt depar ture from the tradition al dispatch control or supervisory-control systems. Firstly, the proper integrationof integrationof all the necessary automatic- and manualcontrol functions requires a total systems approach with the human operator being being an inte gral part of the control-system design. Sec ondl y, he mission mission of secur itycontrol itycontrol is all-encompassing,recognizing tha t con tro l decisions by th e mancomputersystem computer systemmust must be made madenot not j u s t when the power system is operating normally but alsowhe alsowhen n it is operating under abnormal conditions. As power systems have grown in size and have bec become ome more tightly coupled , the problem problemof making he ightoperatin ightoperatin g dec decisi isions ons under undervarying varying conditions has become extremely difficult.
The Three Operating States In [ l ] and [ 2 ] the oper ating co nditions of a power system are characteriz ed in terms of th ree oper atin g states-normal, states-normal, emergency, and restorative. Let u s revi review ew this concept as it is helpfu helpfull in the d discu iscussio ssion n of of secu rity c ontrol and the c urrent st ate of of its development . The power systemmay systemmay be assumed as being operated under condit ions expressible expressible in the form of t wo sets of constraints [2], [ S I :
G x, u)= O
H x , u)ZO
load constraints operating constraints
where G and H are function vectors and x u
vector of dependent dependentvariables; variables; vector of independent or “control variables.”
Sollberger [SI i n extendin g he concept of securitycontrol securitycontrol to industrial systems in general, general, call callss one set of constrain ts the “load constraints” and the other set the “operating conconstr ain ts, ” and gener generaliz alizes es both nto nequalit ies. In our discussion cussio n the equalitie s are the load constra ints while the inequalities are the operating constraints. Fundamentally, the load constraints impose the physical equations which satisfy the requireme nts that the load demands w wil illl be met by the syste m; whil while e the operating constraints impo impose se maximum maximum or minimum operating limits e.g., loading limits, voltage limits, etc.) on variables associat associated ed with the compo nent parts of of t he system. If both he oad and operating constrai nts are satisfie satisfied, d, the system is said to be in the norm normal al ope rating state. In response to herelatively herelativelysmall smallminute-to-minute minute-to-minutechanges changes n load, a power system may be considered considered as going from one normal norm al state o anot her and each norm normal al state may be assumed to be a quasi-steady-state condition. On th e occurrence occurrence of of a severe distu rban ce (e.g., a large load change, a loss of generation, a short circuit) a system may settle down to a ew normal state or may go to either an emergency or restorativ e operating state. In the emergenc emergency y state the operating constraints are notatisfied. a tisfied. In the restorative state the operating constraints are atisfied but not the load constraints. Two types of of emergency may be n oted. The first first typ e occurs when, aftera afteradisturban ce, he power power system emains stable and continues operating but with he operating con con-straintsnot straints not fully fully met, i.e., i.e., with some some equipmen t oading limits exceeded or with withabnormal abnormalvoltage voltage levels a t certain locations. This type of emer gen cy con diti on which we shall refer to as “steady-s tate emergency ” may be tolerated for a reason able period period of of tim e, generally generally allowing correc tive action to be taken. Such corrective action shoulde shoulde effective enough to toprevent prevent o r limit limitdama dama ge o he overloade overloaded d equipmen t. T h e second ty pe of of emergency occurs when, because of a disturba nce, the powe powerr system become becomess un stable, during which which timeboth time both heoperating heoperatingand and he load constrain tsare tsarenot not being met . Th is typ e of emer gency condit ion which we shall referr t o as “dynamic instability” takes place refe place in a very short period of of t ime and , unless unless a prope r fast correct ive action is taken, ends up with theowe theowerr system in the restorative state, i.e., in a partial or a total shutdown. Th e overall objective of of secu rity control may be restate d as fol follow lows: s: to keep the power power system operating in t he normal normal state, i.e., to prevent o r to minimize the departures from the normalstate normal state ntoeith ntoeith er he emergenc emergency y or the estorative sta te. To realiz realize e an effectiv effective e stra tegy for carryin g out his objective let us look more closely into the concept of system security.
The Concept of Sy ste m S e c urity Systemsecurity System securitymay may be considered as the theability ability of a power syste m in normal power normal operation to undergo a disturbance without getting into an emergency condition. The system is then said to be “secure.” On the other hand, a normal operat-
886
ing system would be ‘insecure” if there is a disturbance which could bring ab ou t an emergency operating condition. If one considers all possible disturbances i t would be impossible impossible to find a secure securepower powersystem. system. In practicesystem practicesystemsecurity security is determined with reference to an arbitrary subsetof subsetof the completedisturbance plete disturbanceset. set.This Thissubset subset s call called ed the “next-co “next-conntingency”set. tingency” set.Th Th e choice of th e compo sitio n of th enextcontingency set is dictated by the probabili ty of occurrence of th e contingency within th e next sho rt perio period d o off time (in the theorder order of minutes) and he conseque consequences nces to he sys tem should hecontingency hecontingencyoccur. occur. In mostpower p ower syste ms he next-contingency set includes, as a minimum, the following types of disturbances:
PROCEEDINGSOF THE IEEE, JULY 1974
Security monitoring involves th e processing, of th e measured data to determine the system operating conditions and the violations violati ons o off t he oper atin g cons trai nts H w , u ) B O . Also part of securitymonitor securitymonitoring ing s the on-line on-linedeterm determ ination of th e network topology as required for display and for models used by other on-line functions. Th e large amou nt of telemetry required, with its attendattendan t problems of errors in meas ureme nt and n data ran smission, has justified considerations of exploi ting redun dancies in measurement so as to obtain “best” estimates of the system variables using Kalman filtering and other stochastic
approximationmethods.T This his un unct ctio ion, n, refe referr rred ed t o h the [26] sbecoming industry as “state “stateestimation” estimation”[lo]-[14], [lo]-[14], accepted as a necessary part of security monitoring. 1 any circuit out; We see then that just the addition of security monitoring 2 any generating unit out; to conventional functions creates a new typ e of of syst em con3 any phase-to-phase or 3-phase short circuit. trol using state-of-the-art hardware systems and sophisticated data-processing methods. Other types of disturbancesmay disturbancesmay hemore disbe added, hemore Security Analysis: Security analysis is the determination turbances ncluded n henext-contingency set the more o f th e s ecur ity of the sy stem ba sed on a next-contingency set. stringen t the sy stem-secu rity requ irements ecom ecome. e. While a security analysis may be made for both steady-state For a given givennext-contingency next-contingency set th e se t of all allnormal normal emergency anddynamic nstability he rendhas anddynamic rendhasbeen been o operating states may be partitioned nto wo disjoint subsets -se cur e and ins insecu ecure. re. Tha t is, is, a normal operating system have a separate analysis for each of the two types of emeris either secureor insecure. We see the n tha t for security con- gency. One reason for this is the extre me difficu lty of implementing a dynamic security analysis with present methods f trol to accompli accomplish sh its objective of preventing or minimizing for stead y-sta te se departures from the normal state itwould be highly desirable stability evaluation. On the other hand, curity analysis several approaches arepossible arepossible and are in use. t o be able to manipulate the system so t h a t i t s t a y s as long Basically, heseappr heseapproach oaches es start with a knowledge of he region on of of the normal st ate . Th is is is a as possible in the secure regi present state of the thesystem system obtainedfrom from hesecurity hesecurity as obtained remarkablydifferentcontrolphilos ophy rom hat of th e monitoring monitor ing function. function. Th e syste m s h en este d for various traditionaldispatch traditional dispatchcomputer. computer. ts implicat ions n erms of next contingencies contingencies by, in effect, effect, solving solving for he chan ges in the data requirements,nformat requirements,nformat ion processing, controdesign l andalgorith and algorith ms, manman-mach machine ine nter nterface faces, s, and compu ter re- system conditions for a given contingency and checking the new values against the operating constraints. quirements canbe canbe readily appreciated. Pat ter n recognition recognition has been ried with some promising promising Strategy for Security Control results [SI, [IS], [I61 as an alternative approach to security on-line. EsT he effective effectiveness ness of security control in keeping the power analysis but the method has not been pu t to use on-line. system secure for as long as possible depends heavily on th e sentially, a small set of on-line meas urem ents are used as security indicators. A securit y functi on of thes e indicat ors wil willl control done during the normal operating state. Wewill refer then clas classify sify the st at e of the system as being secure o r inseto his control as ‘preventive ‘preventivecontrol” control”since since its preventiv e cure. The security function or pattern classifier is derived by character is what distinguishes it from conventional control. based on a trai ning set of off-line studies. Preven tive cont rol shoul d not only ta ke care of all of th e tr a- a learning algorithm based of the pattern-recognit ion approach is the transditional functions involved in controlling generation to meet Th e appeal of the load but should also determine the actual operating confer of t he need or full-scale full-scale simulat ion ionstudies studies (especially dynamicstability dynamic stabilitystudies) studies)from from heon-line he on-linecomputer computerwith with ditions of the syste m, assess system secudty or bgth types of limitedcapabil limited capabilities ities to an off off-li -line ne computerwith computer withmore moreand and emergency emergen cy (i.e., (i.e., steady-state emergency and dynamic instabetter modeling andomputation resources.owever, owever, be taken n bility), and determine he corrective action o further furth er work remains remains to be don e to develop develop a satisfactory case the syst em were insecure insecure.. Consideration Consideration of these tasks pattern-recognition method for on-line implementation. leads u s to t he discussi discussion on of t hree important ideas: security Security-ConstrainedOptimization: O ptimization: If the yste m were monitoring, ecurity nalysis, and ecurity-constrained found to be insecur insecure e by the se curity-an alysis function the ne xt optimization. problem s todetermine todeterminewhether whether hesystem hesystemcan can be made SecurityMonitoring: Security Monitoring: Securitymonitoring Security monitoring is th e on-line secure. This becomes a security-constrained ptimization identifi iden tificat cation ion and the dynamic disp lay (to the human operato r) of of the actual o perating c ondition s of t he power system. problem where we have to find the best operating condition which satisfies not only the load constraints and the operating a systemwide nstrumentation Security monitoring requires is, is, miniminion a greater scale and variety than that required by contro controll constraints but also the security constraints. That mize u ) subject to F x , system without security monitoring. To gat herall the system data every few few seconds seconds and bring it to a control location reG r , u)=O load constraints quires a high-speed digital data-acquisition system interfaced H w , u ) h O operating constraints with a central computer system. Th e central computer sysS x, u ) 2 0 security constraints tem operates in real-time and upervises the data acquisition, processes proce sses the da ta received received,, andsupports andsupports hedisplay he displaysubsubwhere F s the operating cost function. system.Finally, system. Finally, he man-mach man-machine ine requireme nts ntsentail entail he The security-function vectorS may consist of all the load the u se of devicessuch devicessuch as colo colorr CRT’s for display and as and operating constrain ts for each of the next contingencies contingencies interface for operator inputs and operator-initiated controls. whoseoccurrences whose occurrenceswould wouldcause causeemergencies emergencies [17], [18]. Or
DY LIACCO:
COMPUTER CONTROL O
POWER SYSTEMS
single e security function S x , u ) L O may be t he singl
used used in he pattern-recognition method of se curity analysis [SI. Th e sesecurity-constrainedoptimization would would determinehe determine hebest corrective action for making a system secur secure. e. Th e extr a cost involved in mplemen ting hiscorrectiveaction would be presented to the operator who would would then decide whether to carry it out or not. Theoreti cally, he optimizin g lev level el o off preventive control may be formulat ed from the very outset using using the complete set of constrain ts, as ndicated in the pr preced eceding ing paragraph. I n practice this is not necessa necessary ry nor is it desirab le. The opti-
887 MODERN CONTROL-SYSTEM INSTALLATION The mplementa tion of securitycontrol ystems which which star ted with the coming of t he 1970’s is gaining gaining mo mentum throughout thepow thepower er systems in the world. world. By the endof endof 1973 there were approximately 30 systems in operatio n, under construction, or on order, which which fall in the category of the new generation of modern, state-of-t he-art control systems. Based on what is being developed in the power industry a compositemodern composite modern controlsystem for a bulk power system controlsystem would hav e the follo following wing feat ures and andon-line on-line functions. 1 Hierarchic Hierarchical al str uct ure consisting of of several evels of
mizing-control algorithm would not only be too complex to implement on-line but would result in anunnecessary ncrease crea se in operatin g expense expense and in computer overhead .
computer systems. Examples: a) National center, region regions, s, divisions. b) Power pool, member companies,ivisions. companies,ivisions. c) System center,divisions center,divisions or substations. Th e com combine bined d s trate gy o off securitymonitoring, securitymonitoring,security security 2) Hierarchicalstru Hierarchicalstru ctur e of control functio ns, .e., preanalysis,and a nd astly ecurity-constrained ecurity-constrainedoptimization is a ventive, emergency, restorative, and each divided into 3 layers correct and practical ap approach. proach. For th the e present it is the only d i r e c t , op opti timi mizi zing, ng, ad adapt aptiv ive. e. viableapproach viable approach oon-line o on-linepreventive preventivecontrol. control. I t is a good good 3) Dual rocessors or multiproc multiprocessors essors lus red und ant illu stra tio n of th e power of decomposition a nd multilevel peripherals. org aniz ati on of a acontrol controlsystem system o accomplish a complex 4) High-s High-spee peed d digital digital d at a teleme try and state-of-the-art controlobjective control objective[6]. [6]. To summarize, he strategy works works as data-acquisition equipment. follows: Theoptimizing Theoptimizingcontrol control sformulated s formulatedas as simply simplyas as 5 Color CRT’s with graphics for interactive display. possib pos sible le using only the load constraints and a very s mall su b6) Dynamic, color, color, wallboard wallboard group display. se t of the opera ting const raint s (e (e.g. .g.,, only the generator lim7) Automatic generation control. its but not the equipment loading limits). In most cases, cases, the 8) Econo Economic mic d ispatch control. optimizing control is is decompos decomposed ed i nto the real power optimi9) Automatic voltage (var) control. zation (or economic dispatch) and the reactive power optimization (or var dispatch ). Sinc Since e the changes n var dispatc h are for many systems less frequent than the changes in economic dispatch,var v ardispatch is usual usually ly treat edasan asan open-loop adaptive control. Further simp simplifi lificati cation on of the economic dispatch is obtained by usi using ng for load constrain ts a single single function equating the algebraic sum of the pow power er injectio ns and the system losses to zero. zero. The security-m onitoring function working indep ende ntly of th e economic dispatch detects online when when certa in operating constraints become become semi-binding or bind binding. ing. When this happens the economic economic dispatch should be modified so that the semi-b inding and bindi binding ng constraints are includ included ed in the set of operating constraints. That is, is, th e size of th eoperating operatingconstraint constraintset set considered is adju sted with the numb er of likel likely y and actual violatio ns. Th e secur ityanalysis function, either working working indepe ndently or triggered by thesecurity-monitoring thesecurity-monitoringfunction, function, chec checks ks the secu rity of the system. f the system systemwer were e insecure, a security-constrained optimizing program wou would ld then be run to fin find d the correctiv e action. I t is readily ev evident ident tha thatt the development of a good preventive control enhances tremendously the ability f a system to tostay staysecure secureand and herefore minimize minimizess thedepartures thedepartures nto the theabnormal abnormalcon con diti ons of theemergency the emergencyand and estorative states.Both states. Bothemergency emergencycontrol controland and estorativecontrol estorativecontrolare are still needed for a complete security-control system. However, However, these controls are diffic difficult ult to develop and mplement for a variety of reas reasons. ons. L et it suf suffic fice e for this pap er to remark th at preven tive control is relativ ely easier easier to mple ment being well within thecapab ilities of pres ent- day technology and analysis.urthermore, analysis. urthermore,he heact acthatreventive h atreventiveontrol ontrol
Automaticircuit ircuitestoration e storation19]. (Re-energizes 10)s Automatic circuit which which have been d ropp ed ou t of service by a disturbance.) 11) Supervisory contro controll (breakers, capacitors, transformer taps, generating unit startup, and shutdown). 12) Sche duling of of genera tion resources, Le., Le., fossil, fossil, hydro, nuclear (long term, medium term, and short term). 13) Security monitoring. 14) State estimation. 15) Steady-state security analysis. 16) Dynamic security analysis. 17) Aut omatic ystem roubleanal roubleanal ysis [19]. Processes breaker and protective relaying operations to analyze system disturbances.) 18) On-line load flow. flow. (An inte ract ive load flo flow w available to the syst em oper ator to deter mine power power flows flows in in th e network for a given set of of cond itions.) 19) Optimum power flow. (An optimal load-flow solution where the loadload-flo flow w equations aresolved aresolved with an optimization routine to minimi minimize ze t he objective function , usually, usually, but not necessari neces sarily, ly, the total oper ating ost.) 20 20)) On-line short- circuit calcul ation. 21) Short-term load forecast.Predicts forecast.Predictshe he ext 24-h load curve.) 22 BusBus-load load forecast forecasting. ing. (Pred icts the load a t each individual load point. ) 23 23)) Various Various support programs for generation dispatching, such as: schedule for startup and shut down of units, power interchange ransactionswith ransactions with neighboring syst ems, calculation of of gene ration reserves, etc. 24) Logging Logging and historical historical data mana geme nt.
offers a workable control strategy gives impetus to its higher priority of developme nt. T hus most most of of the applications of the security-control oncepthat oncepthat re being evelopedn the power industry are i n the area of of pre ventive control . The development of emergency controland controland estorativecontrol estorativecontrol is curr entl y of a lower lower prior ity and no significan significantt innov atio ns have as yet ee een n implemente implemented. d.
Ther e is is no control system th at has all of the fun ctio ns just enumerated. Ingeneral, Ingeneral, except in thecase thecase of of a few power systems, the control systemdesigns systemdesigns that have been developed posses pos sesss only rather small subsets of this list. In the Appendix, Table I ists some somebasi basic c data on cont rol syste ms which which are in serv service, ice, under construc tion, or on order, as of November off t he 1973. This table is intend ed to convey a rough picture o
888
PROCEEDINGS
. w
w . 1
sW,.avc
O
THE IEEE JULY
1974
889
DY LIACCO: COMPUTER CONTROL OF POWER SYSTEMS
TABLE I
Continued)
I
la
n
/
1973
I s A
Implemented by analog controller. Legend for on-line functions: ACRAutomaticCircuit ACRAutomatic CircuitRestora Restora tion. Generation Control AGCAutomatic AGC Automatic Control..
OL F OP F
On-Lin On-Line e Load Flow. Optimum Power Power Flow.
ASTA Automa tic System Trouble Analys Analysis. is. OSC On-Line Short Circuit. AVC AutomaticVoltage/Var AutomaticVoltage/VarControl Control.. SA St ea dy at at e Security Anal Analysi ysis. s. Supervisory Breaker EC Emergency ontrol. BC ontrol. SE Stat e Estimation. Estimation. EDC EconomicDispatch EconomicDispatch Control. NOX Minimum NO. Emissionispatch. S M Security Monitoring. Emission ispatch. SVCSupervisory SVC Supervisory Voltage Control.
PROCEEDINGS OF TH E I E E E , JLXY 1974
890
types of of sys tems tha t have bee been n or are th e po power wer industry napp rox imat ely he decade. In Tabl e
beingdeveloped in first hal halff of this
I several hierarchic hierarchical al computer systems are n
evidence, mos t of of them being 2-level 2-level hierarchies. hierarchies. S yste m 24 is a 3-level 3-level hiera rchy, while the c omb inat ion of Sy stem s 2 and 15 forms a 4-level hierarchy. Of the systems already inervic in ervice, e, Syste m 7 is of par ticu lar interest.Controlling interest. Controlling a small power system, consisting of 9 bussess and 3 major transmission lines busse lines [25], the Tok ke cont rol center of the Norwegian Water Resources andElectricity andElectricity
ing syst em usually usually has to be modif modified ied in in order to s upp ort the requirements for the following. configurat configuration ion with ntercom puter 1) A dual ompute r communication. Provision sionss for maintaining backup for critical control 2 Provi funLtio ns and for auto matic failover failover in case case of of various types of failure. 3) A multiprogrammingnvironment where various modes of progra m execution are re qui red. non4) Interfacing with the data-acquisi tion system and nonstandard peripherals.
Board has the distinc tion of having the first first on-l on-line ine applicamodes of s ta rtu p. 5 System initialization for various modes 6) Provisions for testing and for training. tion in the world world of of s tate estima tion for security monitoring of a po power wer syste system. m. I t als also o has an emergency control function CONCLUSIONS which automatica lly educes educesgeneration generation a t the theappropriate appropriate plan t in case of problems resulting from transmission outage. Th e need need for security control is not unique to th e power For a large, interc onnec ted ystem he first succe successful ssful ind ust ry [SI. However, because of t he uniqueness uniqueness of of th e pr odexperi mental esults of sta teestimation teestimation wer were e obtained n uct tha t t he po power wer industry supplies plus the unceasing need Septemb er 1972 1972 in a portion of of the American Electric Power to maintain that supply, the implementationo implementationof se cur ity consystem [26]. trolwith trol with heaid heaid of real -tim ecompute rs is being being pursued Another interesting development is the use by System 11 perhaps more vigorousl vigorously y in utilities than in other industrial of the opt imu m power flo flow w for for calcu lating cost coeff coefficien icients ts pro proces cesses ses.. T he power indus try, in developin developing g this new typ e of which whi ch a re factore d in the economi c-dispa tch-con trol calc calculaula- control system, has taken advantage of advances in real-time tions. The ve cto r of cost coefficien coefficients ts is given by computer design design,, state-of-the -art evelopmen tsn. tsn. atar acquisitionsystems, acquisition systems,dynamic dynamicand and nteractivedisplays, nteractivedisplays,syssys= F -*rad, P r tems engineeringconcepts and algorithms algorithms n mathematical where J T is the transposeof transposeof the Jacobian matrixo matrixof t h e powerflow equation G(x, u ) =0, and grad, P , is the gradient of the power a t the reference bus used in the power-flow equation. This is th e first onon-line line app lication to economic economic dispatc h of a concept rigorous rigorously ly tr eate d for both real and reactive power dispatching with equality and inequality constraints by Carpentier [24] in 1962. Unique among those systems in serv service ice or under develop develop-mentsshe utomatic ystem ment ystemrouble rouble nalysis unction implemented in System 11. This functio n require s the monitoring of protective relaying operations, both primary protection and backup protection, n addition to monitoring circuitbreakeroperations. Inother ystemsonlycircuit-breaker status is monitored to obtain information about short-circuit type of disturbance. The inclusion of a short-circuit calculation as an on-line function by System 4 is another unique feature feature not found in other systems. Thecontrol-system The control-systemfeatures featuresand and on-line functi ons mentioned tione d in this sect section ion ha ve been descr described ibed rather sketchil y, the objectbeing objectbeing to pres ent an overview overview o off what is being done. More detai ls of specific con tro l syste ms and of v arious functions, some of of which may not have been been refe referred rred to in thi s paper, may be found in in the lite ratu re [8], [9], 19]-[23 19]-[23]. ]. As mentioned earlier earlier most of the sys tems are implement ing only a small numb er of of on-line functio ns. Howe ver, even security-controlunctions woul would d entail a a minimum of security-controlunctions rather ophistic ated assemblage assemblage of computers,display displaydedevices, vice s, d ata-acqu isition systems, and a host of real-time programs. The require ments for the ne new w ty pe of of control system ha ve
programming,tochasticpproximation,n nd dstimation thepry. More work remains remains to be done in the other more difficult areas of security control such as dynamic security analysis, emergency ontrol, nd estorative ontrol.Research R esearch nd analysis effort effortss are cont inui ng o resolve resolve some of these remaining problems problems.. Doubtle ss th e yea rs be yond 1975 1975 will will bear witness to further advances in the real-time computer control
also caused demand s or real-time computer operating systems which initiaIl initiaIly y wer were e either nonexistent or inadequate for the application.Development effo efforts rts have been expendedn expendedn creating special special operat ing systems where where none were were availab le or in augmenting the real-ti me capabilitie s of w hat the computer manufacturer had tooffer. A standard real-time operat-
omissions. Many new, but strictly supervisory-controlor generationI except as they control systems, are not included in Table bel belong ong to a computer hierarchy and satisfy criteria2) criteria2) and 3). I t should should be mentioned, mentioned, howev however, er, th at these strictly supervisory-control systems have becom become emod modern ernize ized d to the ext ent
of powe powerr syst ems [ 7 ] . APPENDIX Tabl e I is presented to give an overall picture of of t he new types of of cont rol syst ems that are being being impl emented in the power pow er ind ustry. These control systems are desig designed ned fo r the central control of the overall power power system and are distinguished from earlier types of automat ic control by the provision visi on for system security. Tabl e I is fairly complete as far as the following criteria ar e concerned concerned 1) 2) 3) 4)
central control for generation and transmission; use of d ual or multipro cessors; use of col color or CRT’s for interactive display; inclusion of, a t least, the security-monitoring function for security control.
Some systems not fully fully meeting these criteria have been been included because of certain security-control features. Althoug h the author tried his best to gather as much information as po poss ssib ible le about what is goi going ng on in the indus try, there re oubtless some some state-of-the-artystems state-of-the-artystems which have escaped his atten tio n, The aut hor apologize apologizess for these
DY LIACCO: COM PUTER CONTROL OF POWER SYSTEMS
tha t dua l processor processors, s, mostly mostly minicomputers, and col color or CRT’ s are being used. I t should also be noted t h at many of the companies n Table I whi which ch have new contro l systems under developm ent alreadyhave exis existing ting generation- nd upervisory-control systems using digitalcomputer digitalcomputer s, some with withCRT’s, CRT’s, which had been placed in service only a few years ago. The last column of Tab le I lists heon-line heon-line unctions planned for each control system. Only the majoron-line majoron-line functions are enumerated. Standard support programs andff-line functions which form a lar ge pa rt of the soft war e of these
89 1 [l l ] a) R . E. Larson, W. F. Tinney, and J. Pesch Peschon, on, “Stat e estimation estimation in power systems, Part I: Theoryand Theory and feasibility,” ZEEE Trans. Power A p p . S y ~ t .O, ~ PAS-89, pp, 345-352, Mar. 1970. b R. E. Larson, W. F. Tinney, L. P. Hajdu, S. Piercy, Hajdu,and andD. D. ‘State estimation in power power systems, Part 11 11:: Imp Implementat lementation ion and applications,” ibid., vol. PAS8 9, pp. 35 3533-36 363, 3, Mar. 1970. [12] F. W. Stagg, J. F. Dopazo, 0. A . Klitin, and L. S . VaS l yck, Techniques for the real-time monitoring of power system operations,” Z E E E T r a n s . P o w er er A p p . Sy s t . , vol. PAS-89, pp. 545-555, Apr. 1970. [13] F. Galiana and E. Handachin, “Combined network and power sy te,m dynamic state estimation,” in Proc. 4 h ower S y s t n n s C o m p u b Wdion Cmf. (Grenoble, France), Sept. 1972. (141 F. Ariatti, P. Ricd, V Castagnoli, and L. Marzio, “Methods for electric power system state estimation:Comparative estimation:Comparativeanalysis analysis of
control systems are not included.
computing simulations based upon generalized load flow, tracking, and least squares methods,” in Proc. 4th Power Syskms,Computation Cm f. Grenoble, France, Sept. 1972. ACKNOWLEDGMENT [15] “Bulk power security asse sment, IBM Res. Div., Final Rep. Edison Elec In st. Project RP90-3, Nov. 197 1970. 0. The autho r wishe wishess t o tha nk his many i ndustry coll colleag eagues ues [la] C. K. Pang, F. S. F’rabhakara, A. H. El-Abiad, and A. J. Koivo, in the United States, Europe, and Japan for their cooperation ‘ k u r i t y evaluat evaluation ion in power power systems systems using pattern recognition,” ZEEE Tra ns. Power Ap p. Syst . , vol. PAS-93, PAS-93, pp. 9 96969-976 976,, Ma y/ in putting together as up-to-date an information summary as June 1974. possible for the tabulation (Table I of modern, state-of-the- [If] J. W. Carpentier, ‘Differential injection methods, a general method for secure and optimal load flows,” flows,” in Proc. 8th Power Zudushy Comart, cont rol syst ems. Permissi Permission on was obtained from each of Applications puter (Minneapolis, Minn.) , June 4-6, 1973, pp. Cmf. the utilities represented o use use the dat a in this pap paper. er. Th e 255-262. da ta for the apanese ystem wer were e obtained hrough he [l8] 0. Alsac and B. Stott, “Optimal load flow with steady-state wur er A p p . Syst . , vol. PAS-93, pp. 745-751. ity,” Z E E E T r a n s . P o w er good offices of the theCen Cen tral Research Research Inst itut e of Electric May/June 1974. Power Pow er Industry in Japan. I191 T. E. D y Liacco, B. F. Wirtz. and D. A . Wheeler. “Automation of the CEI-system for security,” I E E E T r a n ss.. P ow ow er er A p p . S y s t . , vol. REFERENCES PAS-91, pp. 831-84 831-843, 3, Ma y/Jun e 1972. [l] T. E. D y Liaqm, ‘The adaptive reliability reliability control system, system,” ” ZEEE [ZO] D. W. E. Crook and J. H. Ha ms, ‘The w e of on-line computing T r a n s . Po Po w eerr A p p . S y s t . , vol. PAS-86, pp. 517-531 517-531,, Ma y 1967. faalities at the NationalControl National Control Centr e of the C.E.G.B.,” presented he emerging emerging concept of s ecur ity control,? in Proc. 1970 -S,y m p .“T P w e r S y s t n n s (Purdue Univ., Univ., Lafayette, Ind.) , May 19 1970 70.. [3] -, ‘Control of power systems via the multi-level concept,”
Syst. Res. Center, Case Western Reserve Univ., Cleveland, Ohio, Rep. SRC-68-19, 1968. [4] G. Quazza, “Automatic control in electric power power systems,” in Auto matica, vol. 6. New York : Perga mon, 1970, pp. 123-1 123-150. 50. (51 M . H. Sollberger, ‘Security of an industrial system,”Syst. system,”Syst. Res. Center, Case Western Reserve Univ., Cleveland, Ohio, Rep. SRC 71-3, Ja n. 1971 71;; [a] I. Lefkowitz, Multi-level approach applied to controlsystem controlsystem design,” A S M E T r a n s . , ol. 88, ser. B, June 1966. 1966. [7]A. H. El-Abiad, “Ne w d directions irections in power systems reliability and ‘security,” presented a t the 3rd Southeastern Sy mp. System Theory, Georgia Inst. Technol., Atlanta, Apr. 1971. [8] W. K. Pence, “New Michigan pool center stsystem security,” E l c c W o r l d , Oct. 21, 1968. 191 C. F. Ham and K. B. Rennie, ”Control center ups system performance,” Elcc. WorM Aug. 15,1973. [lo] a) F. C. Schweppe and J. Wildes, “Power system stat ic-s tate estimation, Part I: Exact model,” Z E E E T r a n s . P o ow w e r A p p . S y s t . , vol.
by Digital Computer, Imperial College College Sc Technol., London, England, Sept. 1973. (2 (211 11 R. G. Rice and G. W. Stagg, “Application program design criteria for the Southern Company contro controll system,” in Proc. 8t h P w n Znd u s lr lr y C o m pu pu ttee r A p p l i c a t i a s C m f (Minneapolis, Minn), June 4-6, 1973, pp. 128-134. [22] K. no, M. Na rba, and T. Nishimori, ‘Automatic ‘Automatic load dispatching system in Kansai Electric Power Co., Inc.,” presented a t the Symp. mplementation of Real-Time Power System Control by Digital Computer, Imperial Colle College ge Sa. Technol., London, England, Sept. 1973. [23] J. C. Russell and W. Rades:, “The Wisconsin Electric Power Company energy energy control system, in Proc. 8th Power Industry Compukr Applications C m f (Minneapdis, Minn.), June 4-6 1973, pp. 135141. [24] J. W. Carpentier, ‘Contribution i ’6tude du dispatching 6conomique” (“Contribution to economic dispatching”), Bull. Soc. Fr. Elec.. vol. 3. ser. 8, Aug. 1962. [25] J. Svoen, S. A. Fisman, H. H. Faanes, and A. Johannesen, “ The computer in the power system. The on-line, cloeed-loop, cloeed-loop, approa ch
PAS-89, pp. 120-125, Jan. 1970. b) F. C. Schweppe Schweppe and D. B. Rom, ‘Power system static-state estimation, P ar t 11: Approximate model,” ibid. vol. PAS-89, pp. 1251 3 0 Jan. 1970. c) F. C. Schweppe, “Power ystem stat ic-s tate estimation, Part 111: 111: Implementation,” ib id ., vol. PAS89, pp. 130-135, Jan. 1970.
for control of generation and overall protection at Tokke power plants,” plants, ” presented a t the Conf. Int. Grands Resea Reseaux ux El Electrique ectriques, s, Aug. 1972. [26] J. F. Dopazo, S. T. Ehrmann, 0 . A. Klitin, nd A. M . Sasson, ‘JustiETrans. fication of theAEP theAEP real time load flow project,” Z E E ETrans. Power Ap p . SySt. SySt.,, V O ~ . AS-92, pp. 1501-1509, Sept./Oa. 1973.
[2]
Real-Time Power System Control a t the Symp. Implementation of Real-Time