Engineering and Research Center

Bureau of Reclamation

October 1978

1 October 1978 AUTOMATIC GENERATION CONTROL - NOTES AND 1

6. P E R F O R M I N G O R G A N I Z A T I O N C O D E

OBSERVATIONS

7. AUTHOR^) 8. P E R F O R M I N G O R G A N I Z A T I O N R E P O R T NO.

W. B. Gish REC-ERC-78-6

9 . P E R F O R M I N G O R G A N I Z A T I O N N A M E A N D ADDRESS 10. WORK U N I T NO.

Bureau of Reclamation Engineering and Research Center

11. C O N T R A C T OR G R A N T NO.

MS-230 (1-76) Bureau of Reclamation TECHNICAL REPORT STANDARD TITLE PAGE

-

- 1

- 1

1

- I

1

-

Same 14. SPONSORING A G E N C Y C O D E 0

I

IS. S U P P L E M E N T A R Y N O T E S

16. A B S T R A C T

The concept of automatic generation cbntrol is often considered complex because of the freedom each utility has i n choosing individual characteristics within a basic control philosophy. This report endeavors to separate the basic philosophy from the individual characteristics to allow a clearer understanding of the control philosophy. Discussions of individual characteristics used by the Bureau of Reclamation are also presented.

7 . K E Y WORDS A N D D O C U M E N T A N A L Y S I S

0. DESCRIPTORS-/ +load-frequency control/ 'interconnected systems/ *power system operations/ 'power interchange/ algorithms/ automatic control/ control systems/ governors/ power dispatching/ baseloads/ generating capacity/ hydroelectric power/ electric generators/ computer applications

5. IDENTIFIERS-/ automatic generation control

c . COSATI F i e l d / G r o u p 09B COWRR: 0902 18. D I S T R I B U T I O N S T A T E M E N T

4 v a i l a b l e f rom the N a t i o n a l T e c h n i c a l In format ion Serv ice , o p e r a t i o n s I i v i s i o n . Spr ing f ie ld , V i r g i n i a 2 2 / 5 1 .

19. S E C U R I T Y C L A S S (THIS REPORT)

U N C L A S S I F I E D 2 0 . S E C U R I T Y CLASS

ITHIS P A G E )

2 1 . N O . O F P A G E

25 2 2 . P R I C E

AUTOMATIC GENERATION

CONTROL- NOTES AND

OBSERVATIONS

by William B. Gish

October 1978

Electr ic Power Branch Divis ion o f Research Engineering and Research Center Denver, Colorado

UNITED STATES DEPARTMENT OF THE INTERIOR * BUREAU OF RECLAMATION

ACKNOWLEDGMENTS

The contributions of Mr. Brich of the Watertown OperationsOffice, Western Area Power Administration, and Messrs.Knutson and Realson of the Phoenix District Office, WesternArea Power Administration, in reviewing the text andsuggesting modifications are greatly appreciated.

CONTENTS

Abbreviationsandsymbols. . . . . . . . . . . . . . . . . . . . . . .Introduction. . . . . . . . . . . . . . . . . . . . . . . . .Summary. . . . . . . . . . . . . . . . . . . . . . . .BasicAGCcontrol concepts . . . . . . . . . . . . . . . . .Areacontrolerrorformation. . . . . . . . . . . . . . . . . . . .

Interchangepower. . . . . .Frequency and time signals

. . . . . . . . . . . . .

. . . . . . . . . . . . . . . .

Area control error disposition . . . . . . . . . . . . .

Modifyingthe ACE. . . . . . . . . . . . . . . . . . . . . .The allocator. . . . . . . . . . . . . . . . . . . . . . . .

Quality of the AGC . . . . .Bibliography. . . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . .

TABLE

Table

AGC minimum operating criteria . . . . . . . . . . . . . . . . .

FIGURES

Figure

123456

Simplifieddiagramof the systemsynchronizingloop. . . . . . . . . .SimplifiedAGCcontroller. . . . . . . . . . . . . . . .TypicalPIDcontroller. . . . . . . . . . . . . . . . . . . . .Dynamicscheduleof a joint-ownedgenerator. . . . . . . . .Componentsof ACE. . . . . . . . . . . . . . . . . . . . .Example allocator with variable-gain, permissive-control with

area requirement and plant requirement. . . . . . . . . . . . . .Example allocator with mandatory control with area

requirement and plant requirement. . . . . . . . . . . . . . .Example allocator with permissive control and participation

factors basedon plant limits. . . . . . . . . . . . . . . . . .

7

8

1111111

Page

iii1112

49

10

1214

2122

21

2336

11

16

17

19

ACE area control error P participation factor for a plant

AGC automatic generation control PA actual total interchange

AR area regulation Pc contracted power schedule

B frequency bias in megawattsbetween utilities

per 0.1 hertzPDY dynamic schedule

BGi base point for generatingPGi power generated by generatingplant "i"

plant "i"

FA actual power systemfrequency PLL lower limit power of a plant

FS scheduled frequency Pm measured intertie power

the number of the generating Pnm intertie power not measuredplant

II inadvertent interchange Ps total interchange schedule

k the index for summing plants PUL upper limit power for a plant

KD derivative gain PID proporti ona I-integ ra I-derivative

KF total frequency bias PR plant requirement or stationrequirement

KI integral gains LaPlace variable

Kp proportional gainSHADE measurement error

Ks system synchronizing coefficient correction

KSL speed-level motor gain SLM speed-level motor

KT time error sensitivity TA actual or clock time measured

megavolt ampere rating ofmeasured from power

MVA system frequencyequipment

MWS megawatt second inertia constantTg governor equivalent

ti me consta nt

N the total number of generating plantsTmin a control area mechanical time constant

(related to inertia)NAPSIC North American Power Systems

Interconnection Committee Ts standard time from WWV

ABBREVIATIONS AND SYMBOLS

iii

,w.".

INTRODUCTION

The electric power systems of the United Statesand Canada contain control systems to regulateand control energy flow of immense proportions.These control systems work together to maintainclocks to within 4 seconds of WWV standard timewhile simultaneously controlling the weeklygeneration and distribution of 40 to 50terawatthours of energy (145 x 1015 to 180 x1015 joules). This very significant engineeringaccomplishment is often overlooked.

The control system for this major task uses over100 separate primary control units operating inparallel to command thousands of generatorspeed controllers (governors). Each of theprimary control units has different character-istics and control philosophies, and only the mostbasic concepts are similar. These basic conceptsfor control are contained in the NAPSIC operatingmanual [9].1

The primary controller is the AGC (automaticgeneration control) system. The AGC systemconcepts have been described and discussed in amultitude of books and technical articles. A verybasic text for help in understanding AGC systemconcepts is Control of Generation and PowerFlow on Interconnected Systems by Mr. Cohn[12]. The comments and notes on AGC containedin this report were developed on the basicprinciples given by Mr. Cohn. This report is aseries of discussions of techniques used in AGCthat are relative to applications within the Burea uof Reclamation and the Western Area PowerAdministration.

SUMMARYThe AGe concept is not a complicated controlphilosophy. Complications arise as the require-ments of energy marketing and resourcemanagement are factored into the controlphilosophy. The philosophy becomes moreinvolved as the opinions of good performancecriteria are included. Although the basic AGCconcept is used throughout the power industry,each utility will interpret the energy marketing,resource management, and performance criteriadifferently, and an individualized control conceptwill be developed. This report is to help separatethe basic control concepts from the individualizedcharacteristics and promote understanding ofthe automatic gene'ration control.

1 Numbers in brackets identify selected references listed inthe Bibliography.

BASIC AGC CONTROL CONCEPTS

The actual operation of the power systemrequires a form of speed control. The basiccontroller for speed is the governor whichmaintains control of the "prime mover" energy ofa generator in response to generator speed. Allgovernors on all the generators within a powersystem work in parallel because all generatorsare locked together in synchronism. Applicationof this concept alone could adequately control apower system provided all the tielines couldwithstand the resultant power flows. However,several improvements to system operations areimplemented when a master control supervisesthe governor settings. These improvementsinclude:

. Abilityto determine the individual systemloads and allow each system to carry itsown load.

. Maintaining tieline flows to interconnectedutilities to allow accounting of power salesbetween utilities.

. Maintaining system frequency at 60 Hzwhich permits the widespread use of elec-tric clocks.

Note that the AGC is a supervisor of the normalgovernor. The AGC should never attempt tooverpower the governor during system tran-sients. Although a control to dampen intertieswings could be developed for AGC [29], thebasic concept of AGC is to allow such swings tobe controlled by governors, voltage regulators,power system stabilizers, and doc interties.

Power systems operating personnel view AGCsystems differently than control engineers.Operating personnel associate the AGC directlywith interchange schedules, frequency, and timeerror. The control engineer views these as resultsof the AGC. The control engineering conceptspresented in the next few paragraphs is an effortto provide as broad an understanding as possible.The remainder of the report does not requirethese concepts for a basic level of understanding.

The basic system response which the AGCattempts to control is "accelerating power."Accelerating power is the difference between allsystem generation power and the total systemload including losses. If the power system isconnected to a neighboring utility, the power onthe tieline flowing out of the system is considereda load.

The frequency deviation is a measure of theintegral of the total accelerating power on theentire power system. Because of the synchron-izing effect of all interconnections, and becausethe AGC should ignore power changes due tosynchronizing effects, the AGC frequency inputis assumed the integral of accelerating power.The tieline flow error from the scheduled flow(orinterchange power deviation) is proportional toaccelerating power. In figure 1, Tm is themechanical time-constant related to inertia andKs is the synchronizing coefficient related totieline strength [66); system damping is notshown for simplicity. The rema.inder of the basicAGC system is shown in figure 2. The constant KFis normally known as the AGC bias term, 1OB. Theconstants KSL and Tg are related to speed-levelmotor gain and governor response time, res-pectively. The LaPlace variable, s, is used in thedenominators to indicate integration. This type ofcontrol system is similar to a PID (Proportional-Integral-Derivative) control system shown infigure 3, except the derivative gain, KD.is zero.The input reference to the AGC is zero levelaccelerating power which always remains thereference. The interchange schedule and fre-quency schedule are used to facilitatethe energyand time accounting and are not the actualreference to the control system.

Signal ApproximatelyProportional toAccelerating Power

+

Tieline

Power( Interchange

Power)System

- AcceleratingPower I

TmSSystemInertia

Sys t emGeneration

Sys temLoad

Since the accelerating power is constantlydisturbed by random load changes and powersystem disturbances, the AGC attempts to keepthe first and second integral of acceleratingpower (frequency and time) at zero. Theproportional and derivative of accelerating powerare controlled by the synchronizing and dampingforces of the synchronizing loop. The first integralof accelerating power is also controlled by thegovernor in parallel with AGC. However, thedroop of the governor forces accelerating powerto zero after a frequency shift occurs (propor-tional control of accelerating power). Again, theproportional-integral-derivative control from thesynchronizing loop and the governors must neverbe overpowered by the first and second integralcontrol from the AGC unless all governor andsystem characteristics are completely known.

AREA CONTROL ERROR FORMATION

The discussion of AGC control usually falls intotwo categories. The first category includes theformation of the ACE (area control error) and thesecond is the disposition of the ACE or theallocation process. The formation of the ACE isdiscussed in this section. The ACE is an errorsignal which, when positive, indicates excessivegeneration. A positive ACE always requires a

Frequency Deviati onof Nearby Systems

Signal ApproximatingIntegral ofAccelerating Power

+

SystemFre quenc

Deviation

Figure 1. Simplified diagram of the system synchronizing loop.

2

lilllll

+

IntegralPath(-ACE)

KFA/locations Attempt

to Hold ConstantGain Proportional

PathReference

EquivalentSystemGovernor

I1+li s

InterchangeTieline Power

~5

KSL+ -L

T sEquivalent

Speed -levelMotor

III

I II System I

L~roo~ ~oa~ J

SystemGeneration

Figure 2.-Simplified AGC controller.

Feedback

Kp

Output

Integral

ErrorK05

Derivative

Figure 3.- Typical PID controller.

3

reduction in generation. The ACE signal is alwaysopposite in polarity to the normal error signalused in most other types of control systems.Normally, the ACE has the units of megawatts.

The equation assumed by NAPSIC for all AGCcontrollers is

ACE =(PA - PSI - 10B (FA - FS)

where PAis the actual total interchange power inmegawatts with positive power flowing out of thesystem; Ps is the total interchange schedule forthe system; FA is the actual frequency of thesystem in hertz; FS is the scheduled frequency;and 10B is the system frequency bias inmegawatts per hertz (B is always negative). Thesubtraction of reference from the actual quant-ities is the reverse from normal control systemprocedures.

Interchange Power

Although the interchange power deviation is ameasure of accelerating power, the primarypurpose of the interchange power is foraccounting. The AGC is primarily used tomaintain interchange schedules and it ac-complishes this task by holding the acceleratingpower to zero, using the generation, rather thanthe tieline, to adjust accelerating power. Becausepower generated in the system will use the pathof least impedance to seek a load, there is no wayto constantly determine the exact power flowpath between several interconnected utilities. Anenergy meter on a specific intertie cannot easilydetermine the energy flow for the several utilitieswhich may have power contributions on theintertie. Thus, the schedules and related rates forthe buying and selling of power between utilitiesis recorded as contracted schedules and the totalschedule of all interchanges is developed. If theactual total interchange deviates from this totalschedule, it is not usually possible to properlyadjust any individual schedule or to determinethe correct charges. The total interchangedeviation from scheduled becomes "inadvertentinterchange," and special arrangements must bemade to pay for the energy. Normally, agree-ments exist between utilities to allow payback ofinadvertent interchange at agreeable times andthus reduce the inadvertent interchange to zero.

Measurement errors and SHAD£.- The accuratemeasurement of power (and frequency) and theaccurate setting of schedules is important;however, errors in measurement and setting doexist [11]. Each utility creates their own special

way of compensating for these errors. One way isto adjust the schedule to account for the errors.This adjustment is called "SHADE."

SHADE is defined as the integral of the errors andis added to the schedule. Thus:

SHADE =OLD SHADE - ADJUSTED ACE

This is calculated every hour or every day as the,system requires. The ADJUSTEDACEis the errorcalculation and is divided into several com-ponents;

ADJUSTED ACE = inadvertent interchange,adjusted time,adjusted frequency,adjusted ramps, andscheduled inadvertentinterchange.

If the components are broken up, SHADEcalculated for 1 hour can be shown to equal thequantities: (SHADE from last hour) - (Meteredenergy sum from all boundary ties over last hour)+ (The scheduled power over the last hour times 1hour) + ~ (Bias setting for frequency bias inmegawatls per 0.1 hertz (Base reference fre-9uency setting for last hour minus 60.00)} +1(Scheduled power at the start of the hourminus scheduled power at the end of the lasthour minus scheduled power at the start of thelast hour plus scheduled power at the end ofthe hour before last)/48f + (Total scheduledinadvertent interchange desired by the dis-patcher).

The dynamic schedule.- The dynamic scheduleconcept helps the smaller utilities which own orshare in ownership of large generators. Thesmaller utility is not required to regulate the largegenerator alone but can share regulation overseveral areas more compatible with the gen-erator size.

Before dynamic scheduling, it was customary todevelop a schedule for 1 hour of operation andleave the schedule constant. Many occasionsrequired schedule changes in the middle of thehour or at other times. A dynamic schedulesystem was developed for use with certain largegenerators having multiple ownership to allowcontinual changes in schedules.

The generator must never be considered jointoperated, but rather joint owned. The operatingutility where the generator is physically located

4

must be in control of the unit. Many varioustechniques are used to generate and account forthe dynamic schedule. One method suggeststhat:

. Each owner calculates (or allocates) thedesired power requirement or schedule forthe owner's share of the generation. Thisdoes not necessarily have to be constant,but may be allocated dynamically. Thisallocation can be identical to the normalallocation of any generation within theowners' system.

. All the owners' power requirements aretransmitted to the operating owner's systemwhere they are summed.

. The operating owner controls the gen-erator power output with a closed loop powercontroller.

. The generator power output is continual-ly divided into the percentages for eachowner. This percentage is determined bythe percentage of the owner's power re-quirement to the total power requirement.

. The owners' share of power output isdynamically (continually) added to theowners' system boundary.

. The energy reading for each owner,except the operating owner, is the owner'sshare of generator power output integratedover each hour. The operating owner sub-tracts the total energy for all other ownersfrom the actual energy reading and uses theresult as his energy usage.

The operating owner must treat all other owners'power output as power out of the system ratherthan input. See figure 4.

Another use for dynamic schedule involves theaccommodation of another utility's loads (orgeneration) completely enclosed by the controlarea. In this situation, there is no way ofpredicting the actual load ahead of time. Thus,there can be no scheduling of the load. If the loadpower requirement is used as a dynamicschedule for the control area surrounding theload, that control area can eliminate power flowto the isolated loads from their interchangereadings. Likewise, the neighboring utility whichowns the load uses the same dynamic scheduleto adjust his interchange readings. The load then

Ijl111

seems to be in the control area of the utilityowning the load. The schedules used forcontracti ng for the transfer of energy throughoutthe control area surrounding the load iscalculated after the fact by averaging the actualload energy used over the hour.

Metering the interties and accuracy.- Themetering of the interties between a system andneighboring systems may be divided into twoareas, the accuracy of the signal and thedetection of errors.

The accuracy of the signal is again of moreconcern to the accounti ng procedures than to thecontrol system. The accuracy of transducersrange from 1 to 0.1 percent error at full scale. Themetering potential and current transformershave an accuracy between 0.1 and 0.3 percenterror at rated voltage or current. With scalingresistors, temperature coefficients, and tele-metry electronics, and accuracy of 0.5 percent ofthe reading is normally obtained at the location ofthe AGC equipment. If the energy (MW'h)monitoring equipment is assumed errorless (as isnormal for accounting), the transducer accuracymay be somewhat improved by a method similarto SHADE. The energy is read every hour and thepower output read from the transducer iscontinuously integrated over the hour. Bycomparing the actual and calculated energyused, an error correction may be applied to thenext hour of power data. Thus:

Calculated power (MW) =Actual power reading (MW)

XEnergy (MW'h) last hour

Calculated energy (MW.h) last hour

This concept works well if the power readings donot cha nge often and are not below 10 percent ofthe transducer full scale. An alternative methoduses the same actual energy value divided by thecalculated energy, and then a table of values isformed for each range of average power readingsover the hour. This table is stored and used tomultiply the actual power reading for eachreading. In other words, a semidynamic calibra-tion curve is used. This concept is rathercomplicated to implement and update, and doesnot work well unless the transducer is frequentlyoperated over its range.

Analog telemetry channels usually involve somedelay or filtering of the power reading. In digital

5

OPERATINGOWNER

PowerOutput ~

0

Figure 4.-Dynamic schedule of a joint-owned generator.

6

systems, the sampling rate forms a delay andoften several computers must transmit the datato the AGC computer, with a delay in the data ateach computer. In the Bureau of Reclamation andWestern Area Power Administration systems,the sampling rate is normally 2 seconds withtransport delays often as long as 6 seconds. Thus,the coherency of the various power readingsbecomes an important question. If the primaryconcept of AGC is not to control intertie swings,then AGC response to signals faster than 0.016Hz ( 1 cycle per minute) is really not necessary. Itshould be realized that with a frequencyresponse of 1 cycle per minute, the transientresponse of the AGC is not delayed by 1 minutebut rather responds slowly at first to an error.There are no intentional delays of the ACE signal.In addition, the governors which the AGCcontrols usually cannot function effectivelyabove this frequency. Thus, filters with timeconstants above 0.16 Hz (10 times the 0.016-Hzfrequency) will cause very little phase shift orgain reduction. Sampling theory indicates thatthe samples should be at least 10 times faster thanthe highest frequency of the controller (0.16 Hz).Thus, a sampling frequency faster than 0.16 Hz(or a sampling period of 6.25 seconds) should beused. It is always best to sample as often aspossible. Randomness in the sampling time of 6seconds can be tolerated (the maximum sam-pling period) without seriously affecting the con-trol. Thus, a sampling period of 2 seconds witha maximum of 4 seconds of random transportdelay would not seriously affect the AGC con-trol. If frequencies above 1 cycle per minuteare required to be controlled, the sampling per-iods and filter frequencies must also be raised.The governors seldom respond faster than ona 10-second time constant which has a filterfrequency of 0.16 second. Many hydroelectricgenerator governors have a 30- to 40-secondti me constant or more with response frequenciesof 0.5 to 0.4 Hz. Normally, good control of thegovernors can be achieved at one-fifth the time-constant frequency or 0.3 to 0.08 Hz (the gainand phase are seriously dropping by this time).This becomes the fastest an AGC can be expectedto control. Fortunately, these fast response timesare not usually required and tuning of the AGC isnot as difficult. Actual AGC systems usually haveequivalent time constants of less than 0.0003 Hzor 0.018 cycle per minute. At 1 cycle per minute,the gain may be down more than 40 decibels.

Since aliasing (or sampling frequencies higherthan one-half the sample frequency) is normallyexpected from intertie oscillations between

LlllIl

systems, an analog filter before the analog todigital converter is absolutely necessary. Thetime constant of the filter should be no higherthan one-forth the sampling frequency or 0.7second for a 2-second sampling period. Goodresults with aliasing reduction utilizes 3-secondfilters with a 2-second sample period. The gain isthen down for the high frequencies withoutundue phase shift at a 1 cycle per minute controlfrequency.

Errors in metering the interties.-A very seriousproblem to the total summation of the intertiepower is the loss of an intertie reading. If thetransducer or telemetering equipment fails, largechanges in ACE will occur falsely causingunnecessary control, inadvertent interchange,and time error. Since most of the failures occur inthe transmission link, the easiest failure to detectis loss of carrier frequencies. Normally, contactsare provided on telemetering or microwaveequipment which close on loss of carrier signals.The AGC equipment can monitor these contactsand should one close, the AGC is tripped or"suspended" until the trouble can be remedied.With digital computers, the last good reading canbe saved and a timer started, but the AGCcontinues to operate. Ifthe timer, usually 20to 30seconds, times out without the carrier signalreturning, then the AGC is suspended. If thecarrier returns a second timer, 10 to 20 secondsmay be used to insure the signal is steady beforethe reading is again updated. Should the secondtimer, or the "thawing" timer, not completelycycle, the first timer is not reset but continues.Thus, bouncing contacts will cause a time-out.Once the AGC is suspended, the operator mustremove the tie from monitoring status on AGCand adjust the schedule, usually with a manualsubstitution interchange value, before allowingthe AGC to become active again.

Failure of a transducer may be detected bymonitoring the rate of change of the meteredpower, and if it is large and the final powerreading is near zero, an alarm may be sounded.Unfortunately, the same effect may be producedby transmission line tripping or intertie oscilla-tions and care must be taken not to "save the oldvalue" unless the transducer is the device knownto have failed.

A help in increasing the security of the intertiereadings is to have alternate transmission paths.These alternate paths are then used when theprimary path has a carrier signal failure.However, the secondary source must have the

7

same error detection as the primary and shouldbe constantly checked to ensure the channel isworking when the secondary channel is needed.

Another source of error of the intertie powerreadings is channel noise. These usually aremost serious in the form of noise spikes causedby electromagnetic interference. One method ofreducing the effect of noise spikes is to digitallyfilter the input reading. Single time constantfilters (recursive filters) with approximately4-second time constants work well in 2-secondsampling systems. Averaging filters which sumthe last one or two readings with the presentreading and divide by the number of readings willalso help with noise spikes. The averagingprocess should not have more than three stagesin order to preserve the phase shift of the signalat a frequency one-tenth the sampling frequency.The AGC control will automatically filter thesenoise spikes because of the integrator (speed-level motor) in the control path. The noise spikesdo increase the activity of the ACE.

If PMSC (programmable master supervisorycontrol) systems are used to obtain the intertiereadings, it is not always possible to determinethe loss of a reading because the reading mayhave been skipped when the equipment mayhave been busy transmitting status events. Amethod of flags and timers is sometimes used toensure that a reading is indeed transmitted.Another method uses a separate square wavevoltage signal applied to an AID input at thesame remote terminal unit metering the intertie.The square wave has a period of 4 to 6 seconds. Ifthis change is not detected by the AGC computer,the signal is assumed to have failed. The bestmethod depends on the communication pro-cedures of the particular PMSC.

The summation of power or the intertie powerdeviation.-When all intertie readings have beentaken, and the schedule entered, includingmanual entry of unmonitored interties, theschedule is subtracted from the total intertiereadings and the "intertie power deviation" isformed. This quantity is used in the formation ofACE and the determination of inadvertentinterchange.

If the total intertie reading is subtracted from thetotal generation of the system, the result is anapproximate value of the system load. If thefrequency is varying, some of the powercalculated as load is actually accelerating power.The tieline losses are also included in the valueof load.

Inadvertent interchange.-Inadvertent inter-change is a concept developed for accountingand is not directly used by the AGC control.Inadvertent interchange (II) is defined as

II = fPsdt - 2; MW.h1

over 1 hour

and is calculated when the energy (MW'h)readings are available, normally once an hour.The Ps is usually calculated from the start of thelast hour to the start of the present hour. Thiscalculation does not include the ramping of theschedule from normally 5 minutes before thehour to 5 minutes after the hour which isreflected in the energy readings. Thus, a rampcorrection must be made to the integral of theschedule.

Integration of intertie power deviation frompower measurements also generates inadvert-ent interchange as a continuous value. Althoughthis value is not based on the energy readings(which are assumed perfect by the accountants),the value can be used to aid the dispatcher indetermining the ability for the AGC systems tomaintain the interchange schedule. It should beremembered that inadvertent interchange is anenergy error between systems and may becaused by many circumstances outside aparticular AGC system. Therefore, inadvertentinterchange cannot be used alone to determinethe quality of AGC control of a particular system.This calculated interchange may be used in acoordinated inadvertent interchange paybackmethod proposed by Mr. Cohn [11]. It also can beused as a component of the integral of ACEdiscussed in succeeding material.

Reducing the inadvertent interchange.-Caremust be taken when reducing the inadvertentinterchange since reduction of inadvertentinterchange for one system may increase theinadvertent interchange in a neighboring system.The payback of inadvertent interchange shouldbe negotiated between two systems so that thepayback will help both systems. This is called"bilateral payback." Ifthis coordinated payback isnot possible, then a "unilateral payback" may bemade but only if time error is decreased. Anotherconcept would be to institute a "synchronized" orcoordinated payback system using time error aswell as IIwhere all members of a large power poolwould payback II according to a signal from a"master utility." [11] This does require that allutilities use the same metering methods and thesame methods for calculating II.

8

Frequency and Time Signals

The ACE is formed by two types of signals - theinterchange deviation power and the frequencybias. If only interchange power is used, thecontrol is considered "flat tieline" control. Theutility must then depend on other utilities forfrequency and time correction and the utility willunnecessarily participate in disturbances withinneighboring utilities. Thus, this control is usedonly when frequency is temporarily not available.If only frequency bias is used, the control isknown as "flat frequency controL" This controlforces a utility to provide power for alldisturbances on the entire system and does notallow for reasonable accounting procedures. Flatfrequency control is seldom used. "Tieline biascontrol" exists when both frequency bias andtieline deviation is used and is the normal,recommended method of control. Each utilitythen shares equitably in a system disturbance.Western utilities also add time deviation signalsto automatically correct time. Eastern utilitiescorrect time using the frequency reference offsetmethod. Western utilities will also correct largetime errors with the frequency reference offsetmethod.

Frequency signal accuracy.- The frequencysignal should have accuracy to 0.001 Hz or betterand have near perfect long term stability.Fortunately the frequency signal need not be fastsince the AGC need not attempt to correct intertieoscillations. Therefore, a filtering system or amethod of measurement can have as much delayas a simple filter with a 3-second time constant.The phase lag should be maintained as low aspossible in the 1 cycle per minute range. Also,since the AGC is to ignore intertie swings,frequency is assumed the same over the entiresystem and the frequency is monitored at thenearest power system bus to the AGC equip-ment. Failure of the signal can usually bedetected by the signal going out of limits. Theselimits are usually set at 59.8 and 60.2 Hz. If thesystem frequency is outside these limits for anyreason, it is customary to suspend AGCoperation. The power system is in serious troubleif frequencies of these magnitudes are ex-perienced, and the AGC can do little to help. Ifthefrequency is telemetered, the loss of carrier mustsuspend control. A similar timing system to theinterchange power monitoring system may beused and alternative sources of frequency shouldbe provided.

Frequency bias.- The setting of the frequencybias is a very controversial concept. Figure 1

Ijlllll

shows that the frequency signal is essentially theintegral of the accelerating power. Because theaccelerating power is the variable to becontrolled, the integral of accelerating power canbe used as a valuable signal for the controller.However, the frequency is the integral ofaccelerating power times Tm-1 where Tm is themechanical time constant of the system withinthe AGC control boundaries including the inertiaof the lo!,!ds. To obtain the pure integral ofaccelerating power when frequency is summedinto the ACE signal, the frequency bias, 10B,should be equal to Tm, where

MWSTm!::::!2~ rated

L.J MVArated

allequipment

where the MWS is the inertia constant for theequipment in megawatt seconds and MVA is themegavolt-ampere rating of each piece ofequipment. Because this quantity is continuallychanging as different generators and differentloads are connected to the system, the biasshould be continually changing. If the AGCcontrol were required to perform excellentcontrol for tieline oscillations, then the biasshould theoretically be less than this value(usually indicated as (3 in many references).However, the AGC is not normally used for tielineoscillations and a bias equal to or greater than theTm or B is usually recommended. The method ofdetermining the bias, 10B, is described in theNAPSIC manual [9].

In practice, the power system is far toocomplicated to adequately measure Tm of aspecific control area, and the bias is based on thecontrol area peak generation during the year. Thevalue of B may be 1 to 2 percent of the peakgeneration.

Frequency Schedules and Use in Time Correc-tion.-Since the inability to maintain the intertieschedule results in movement of acceleratingpower, a frequency error is introduced. Thefrequency error then results in a time error thatcan be seen on electric clocks connected to thesystem. Therefore, the time is normally moni-tored by one utility in the system and compared toan accurate time standard such as WWV orWWVB radio stations. When ti me corrections areneeded (when the error is 3 to 6 seconds) an

9

order is sent from the "time" utility to all otherutilities requesting a time correction. All theutilities change their frequency schedule (usuallyby 0.02 Hz) for a period of several hours. This isthe same as changing the schedule interchangeif the bias, 10B, is included; however, it is moreconvenient to change the frequency referencethe same for all AGC controls working within theoverall system.

The time error may not be corrected during peaksystem loads in some situations because theadded energy for time correction utilizes highercost fuel. The time may be "overcorrected" oradvanced using lower cost fuel in the earlymornings in anticipation of the heavy systemloads.

Time Deviation Signals. -It is now possible for allutilities within the overall system to monitoraccurate time from WWV at a reasonable cost.If all the utilities monitor time and bias ACE fortime correction, then time can be continuouslycorrected without resulting inadvertent inter-change (from time correction). This is the verysame concept as changing the frequencyschedule except the schedule is changedcontinuously and automatically. To be success-ful, all utilit.ies must use the same amount ofoffset for the' same amount of time deviation.This quantity is usually called "sensitivity" ofthe time deviation correction system. In theWestern systems, a sensitivity of 0.02 hertz persecond has been used. In reference 11, a systemis presented that uses a signal from the "time-keeping" utility to all utilities to synchronizeboth time and inadvertent interchange. Un-fortunately, all utilities must participate to allowaccurate corrections.

Usually the time deviation signal is limited to apreset value so the ACE is not greatly biasedfor large time errors. For minimum inadvertentinterchange, all utilities must use the same limitvalue as they must use the same sensitivity. Thetime deviation error must not have much gain (ortime must not be corrected quickly) since the timedeviation is the second integral of acceleratingpower control.

The Final Form of ACE.-The final formation ofACE is shown in figure 5. The basic equation is:

ACE = (PA - PsI - 10B { (FA - FS) + KT(TA -T s)}

and PA = (2; Pm) + Pnm

where Pm is the measured tieline power flows inmegawatts and is positive for power flow out ofthe control area. The Pnm is an entry from thedispatcher to account for interties not measuredbecause of equipment failure or other technicalor economic considerations. The scheduledpower Ps = (~ Pd + PDY + SHADE where Pc is thecontracted power sold or bought betweenutilities, PDY is the dynamic schedule fromjoint-operated generators or loads, and SHADE isthe measurement error correction factor. Thevalues of Pc, PDY, and SHADE are in megawattsand are positive for power or energy leaving thesystem (or energy sold). The frequency bias, B, isthe bias value which converts a frequencydeviation into an equivalent power change; it is inmegawatts per 0.1 hertz and it is always anegative number. The measured or actual systemfrequency, FA, is measured on the power systemnear the AGC equipment and is in hertz. Theschedule frequency is the frequency desired bythe dispatcher; it may be different than 60 hertzfor time error correction. The constant KT is thetime sensitivity factor and is related to the totaltime desired for correction. The units are hertzper second. All utilities with automatic timedeviation correction must use the same KT'Sometimes the time sensitivity is reported as1OBKT and is given in megawatts per second. Thetime, TA, is measured by integrating systemfrequency and has the units of seconds. The time,T

s'is the standard time from WWV.

AREA CONTROL ERROR DISPOSI-TION

Once the ACE has been formed, the signal mustbe routed to generator governors so that thegeneration of the system can be changed. Theprincipal function of this part of the AGC controlis to establish control loop gain. Before the ACE isformed, the control gain could not be easilyimplemented since all signals contributing toACE would have to be multiplied by the samefactor. After ACE is formed, a multiplying factorcan be applied to a single signal. After the ACE isformed, the gain should be as high as possiblewithout instability. There are two major concerns.First, the gain of the speed-level motors differfrom unit to unit and more than one SLM(speed-level motor) may respond. There is alwaysthe problem of deadband and hysteresis in theSLMS and governors. Second, the control mustactivate the SLM, governor, and prime movercontrols such as throttles or wicket gates. Thesedevices are mechanical in nature and wear outwith constant use. Unnecessary movement of

10

Ij

IIj~I

StandardTime fromWWV

Is

MeasuredTime +TA

TimeError Time Sensitivity

ErrorLimit Bias B is

lOB Negative

Measured Schedu eFrequency from

Operator

Ps

DynamicSchedule

Intert~angeDeviation

Freq uen cyFSchedule s

fromOperator

Frequency BiasDeviati n

lOB

B is Negative

ShadePA

Power NotMeasured fromOperato r +

++ of

+ +

,v

/

Mea suredInterchangePower

Figure 5.-Components of ACE.

11

these devices creates additional maintenancecosts. Thus, a strong incentive is present toreduce the number of movements or controloperations if such a reduction does not greatlyincrease the inadvertent interchange. Thisreduction of control activity is implemented bymodifying the ACE and by the design of theallocator.

Modifying the ACE

Modification of ACE is usually an attempt toreduce the random noise inherent in the ACEsignal. The power system load is constantlychanging in a somewhat random manner and thepower system often restores itself by randomdisturbances which cancel the previous change.Thus, the accelerating power is constantly inmotion and, for a large percentage of the time, isconstantly fluctuating about zero. The AGCcontrol cannot remove this "noise" from thepower system with the slow speed-level motorsand governors. Further, to control such self-correcting variations causes unnecessary wearand tear on the mechanical parts. Therefore, twocriteria are usually used to decide when ACE isvalid. First, slow drifts within the ACE must becorrected to minimize the inadvertent inter-change and the time error. These slow drifts, dueto slow load changes, are of the range of 5- or1O-minute duration or more. Second, large rapidchanges of ACE indicate the loss of generation orload. A reasonably rapid response of the AGC willallow a control area to rebalance generation toload without unnecessary or lengthy assistancefrom neighboring utilities. It should be noted thatmany AGC systems do not modify ACE and dowork satisfactorily. However, the use of someACE modifications has significantly reducedactivity of certain AGC systems.

Limiting and Deadband. -The most si mplemodification to ACE is to provide a deadbandsized to remove some of the random "noise." Thedeadband works well to reduce activity and toallow rapid recovery of system disturbances.However, the deadband may mask a small, slowchange or offset in the ACE. Another modifica-tion to ACE is to provide a limit. The limit isespecially useful in systems where a larger ACEcould not influence more generation anywayand, thus, the dynamic range of the ACE is notexcessive for the hardware. Often, ACE levelscause a switch of gain on the ACE signal bychanging the type of allocator used. Such aswitch may be called "emergency assist" mode.

It is generally acknowledged that deadbands andlimits can be applied elsewhere in the controlsystem with perhaps better results and that littleis accomplished by modifying ACE itself.

Smoothing Functions.-Smoothing filters can beapplied to modify ACE. These filters are normallysimple, single-pole time constant filters whichwork to reduce the gain of the "noise" and allowthe slow changes to have full gain. A timeconstant of 120 seconds has been used to reducesystem "noise." [33] This filter does reduce"noise" and allow correction of slow changes,but the rapid, large system disturbances are notresponded to.

A possible remedy for the rapid, large systemdisturbances is to switch to an "emergencyassist" allocator using unmodified ACE when theunmodified ACE reaches a present level. Thefiltered ACE is then used when the ACE returnsto the normal operating levels.

The Probability Filter.-Many AGC systems haveadopted some form of a probability filter. Thebasic filter design was presented by Dr. Ross [33]using the designation "Error Adaptive ControlComputer." The adoption of the filter toBonneville Power Administration is reported byTaylor and Cresap [42].

The description of the filter is not overlycomplicated, but the description of the adjust-ment of the filter parameters is very difficult. Theadjustment depends on the operating character-istics of the power system which are difficult tomeasure and interpret. The following paragraphsattempt to describe the needs for the filter, thebasic design of the filter, and a basic concept offilter adjustment.

The "Error Adaptive Control Computer" des-cribed by Ross is based on the need fordiscrimination of three different types of signalsfound in power system control. These are"deterministic," "probabilistic," and "sustained"error signals.

. Deterministic error signals have character-istics which are known although their occur-rence may be random. These deterministicerrors may be caused by intertie swings, periodicload fluctuation (i.e., rolling mills or arcfurnaces)or large, sudden changes in load or generation.

. Probabilisticerror signals are random in naturebut may have some definable amplitude level and

12

frequency band. These errors may be caused bythe normal, random changing of load on thepower system.

. Sustained error signals are usually caused bythe control system errors from either man orequipment. These errors develop into inadvertentinterchange and/or time deviation.

The filter attempts to discriminate the signals ofeach type into errors which need correction (ormore accurately, which will respond to cor-rection and benefit the system in an economicalsense) and errors which can be safely ignored.The result of the discrimination process is an"amplitude modulation" of the ACE. The ampli-tude modulation is usually in discrete steps,such as multiplying the ACE by zero, one, ortwo. The ACE is never "phase shifted," becausean amplitude modulation is used, however, theerror is used to control governors and powersystem inertia, and the time constants of thesedevices use the overall effect of the modulatedwave and effective phase shifting does occur.The process is very similar to the firing ofSCR's (silicon controlled rectifiers) into an in-ductive load. The individual pulses of voltageare smoothed into a current and the overalleffect is very nearly the same as if th~ voltagewere amplitude controlled and phase shifted(instead of being switched).

As described, the output of the filter is switchedbetween two or three gains. This switchingprocess is basically caused by the ACE crossingzero, and the weighted ACE (a constant timesACE) plus the integral of the ACE exceeding apreset value for a preset time. Usually severalfilter elements are used in a specific filterapplication.

The error signal crossing zero.-When the ACEcrosses zero, an assumption is made that thepower system is restoring itself and no con-trol action is required. Each time the ACEcrosses zero, the gain applied to the ACE isset to zero (this is the concept described byTaylor [42]). The gain remains at zero until atimer times out or ACE again crosses zero. Thisprovides a very effective "filter" of all fre-quencies having periods of less than twice theti mer setting. Taylor sets the timer to 3 secondsand all frequencies above approximately 0.16Hz receives no control. It also should be notedthat this type of zero crossing filter will allowslow or sustained ACE of about one-half therandom noise signal amplitude to continuewithout control action.

IJHI

Ross implies the opposite concept. When thezero crossing occurs, a separate timer is resetto zero, and if no other filter elements areactive, the gain is set to zero. This does notnecessarily filter the high frequencies becauseother filters may be active but does providestrong control of "sustained" ACE. Ross alsosuggests two different timers to provide evenhigher gain for "sustained" ACE that do notseem to respond to control.

The weighted error signal and the integral ofthe error signal exceeding a preset limit for apreset length of time.- This filter elementworks on the concept that if the ACE and theintegral of ACE are not sustained for a givenlength of time, control is "probably" not re-quired. The difficulty lies in the word "probably."To one observer, the control is too active, andto another, the same control is very quiet. Themeasurement of "probably" is usually controlactivity. There is further discussion of the"activity" in succeeding material.

Both Ross and Taylor describe a probabilityfilter with two of these filter elements. How-ever, the method of determining the integralused by these elements differs. Ross indicatesthat the integrator should be periodically resetto the inadvertent interchange and time devia-tion as calculated from sources other than ACE(such as watthour totalizers, etc.). This impliesthat the integral stage can have a substantial,sustained, but limited value. Taylor uses anintegral with very low limits and thus providesa hysteresis effect on the weighted ACE.

The weighted ACE signal is essentially anartificial error energy signal in the samemanner that the integral of ACE is the actualerror energy signal. (Another way to look at thesignals is to realize that a power error, ACE,applied to the system inertia procedures fre-quency error.) The artificial error energyindicates what energy is in the ACE signal if ithad continued at the present level for a preset(or weighted) time. This can be considered amethod of predicting for a short period of timewhat the error energy will be. The sum of theactual error energy and the "predicted" orartificial error energy give a total error energyor a total expected energy that indicates howserious the error is. If the error energy exceedsa preset level, the error is serious and mayrequire control. A timer is included to ensurethat the serious error energy remains seriouslong enough to "probably" require control (orthat the expected time error is enough to"probably" require control).

13

If the error energy builds to a high level in ashort ti me, the error is considered serious. But,if the error energy builds to a medium level fora longer period of time, the error is moreserious and the gain is then doubled. This longduration may indicate that the error is notresponding to control action. The result of thefilter element is that a certain area of frequen-cies include the higher frequencies. Also, adeadband effect occurs because the filter willnot respond to low level signals.

It is important as noted by Ross thatthe integraloutput does not ever exceed the preset levelfor switching gain of the sum of the integralplus the weighted ACE. Otherwise, the ACEwould become active (have a gain other thanzero) when excessive inadvertent interchangeis present and the control would become ex-tremely active (but miscontrol would notresult). For this reason, a hard limit on integralof ACE is required.

There are numerous ways to adjust the probabil-ity filter elements. The adjustment is primarilybased on activity and speed of system recovery;neither of these concepts are well defined. Sys-tems have operated (and are still operating)without the filter; however, on these systems,there have been some complaints about thecontrol being too active or not fast enough. It isknown that if activity is reduced, the speed ofrecovery or response can be increased. Becauseincreased response usually implies lower levelsof inadvertent interchange, speed of recoverymay become important. The response speed alsodepends on otherfactors such as the number andsize of the generators responding and the re-sponse of the powerplant controllers, and be-cause the inadvertent interchange depends onmany different parameters, the only quicklyobservable characteristics of an AGC system isactivity. Therefore, a "feel" for systems activity isusually developed by anyone who has respon-sibility for an AGC system. Activity can usually bementally equated to "wear and tear" ofequipment. The speed-level motor may wear outfaster when the AGC is active, but it is debatablewhether the governor suffers from added "wearand tear" since there is always a little governoractivity due to the power system frequency.Because inadvertent interchange is more "im-portant" than speed-level motors, the normalassociation should be "less activity to allow formore response to reduce inadvertent inter-change." Unfortunately, as the activity de-creases, the ability to respond to certain types of

errors decreases a nd the dispatcher becomesaware of circumstances when the control is tooquiet. Reasonable settings for the filter elementsmust be determined despite the conflictingcondition.

One method for adjusting the filter is toprecalculate the settings based on some simpledata measurements and then tune the systemwhile in service. Ross discusses this approach.The "optimum" tuning can be done by a veryknowledgeable installer who fully understandsthe filter. Unfortunately, few people fullyunderstand the filter and often less knowledge-able people will later alter the tuning to suit their"feel" of the system.

A second method of adjustment is by simulationas described by Taylor. This method usuallyallows a parameter study to be made so that thechange in operation from the change of anyparameter can be understood. Usually, most ofthe parameters are then defined as "best," andone or two parameters are left for the tuningprocess during normal operation. System simu-lation is not an easy task and very oftenparameters found from simulation neglect somecharacteristics of the system which the filtermust either accept or reject.

Biasing the ACE with the Integral of ACE.-SomeAGC systems use the integral of ACE to bias theACE; this helps correct slow drifts in ACE. Theprincipal effect is the reduction of oppositepolarity pulses to the generators resulting inreduced activity. As an example, during anextended loading period (such as the morningloading), the integral of ACE will restrict "lower"generator pulses and favor "raise" pulses. Theactivity is reduced because a lower pulse wouldrequire a raise pulse in a short time. Improvedresults with the integral of ACE can be obtainedby also using a filtered ACE (as previouslydescribed). The reduction in activity will bebetween the filtered ACE alone and a well tunedprobability filter.

The Allocator

The a lIocator is the most controversia I andcomplex part of the AGC system. The controlsystem requires that the ACE is delivered to agenerator to alter the generation and thus theaccelerating power. The principal question iswhich generator must be controlled. The first areaof concern is the range between "permissive"

14

and "mandatory" control concepts. These con-cepts are based on the requirement for base loadsetpoints.

Base/oad Setpoints.- The generators used in apower system have many constraints. Somegenerators may be required to run at a fixed loadfor mechanical reasons. Other generators haveauxiliary equipment which runs best betweentwo specific (and usually close together) loadlevels. Other generators may have reservoir levelor river flow constraints. Thus, when thedispatcher selects generators to supply the load,many generators must be maintained at abasepoint generation level.

If the basepoints are allowed to change, often themovement of the generator to the new basepointmust be done as quickly as practical to avoidturbine rough zones or switching of auxiliarysystems. Thus, some generators must be rampedto the new basepoint without regard to theimmediate need of the power on the system.

The same concepts of baseloading and rampingalso occur for entire powerplants. Hydroelectricinstallations may be required to pass largevolumes of water or must be run to maintain acertain river flow, and the requirements of theimmediate system are less important.

Permissive or Mandatory Contro/.-A permissiveallocator can be defined as an alloca'tor whichnever causes a control output that either perturbsor conflicts with ACE (i.e., the immediate systemenergy balance is most important). This is thebest controller from the control theory stand-point. The ACE is always in control and nounnecessary control actions occur. The probabil-ity fiIter used to modify ACE works especia lIywellon a permissive control since the closed filter (orzero ACE) will not cause any control which willattempt to open or activate the filter. All filteractivations are caused by the power system loadsand frequency shifts.

A mandatory allocator attempts to hold all plantsat a basepoint. The plants move their power inresponse to the ACE but when the ACE is zero,the plants return toward the basepoint. Thissystem works well when generators have severeconstraints. The ACE is always perturbed as soonas it is zero unless the unlikely situation occurswhere all the plant basepoints match therequired system generation.

Actual allocators of the various AGC controllersusually use a combination of permissive and

mandatory operation. Some examples include (1 )placing baseload on mandatory control withbasepoints and the AGC generators on permis-sive control without basepoints; (2) executing allbaseload changes with all generators in man-datory and then operating the rest of the time inpermissive mode; and (3) not allowing baseloadchanges to oppose the ACE (but ifACE is zero, themandatory movement is made). There arenumerous variations. However, the basic con-cept is that the permissive control provides goodcontrol with m in imum activity a nd the ma ndatorycontrol provides good capability to honorgenerator or powerplant constraints.

Participation Factors.-Participation factors arethe crux of the allocating system. Unfortunately,they are also difficult concepts to define. Thenormal definition of participation factor is thefractional part of the ACE that is allocated to aspecific generator or plant. In most systems, theparticipation factor is only a small part of thedetermination of the amount of ACE that isallocated to a plant. A few examples (not anexhaustive summary) of actual systems willdemonstrate the various uses of participationfactors.

Permissive control with plant and arearequirements.-Figure 6 shows the basicblock diagram for the allocator using stationrequirement, area regulation, and variable-gain, permissive control. The effect of variablegain is present because the switch to sendACE to a plant may be closed for more than oneplant at a time and there are a variable numberof plants or generators responding to the ACE.The permissive concept is that only ACEsignals are sent to the plant, and even thoughthe "station requirement" has a value, nocontrol is present until an ACE is present.The AR (area regulation) is:

N

AR =ACE +~ (BGi - PGi)

i=1

where PGi is the monitored generator powerfor plant i and BGi is the basepoint setting thatthe dispatcher has selected for plant i. Thetotal number of plants in the control area isN. The station requirement (or plant require-ment) is

N

PR =PG - BG + p[ ACE + 2.: (BGi -PGi)]i=1

15

ACE

Sum ofPlantErrors

Bose SettingfromOperator

Plant

Error+

ActualGenerationfrom Plant

+

ACE toPlant

ParticipationFoetor fromOperator

Close OnlyWhen ACEand StationRequirementsare SomeSign

StationRequirement

TYPICAL PLANTFigure 6.-Example allocator with variable-gain, permissive control with area requirement and

plant requirement.

where PG is the monitored plant generation,BG is the basepoint for the plant and P is theparticipation factor selected by the dispatcher.The sum of all the participation factors for allthe plants should be 100 percent.

For control analysis, the "participation" of theplant is not the participation factor alone butthe fact that ACE and PR are the same sign.The participation factor actually acts as a biasto determine which plant will be favored toreceive ACE. However, a plant may receive ACEeven if the participation factor is zero. Note thatthe difference between plants on "automatic"and plants on "baseload" is that the plants onbaseload havea zero participation factor. Also,positive ACE, AR, and PR indicate the genera-tion should be reduced.

Mandatory control with plant requirementsand area regulations.-Figure 7 shows thebasic block diagram for the allocator. Theconcept looks very much the same as figure 6as far as the form of the equations. However,a more careful search' will show that ACE isnot ever sent directly to the plant. Rather theplant requirement is sent to the plant and ACEis only one of the components.

The plant requirement is:

N

PR = PG - BG + P [ACE + L (BGi - PGi) ]i=1

where the variables are the same as in theprevious example. If all the plant requirementsof this example are summed together,

N N N

LPR =L: PGi - L:BGi

+ti;t ACE~=t (B~:1-PGk)]

;=1 N k=1

where L P is always 100 percent. Thus

i=1N

LPR = ACEi=1

and the system is "mandatory" in the formthat control signals will be sent even if ACE iszero, but since the sum of all control signals

16

IJI111

ACE+

BaseSettinfromOperator

PlantError

+Actual

fromGenerationPlant

PParticipation

Factor

P= INumber of

Plants onControlSum ofPIant Errors

( PIantControlError)

TYPICAL PLANT

Figure 7.-Example allocator with mandatory control with area requirement and plantrequirement.

is ACE, the system is somewhat like "per-missive" control. In theory, the control is verygood. In practice, the response of the plantsto the control signals will not be proportionalto ACE because each plant may have a dif-ferent gain to the plant requirement. Theconcept will work but perturbations of ACEwill continually occur as the plants seek abalance point among themselves. This controlexibits mo.r:eactivity than the purely permissivecontrols.

In this example, the baseload plants simply havezero participation factor (in practice they alsohave a frequency regulation term while inbaseload). Also, the participation factor issimply 100 percent divided by the number ofplants on automatic control. This implies thatall plants are approximately the same size andwill respond nearly the same for a plant re-quirement signal.

This type of control can be made more permis-sive by limiting the plant requirement. When

this is done, the concept of the total PR beingequal to ACE is lost. Ways of providing morepermissive control are:

. Limit the PR if it opposes the ACE. If ACE iszero, or not in opposition, the PR is not limited.

. Set PR to zero if the PR opposes the ACE. IfACE is zero or not in opposition, send the PRo

. Set the PR to zero if the ACE is zero oropposes the PRo If ACE does not oppose thePR, send the PRo

. Set the participation factor to zero and PR tozero if ACE is zero or opposes the PRoThen theparticipation factor is 100 percent divided bythe number of plants on control and have a PRthat aids ACE.

The progression from the first to the thirdconcept tends to reduce activity but also allowsthe gain on ACE to become more variable. The

17

last concept is nearly the permissive con-trol of the first example except the gain is moreconsta nt. The mandatory changes are removedand a plant will respond only to signals inphase with the ACE.

Permissive control with participation factorsbased on plant limits.-Figure 8 shows thesignals used in a permissive automatic controlwith participation factors based on plant limits.It is noteworthy that the basepoint allocator ismandatory control (with a frequency regula-tion term). This system has basically the sameequations as the first and second example. Amajor difference is that baseload plant errorsare not summed into the area regulation butrather are used to directly bias the ACE forra mp cha nges. Seca use the participationfactors are dynamic (they change with gen-erator loading), the allocator can be reducedto a "floating" basepoint controller where nobase setting is required for plant on automaticcontrol.

If N

PR =PG - Ps + P ~ (PSi - PGi) + ACE

i=1

in the sum of all the participation factors insubsection 1, and if PSi - PGi are summed onlyfor units on control, then

N

~ PRi = ACE

i=1

However, if the basepoint (PS) is eliminated,the same effect can be obtained from

PR =P(ACE) andN

L: PRi =ACE

Ii=1N

because L P = 100 percent

i=1

The floating basepoint is superior to the fixedbasepoint system in practice although theyare the same theoretically. The fixed basepointwill cause ACE perturbation because the gainsof the various plants are not equal. Whereas,the "floating" basepoint concept does notcause such perturbation.

The participation factor itself is calculatedusing the polarity of ACE. If ACE is positive. the

generation is excessive and must be reduced.The actual plant power is subtracted from thelower power limit of the plant (determined bythe dispatcher) and divided by the summationof all the differences of power and lower limit.Thus, for positive ACE,

PG - PLLPi = N

L:IPGi - PLLiI

i=1

where PLL is the lower power limit of the plant.For negative ACE,

PG - PULp. =I N

~IPGi - PULiI1=1

Where PUL is the upper power limit of theplant. This procedure works well since theplant farthest away from the limit will have thehighest participation and a plant at the limitwill have no participation. Two problem areasexist however. If the PG - PLL or PG - PULterms become negative. the control wouldreverse the plant requirement and the controlwould become mandatory. Thus. negativeterms in the numerator of Pi must be set tozero. Also, the absolute value is not necessarysince the negative terms in the summationshould also be set to zero before summing.The second problem occurs when all plants areat or beyond the limit. Then the denominatorbecomes zero and Pi must be limited to anumber below overflow within the computer.

The actual participation is:

Pi

~ PiP=

This relationship provides that the sum of P willbe one. However, as the sum of Pi becomeszero because all plants are at zero, a zero overzero condition occurs which should be avoided.This relationship is not necessary if negativeterms in the numerator and within the absolutevalue of the denominator are set to zero. Thepossibility of having no participation is presentand should be brought to the operators atten-tion.

An additional approach of changing all plantlimits to emergency limits when all participa-tion factors are zero would provide automaticemergency allocation.

18

ACE

Base SettingfromOperator

PG

+Actual

Generationfrom Plant

IJIJ.!!I

p.L

LP'ParticipationFactor

Pi = Px- PG

L I Px - PGI

Sum of PIants

on Control

PlantError

+PlantRequirement+

TYPICAL PLANT

Figure B.-Example allocator with permissive control and participation factors based on plantlimits. (Px =Upper plant limit for - ACE and lower plant limit for + ACE. Limits are entered byoperator. PG =Actual generation from plant.)

The mandatory baseload system is not reallycompatible with the above concept. An attemptto overcome baseload changes by biasing ACEwith baseload changes helps overcome someof the problem, but still causes perturbationsto ACE which the floating base point systemfor automatic generation attempted to elimin-ate. If a plant is required to have specificramp rates and mandatory settings, the planterror (i.e., PG - PS) should be subtracted fromACE rather than the change of Ps only, andACE will perturb less (theoretically, not at all).However, if the loading of a plant can be per-missive, the upper and lower limit could be setto the desired basepoint. Then loading wouldbe done according to the needs of ACE. Alsochanging limits in an emercency would auto-matically include the plant baseload inthe allocator.

There are many forms of participation factorsbesides the "coefficient-of-ACE" factor de-scribed in the preceeding three examples.Circular allocators will load plants according tothe size of the plant similar to the third example,

except small allocations during each pass (whichmay be within a deadband in the generatorcontroller) are avoided. Therefore, a response tothe loading signal is usually more precise, thegain of the system is held more constant, and thebaseload units can be used if the rate of systemresponse becomes too low.

Participation factors are also used in economicdispatch. The concept of economic dispatch isoutside the scope of this report.

Emergency Assist and Standby Generation.-When the ACE becomes very large, most AGCsystems have provision for providing very highgain and, thus, quick generation response.Usually, the resulting gain is actually larger thannormal control would allow. If the gain is activefor more than the minimum time to reverse theACE trend, the AGC will tend to become unstable.The emergency assist may take many forms.

. All plants connected to the AGC systemincluding baseload plants are sent the ACE whenthe ACE exceeds a present limit. There is no

19

participation and all plants receive the ACE, thusmultiplying the effect of the ACE many times.When the ACE recrosses a preset value (a littlehysteresis is used), normal allocation proceduresare then used. On special occasions, the ACEmay oscillate several times but will stabilize.

. A baseload assist system is used when the ACEbecomes large or when the ACE cannot beallocated to control plants (because of Iimits). Thegain is changed by providing a second allocator toallocate to the baseload plants in the samemanner that the main allocator used. Therefore,the gain is approximately doubled (unless themain allocator cannot allocate due to limits). Ifthe ACE is large, plants on "standby" areincluded in the allocation. These "standby"plants have their generation controlled at theplant. and in "emergency" conditions, the ACE isused by the plant.

The emergency assist is actually an attempt toreduce the system droop. The speed-level motorsmove the speed reference of the governor so thatmore generation is changed than would havebeen possible with the droop alone. Care must betaken not to cause a larger disturbance than theoriginal disturbance.

Signal to the Plant.-Once the allocator hascalculated the plant requirement. the signal mustbe transmitted to the plant. This communicationcan take place in several ways.

Pulse systems.- The size of the ACE isperiodically converted to pulse width modula-tion and sent to the plant. There is a definiteperiod to the pulse (2 to 6 seconds) and thewidth of the pulse hasa maximum value. Sincethe speed-level motor will integrate thesepulses (after being allocated to the generatorsat the plant), the system is considered a "rate-limited" control system. The generation canmove only at a certain maximum rate limitedby the maximum speed of the motor and themaximum pulse width. Thus, the gain changeduring emergency assists actually increasesthe rate limit by using more than one plant.

The plant is in a "rate-limited" power feedbackloop if measured plant power is used in theAGC allocator. Since the rate-limited controlautomatically reduces gain when the speed ofthe disturbance is rapid, system damping is notseriously impaired. Further power-rate feed-back is usually not provided and the controlgain around the plant power control isusually low.

Pia nt control error. -The pia nt control erroris sent directly to the plant by digital or analogtransmission. This system provides an ac-curate power controller around the plant andhas a very definite maximum gain for the SLMintegrator. If the gain becomes too high,oscillations develop. One method of controllingthe gain is to allocate the error at the plant inthe form of pulses to each generator using asophisticated plant allocator. This controlsystem must also be rate limited to ensurethat system damping is not impaired.

Power setpoint signa I.-Another method forcontrolling the plant is the use of a power set-point transmitted either with analog or digitalequipment. The allocated error must be inte-grated in the AGC controller (replacing theSLM integration). Then the powerplant mayuse an allocator and closed-loop power con-troller. Th is places the closed-loop controller atthe individual generator and also allows theuse of local constraints (breakers, gate limit,rough zone, efficiency, temperature, etc.) tomodify the individual controllers. Usually arate-of-power or a predictor closed-loop con-troller is used [35, 65]. The rate-of-power con-troller usually has a rate-limit or loading rateincluded. The predictor controller does notrequire a rate-limit system and may respond atnearly the maximum linear rate of the governor(if the windings, fuel handling equipment, orriver level can tolerate the loading speeds).This speed of response is not usually needed byby the AGC, however.

Plant Regulation Included in the Signal.-Since all AGC allocators, with the possibleexception of the floating basepoint concept, use aform of closed-loop power control. the long termeffect of the droop of the governor (or the speedregulation of the governor) will be negated. Thebasic concept of the AGC system is to use thedroop characteristic of the system as a basicassumption and reregulate the system aroundthe droop characteristic. The bias setting on thefrequency component of ACE includes thesystem droop characteristic. Thus, removal ofthedroop by a constant power controller opposes thebasic AGC concepts and opposes the naturalsystem response.

It is preferable to have every generator controllerwhich is not being directly influenced by ACE tohave some type of frequency bias allowing thedroop to work. Even plants directly affected byACE should have some type of frequency bias to

20

aid in system load stability although the ACE mayovercome the frequency bias over the long term.This concept becomes more important as thestrength of the closed-loop power controlleraround the plant increases.

Another problem in plant control is the governorwith the dashpot bypassed, or governors thathave been tuned to give very quick responses.These governors usually depend on the powersystem for control loop damping and they will notoperate with an isolated load. The power systemcan support many governors of this nature butthere is a maximum limit. Unfortunately, whenthe limit is reached, system oscillations maybegin without any obvious cause or warning.Thus, the number of "fast" governors should bekept to a minimum.

Calculations Useful for Plant Allocation.-Several calculations and tabulations are usefulfor operation of the AGC and, in particular, fordetermining the capability of the syst ~m in theevent of a disturbance.

Plant response errors.-A powerplant may notrespond to AGC commands for a variety ofreasons including all generators at limits orloss of communications. This condition may bedetected by finding the plant power outside adeadband from the base generation plusplant requirement. Another method is tomonitor the rate of change of plant power todetermine if the direction of the plant require-ment and plant power correspond. Once theslow response has been detected, a signalmay be sent to the plant to determine if anyresponse is possible. If no response is foundthe plant should be removed from allocationand an alarm generated.

Disturbances outside the AGC system.-Often, severe disturbances outside the AGCarea may require action by the dispatcher orsystem operator. Such disturbances can bealarmed to the operator by using the tielinepower error and the frequency bias termsusually added for ACE. However, the two aresubtracted to give the "ACE" for the neighboring systems. If the disturbance is within thelocal system, the ACE will become large. Whenthe disturbance is outside the system, the ACEshould remain small, but the "reverse ACE"will become large. If the equivalent externalsystem bias can be found, the signal may con-tain the actual disturbance size information.

The regulation margins.- The difference be-tween the present generation and the maxi-mum capability (or limit of generation) for eachplant on automatic should be calculated. Also,the difference between the minimum plantpower limit and the present power output forall plants on automatic should be calculated.These two calculations will provide alarmswhen plants can be put on or taken off auto-matic to ensure an adequate operating rangefor the AGC system.

Spinning reserve.- The spinning reserve isused as an indicator of system reliability andis usually the difference between the totalmaximum limit of generation for every gen-erator on-line and the present power outputof every generator on line.

QUALITY OF THE AGC

The NAPSIC operating guide [9] defines a mini-mum operating criteria. This criteria is sum-marized in table 1.

Table 1.-AGC minimum operating criteria

For normal system conditions:

. Area control error (ACE)must equal zeroat least one time in all10-minute periods.

. The average deviation of ACEfrom zerofor all 10-minute periods must be within aspecified limit, Ld' based on system genera-tion. The value of Ld can be found fromcalculations in the NAPSIC manual and isbased on maximum rate of change of load.

For disturbance conditions:

. ACE must be returned to zero within 10minutes.

. Corrective action must be observable inthe ACE within 1 minute of the disturbance.

This criteria is easy to use since it depends onlyon measurements taken from chart recordings ofACE.

Theoretically, each AGC system is "indepen-dent" of the neighboring AGC systems since theaddition of frequency bias to the tieline errorresults in an ACE that should respond only todisturbances internal to the system and ignore

21

disturbances in neighboring systems. In practice,the setting of the frequency bias is not equivalentto the system characteristics, and interactiondoes occur.

Thus, the ACE of a given system cannot be reliedupon for fine tuning of an AGC system. The ACEwill indicate course adjustments improvement.but the ACE changes due to activity in aneighboring system will mask fine adjustmentsand improvements will not be conclusivelyindicated.

Another problem lies in the difficulty of the lack ofdefinition of good AGC operation. The definitionseems to cover at least three areas. First, theintegral of ACE should be maintained as close tozero as possible. Second, the AGC shouldrespond to transients and disturbances, and,third, the AGC must not create unnecessaryoperation of the mechanical equipment in thesystem. Unfortunately, only the first area can beanalytically measured. The second and thirdareas are defined more by opinion than byanalytical testing.

A possible solution would be to develop an"index" of performance which, over a 10-minuteperiod, would measure the activity of the ACE,the integral of ACE, and the activity ohommandsto the speed-level motor. A single number of"goodness" would be generated. However,interpretation of this number may be difficult andinvolve opinions and may cause more problemsthan it solves.

Another possible solution is to use a fast Fouriertransform calculated every 10 minutes for ACE,speed-level motor activity, and the integral ofACE. Then comparisons of the amplitude of thefrequency components could be made. Becausethe response of various frequencies may not beas dependent on neighboring AGC systems (eachsystem will have characteristic frequencies ofactivity), the effects of fine tuning of filters andother constants may be easier to see. However, inthe final analysis, the improvement will still be amatter of opinion.

BIBLIOGRAPHY

GENERAL AGC OPERATION

1 A. R. Benson, "Control of Generation in theU.S. Columbia River Power System,"Bonneville Power Administration, 1960.

2 W. R. Brownlee, "How to Account for Energyin Complex Interconnections," ElectricalWorld, October 23, 1961.

3 H. J. Fiedler, "Power System Operation,"Indian Journal of Power and River ValleyDevelopment," vol. 21, No.7, July 1971 .

4 C. J. Frank, "Generation Control ResearchOpportunities," Power System ControlWorkshop, San Louis Obispo, CA, April1977.

5 H. J. Holley and J. C. Hung, "Load-FrequencyControl of an n-Area Power System Inter-connected by AC Lines and DC Links,"International Electrical, Electronics Con-ference, Toronto, Ontario, October 1971.

6 Mochon, "Power System Planning andOperations: Future Problems and ResearchNeeds," EPRI EL-377-SR, February 1977,Henniker Conference 1976.

7 R. Pod more, "Recommendations for Controlof Jointly Owned Generating Units,"System Control, Inc., September 1975.

B R. O. Usry, "Inadvertent Energy Interchange-Causes, Remedies, and Balancing:' IEEEPAS, vol. PAS-87, No.2, February 1968.

[9] "NAPSIC Operating Man-ual:' North American Power Systems Inter-connection Committee, April 1976.

10 L. A. Mollman and T. Kennedy, "Interrela-tionship ofTime Error, Frequency Deviation,and Inadvertent Flow on an InterconnectedSystem," IEEE PAS, vol. PAS 87, No.2,February 1968.

AGC THEORY

[11] N. Cohn, "Energy Balancing On Inter-connected Systems," Proceedings of theAmerical Power Conference, vol. 35, 1973.

[12] N. Cohn, "Control of Generation and PowerFlows on Interconnected Systems:' JohnWiley and Sons, 1966.

22

13 N. Cohn, "A Step-by-Step Analysis of Load-Frequency Control Showing the SystemRegulating Response Associated with Fre-quency Bias," presented at the Inter-connected Systems Committee Meeting,Des Moines, lA, 1956.

14 N. Cohn, "Techniques for Improving theControl of Bulk Power Transfers on Inter-connected Systems," IEEE PAS, vol. PAS90, No.6, Nov/Dec 1971.

15 N. Cohn, "The Use of Synchronized Time andFrequency Standards to Improve Control ofInterconnected Electric Power Systems,"presented to North American Power Sys-tem Interconnection Committee, Atlanta,GA. February 1969.

16 N. Cohn, "Considerations in the Regulationof Interconnected Areas," IEEE PAS, vol.PAS 86, No. 12, December 1967.

17 N. Cohn, "Research Opportunities in theControl of Bulk Power and EnergyTransferson the Interconnected Systems," EPRIEL-377-SR, February 1977, pp 6-1 through6-30, Henniker Conference, 1976.

18 N. Cohn, "Some New Thoughts on EnergyBalancing and Time Correction on Inter-connected Systems," Power System Con-trol workshop, San Louis Obispo, CA,April 1977.

19 N. Cohn, "Area Wide Generation Control -A New Method for Interconnected Systems,"Electric Light and Power, June, July,August 1953.

20 N. Cohn, "Some Aspects of Tieline BiasControl on Interconnected Power Sys-tems," AlEE Transactions, February 1957.

21 C. Concordia and L. K. Kirdhmayer, "TielinePower and Frequency Control of ElectricPower Systems, Part I and II," AlEE Trans-actions, June 1953 and April 1954.

22 R. L. Cresap, D. N. Scott, and C. W. Taylor,"Closed Loop Digital Hydroplant Controllerfor Automatic Generation Control," Bonne-ville Power Administration, September 1973.

23 J. F. Engle, "Effect of Grand Coulee Deriva-tive Type Governor on Automatic Genera-tion Control," Bonneville Power Admin-istration, July 1974.

24 D. N. Ewart, "Automatic Generation Con-trol - Performance Under Normal Condi-tions," Henniker Conference 1975.

25 C. W. Ross and K. W. Goff, "An Exper-imental Corr-elation Analyzer for MeasuringSystem Dynamics," IEEE Transactions onCommunication and Electronics, vol. 82,1963.

26 , "IEEE Standard Definition ofTerms for Automatic Generation Control,"IEEE Standard No. 94, IEEE PAS, July/August 1970.

27 F. P. deMello, R. J. Mills, and W. F. B'Rells,"Automatic Generation Control Part II -Digital Control Techniques," IEEE PAS,March/April 1973.

28 F.P. deMello, R. J. Mills, and W. F. B'Rells,"Automatic Generation Control Part I -Process Modeling," IEEE PAS, March/April1973.

[29] C. E. Fosha, Jr., and O. I. Elgerd, "The Mega-watt-Frequency Control Problem: A NewApproach Via Optimal Control Theory,"IEEE PAS, vol. PAS 89, No.4, April 1970.

[30] O. I. Elgerd and C. E. Fosha, Jr., "OptimumMegawatt-Frequency Control of MultiareaElectric Energy Systems," IEEE PAS, vol.PAS 89, No.4, April 1970.

31 K. Honda, "A Filtering Method Applied toLoad Frequency Control," Electrical Engi-neering in Japan, vol. 93, No.4, 1973.

32 T. W. Jenkens, Jr. and B. Littman, Re-sponse Capability in the Control of LargeGenerating Units," IEEE Paper 71 CP 73PWR, IEEE Winter Power Meeting, Jan-uary 1971.

33 C. W. Ross, "Error Adaptive Control Com-puter for Interconnected Power Systems,"IEEE PAS, vol. PAS 84, No.7, July 1966.

23

34 C. W. Ross, "A Comprehensive DirectDigital Load Frequency Controller," IEEEPICA Conference Record, Pittsburgh, PA,1967.

[35] D. N. Scott, R. L. Cresap, R. F. Priebe, D. E.Tebrink, and K. A. Takeuchi, "Closed LoopDigital Automatic Generation Controller,"IEEE Conference paper C 73, July 1973.

36 C. Taylor, "Effect of Sampling Rates onDigital Automatic Generation Control,"Bonneville Power Administration, Aug-ust 1974.

37 J. Weiss, "Digital Load Frequency Controlin Multiarea Power System," Fourth IFAC/IFIP International Conference on DigitalComputer Applications to Process Control,Zurich, Switzerland, March 1974.

38 J. L. Willems, "Sensitivity Analysis of theOptimum Performance of ConventionalLoad-Frequency Control," IEEE PAS, vol.PAS 93, 1974.

39 J. Zaborszki, A. K. Subramanian, and K. M.Lu, "Control Interfaces In GenerationAllocation," Henniker Conference 1975.

GENERAL AGC PERFORMANCE

40 C. W. Ross and T. A. Green, "DynamicPerformance Evaluation of a ComputerControlled Electric Power System," IEEEPAS, vol. PAS 91, No.4, Sept - Oct 1972.

41 C. W. Taylor, "Analysis of Dittmer PowerSystem Frequency Deviation Records,"Bonneville Power Administration, March1976.

AGC SIMULATION

[42] C. W. Taylor and R. L. Cresap, "Real-Time Power System Simulation for Auto-matic Generation Control," IEEE PAS, vol.PAS-95, No.1, January/February 1976.

43 J. E. Van Ness and R. Rajagopalan, "Effectsof Digital Sampling Rate on System Stabil-ity," PICA Conference Record, Pittsburgh,PA, May 1967.

44 K. N. Stanton, "Power System DynamicEnergy Balance," Proceedings of the Sym-posium on Real-Time Control of ElectricPower Systems, Brown, Boveri, and Com-pany Limited, Baden, Switzerland, 1976.

45 L. H. Fink, L. M. Smith, and R. P. Schulz,"Dynamic Mathematical Model of Inter-connected Electric Power System," Mod-eling and Simulation, IEEE/ISA,. vol. 5,Pittsburgh, PA., 1974. '

46 L. M. Smith, L. H. Fink, and R. P. Schulz, "Useof Computer Model of Interconnected PowerSystem to Assess Generation ControlStrategies," IEEE PAS, vol. PAS 94, No.5,Sept/Oct 1975.

47 H. G. Kwatny, K. C. Kalnitsky, and A. Bhatt,"An Optimal Tracking Approach To Load-Frequency Control," IEEE PAS, vol. PAS 94,No.5, Sept/Oct 1975.

48 M. Ehsan and M. J. Short, "Digital Simula-tion for On-Line Load-Frequency ControlStudies," IEEE, 1976.

ADVANCED CONTROL THEORY

49 R. Aggarwal and F. Bergseth, "Large SignalDynamics of Load Frequency Control Sys-tems and Their Optimization Using Non-linear Programming, I and II," IEEE PAS,vol. PAS 87, No.2, Feb 1968.

50 R. K. Cavin, M. C. Budge, Jr., and P.Rasmussen, "An Optimal Linear SystemsApproach to Load-FrequencyControl," IEEEPAS, vol. PAS 90, No.6, Nov/Dec 1971.

51 T. W. Reddoch, P. M. Julich, T. O. Tan, andE. C. Tacker, "Models and PerformanceFunctionals for LO;;ldFrequency Control inInterconnected Power Systems," IEEE Con-ference on Decision and Control, MiamiBeach, FL, December 1971.

52 J. D. Glover and F. C. Schweppe, "AdvancedLoad Frequency Control," IEEE PAS, vol.PAS 91, No.5, Sept/Oct 1972.

24

53 S. M. Miniesy and E. V. Bohn, "OptimumLoad Frequency Continuous Control withUnknown Deterministic Power Demand,"IEEE PAS, vol. PAS 91, No.5, Sept/Oct1972.

54 E. V. Bohn and S. M. Miniesy, "OptimumLoad-Frequency Sampled - Data ContrQIwith Randomly Varying System Disturb-ances," IEEE PAS, vol. PAS 91, No.5,Sept/Oct 1972.

55 M. Calovic, "Linear Regulator Design for aLoad and Frequency Control," IEEE PAS91, No.6, Nov/Dec 1972.

56 E. C. Tacker, C. C. Lee, T. W. Reddoch, T. O.Tan, and P. M. Julich, "Optimal Control ofInterconnected Electric Energy Systems-ANew Formulation," Proceedings of IEEE, vol60, No.1 0, October 1972.

57 E. C. Tacker, C. C. Lee, P. M. Julich, and T. O.Tan, "Models and Optimal Stochastic Con-troller Designs for Interconnected PowerSystems, "Region III,IEEE Conference, 1972.

58 W. R. Barcelo. "Effect of Power Plant Re-sponses of Optimum Load Frequency Con-trol System Design," IEEE PAS, vol. PAS 92,No.1, Jan/Feb 1973.

59 N. G. Malek, O. T. Tan. P. M. Julich:and E. C.Tacker, "Trajectory-Sensitive Design ofLoad-Frequency Control Systems," Pro-ceedings IEEE, vol. 120, No.1 0, October1973.

60 E. C. Tacker, T. W. Reddoch, O. T. Tan, andT. D. Linton, "Automatic Generation Con-trol of Electric Energy Systems - A Simula-tion Study," IEEE Trans on System, Man,

and Cybernetics, vol. SMC 3, No.4, July1973.

61 T. W. Reddoch, P. M. Julich, G. J. Burchert,Jr., O. T. Tan, and E. C. Tacker, "On the Useof Advanced Control Theory in the Design ofAutomatic Generation Controllers," IEEEConference Paper C 73, 101-3, Ja nuary 1973.

62 V. R. Moorthi and R. P. Aggarwal, "DampingEffects on Excitation Control in Load-Fre-quency Control System," Proc. IEEE, vol.121, No. 11, November 1974.

63 O. T. Tan, E. C. Tacker, and D. R. Fischer,"Automatic Generation Control SystemsContaining Governor Dead Band," IEEE Con-ference Paper C 74,127-7, January 1974.

64 T. W. Reddoch, "Load Frequency ControlPerformance Criteria With Reference to theUse of Advanced Control Theory," HennikerConference, 1975.

DIGITAL LOAD CONTROL

[65] W. B. Gish, "Digital Load Control for Hydro-electric Powerplants," U. S. Department ofthe Interior, Bureau of Reclamation, REC-ERC-77-10, August 1977.

[66] W. B. Gish, T. R. Whittemore, and U.Milano, "Grand Coulee Third Interim Con-troller Prototype Digital Load and VoltageControl," U.S. Department of the Interior,Bureau of Reclamation, REC-ERC-76-8, July1976.

25 GPO 848-069

"

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .:

. . . . .-. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .................................................. .# .: ... ....................................................

ABSTRACT

The concept of automatic generation control is often consideredcomplex because of the"freedom each utility has in choosing indi-vidual characteristics within a basic control philosophy. This reportendeavors to separate the basic philosophy from the individual'characteristics to allow a clearer understanding of the controlphilosophy. Discussions of individual characteristics used by theBureau of Reclamation are also presented.

ABSTRACT

The concept of automatic generation control is often consideredcomplex because of the freedom each utility has in choosing indi-vidual characteristics within a basic control philosophy. This reportendeavors to separate the basic philosophy from the individualcharacteristics to allow a clearer understanding of the controlphilosophy. Discussions of individual characteristics used by theBureau of Reclamation are also presented.

ABSTRACT

The concept of automatic generation control is often consideredcomplex because of the freedom each utility has in choosing indi-vidual characteristics within a basic control philosophy. This reportendeavors to separate the basic philosophy from the individualcharacteristics to allow a clearer understanding of the controlphilosophy. Discussions of individual characteristics used by theBureau of Reclamation are also presented.

ABSTRACT

The concept of automatic generation control is often consideredcomplex because of the freedom each utility has in choosing indi-vidual characteristics within a basic control philosophy. This reportendeavors to separate the basic philosophy from the individualcharacteristics to allow a clearer understanding of the controlphilosophy. Discussions of individual characteristics used by theBureau of Reclamation are also presented.

REC-ERC-78-6Gish, W. B.AUTOMATIC GENERATION CONTROL - NOTES AND OBSERVATIONSBur Reclam Rep REC-ERC-78-6, Div Res, Oct 1978. Bureau of Recla-mation, Denver 25 p., 8 fig, 66 ref

REC-ERC-78-6Gish, W. B.AUTOMATIC GENERATION CONTROL - NOTES AND OBSERVATIONSBur Reclam Rep REC-ERC-78-6, Div Res, Oct 1978. Bureau of Recla-mation, Denver 25 p., 8 fig, 66 ref

DESCRIPTORS-/ *Ioad-frequency control/ *interconnected sys-tems/ *power system operations/ *power interchange/ algorithms/automatic control/ control systems/ governors/ power dispatching/baseloads/ generating capacity/ hydroelectric power/electric gen-erators/ computer applications

DESCRIPTORS-/ *Ioad-frequency control/ *interconnected sys-tems/ *power system operations/ *power interchange/ algorithms/automatic control/ control systems/ governors/ power dispatching/baseloads/ generating capacity/ hydroelectric power/ electric gen-erators/ computer applications

IDENTIFIERS-/ automatic generation control IDENTIFIERS-/ automatic generation control

COSATI Field/Group 09B COWRR: 0902 COSATI Field/Group 09B COWRR: 0902

REC-ERC-78-6Gish, W. B.AUTOMATIC GENERATION CONTROL - NOTES AND OBSERVATIONSBur Reclam Rep REC-ERC-78-6, Div Res, Oct 1978. Bureau of Recla-mation, Denver 25 p., 8 fig, 66 ref

REC-ERC-78-6Gish, W. B.AUTOMATIC GENERATION CONTROL - NOTES AND OBSERVATIONSBur Reclam Rep REC-ERC-78-6, Div Res, Oct 1978. Bureau of Recla-mation, Denver 25 p., 8 fig, 66 ref

DESCRIPTORS-/ *Ioad-frequency control/ *interconnected sys-tems/ *power system operations/ *power interchange/ algorithms/automatic control/ control systems/ governors/ power dispatching/baseloads/ generating capacity/ hydroelectric power/ electric gen-erators/ computer applications

DESCRIPTORS-/ *Ioad-frequency control/ *interconnected sys-tems/ *power system operations/ *power interchange/ algorithms/automatic control/ control systems/ governors/ power dispatching/baseloads/ generating capacity/ hydroelectric power/ electric gen-erators/ computer applications

IDENTIFIERS-/ automatic generation control IDENTIFIERS-/ automatic generation control

COSATI Field/Group 09B COWRR: 0902 COSATI Field/Group 09B COWRR: 0902