ieee - improving the ace

8/12/2019 IEEE - Improving the ACE

1/6

IEEE Transactions on Power Systems, Vol. 10, No. 2 May 1995 1 71A Simple Method for Improving Control Area Performance:

Area Control Error (ACE) Diversity InterchangeAD1

Alan R Oneal, ManagerENEREX

140050th St. Suite 200WestDes Moines, IA 50266-592 1Abdrud Control Areas within three major (andessentially separate) areas of North America areinterconnected electrically, thus enjoying vastly improvedreliability and economy of operation compared tooperating in isolation. Each must continually balapceload, interchange and generation to minimize adverseinfluence on neighboring control areas andinterconnection frequency. This requires investment incontrol systems and the sacrifice of some fuel conversionefficiencies to achieve the objective of complying withminimum control performance standards set by the NorthAmerican Electric Reliability Council (NERC). Controlalso increases wear and tea r on machinery in the pursuitof these goals. Area Control Area (ACE) DiversityInterchange (ADI) offers a means of reducing this controlburden without undue investment or sacrifice by anyparticipaut in a group. This paper describes thephilosophy of AD1 and th e ENEREX partnership sfavorable experiences with its actual implementation inIowa.

I. INTRODUCTIONENEREX is a threecompany partnership in Iowadealing primarily in bulk-power exchange among itsmembers: Iowa Electric Light Power (ESC), Iowa-Illinois

Gas Electric Co. IIGE) and Midwest Power Systems Inc.(MPSI). ENERJ2X began as a fivecompany partnership in1984 and has lost two partners through two mergers ofpairs of the original partners. Iowa Power and Iowa Public94 SM 586-8 PWRSby the IEEE Power System Engineering Committee of theIEEE Power Engineering Society for p r ese nt a ti o n a tthe IEEE/PES 1994 Summer Meeting, San Fr an ci sc o, CA,July 24 - 28, 1994. Manuscript submi tted Janua ry 41994; made available for printing June 14, 1994.

paper recommended and approved

SeMce became MPSI in 1992 and Iowa Electric LightPower and Iowa Southern Utilities became IESC in 1993.All ENEREX partners are small-to-medium sized controlareas operating in the Mid-continent Area Power Pool(MAPP). Their peak demands range from approximately1100h4W (IIGE) to approximately 2300MW (MFSI).

ENEREX has the capability to exchange data on anAutomatic Generation Control (AGC)cycle basis (every fourseconds) using a specially-adapted SCADA system. Thissystem has primarily sewed to allow interchange of controlinformation for shares of jointly-owned generating units(JOUs), but is now being used for another novel purpose aswell. That purpose is the interchange of diversity schedulesfor the dynamic correction of Area Control Error (ACE).Control of ACE is a costly affair for every control area in allelectrical interconnections in North America, including theEastern Interconnection of which the ENEREX companiesare a part. However, proper control is a necessaryresponsibility for maintenance of interconnection frequencyand localized stability in the electric systems. The NorthAmerican Electric Reliability Council (NERC) sets standardsfor realization of this responsibility, measured by ControlPerformance Compliance (CPC) percentages [l] Thepurpose of ACE Diversity Interchange (ADI) is maximummitigation of the required control efort to achieve acceptableCPC percentages based on available diversity among a groupof participants. All it requires is two or more control re swilling to participate, a means of communicating data on areal-time basis and calculation software at a common site.

II. AREXCONTROLERRORACE is defined [11as the difference between actual netarea interchange and scheduled net area interchange, with acomponent or frequency bias. Since the ACE calculationuses i nstantaneous information, the interchange is power,stated in Megawatts, rather than energy, as used hourly. In

classical equation form (calculated each AGC cycle):ACE = (Ni, NiJ 1OB(F, (1)

where:Ni, -Net power interchange (actual)Ni, Net power interchange (scheduled)B -Bias (Megawattdo. 1H z unique for each controlarea)Fa Frequency, actual (measured)Fs Frequency, scheduled (normally 6O.OOOHz

0885-8950/95/$04.00 1994 IEEE


2/6

1072As referred to in this document, raw ACE is the errorvalue calculated by any AD1 participant based on its ownschedules and telemetered interchange prior to any ateringor the of ADI. A cycle is one AGC calculationcycle, usually four seconds. ACE is the basis used byAutomatic Generation Control (AGC) a l g o r i h inestablishing the way in which a control area s generation

resourcesare allocated to balance requirement with output.m. HEAD1 SCHEDULING PROCESS

A. PremisesAD1 works on the assumption that instances occur inwhich one or more of theACEs of the participants (of course,there must be two or more) are of opposite sign from one ormore others. Whenever this is the case here is diversitywhich canbe exploited to partially or wholly correct the rawACE of some or all participants. The ENEREX AD1algorithm ensures that no calculated schedule ever makesany raw ACE worse... t will always move raw ACE towardzero or do nothing to it.Protective parameters must be provided to eliminateundesired corrective action on the part of the pracedure,namely:

Maximum allowable (reasonable) raw ACE valueMaximum allowable rate-ofchangeof raw ACEDeadband for acting on raw ACE value receivedMaximum allowable AD1 schedule sent to participantMaximum rate-ofchange of scheduleAn additional switch is provided which can remove a

participant from the procedure instantly. It has rarely beenused but is a convenience to the participants. The parametersare provided for the convenience of the participants and maybe changed at their request without delay.B.Method

Steps n Calculation of Schedules (each AGC cycle):All ACEs are received and limits applied according toeach participant's Maximum values and Maximumrates-ofchange. Further limitations are imposed byignoring (zeroing the AD1 chedule) of any participantwhose raw ACE is inside its deadband.

Diversity is tested for (is at least one raw ACE oppositein sign from others?). If none, all AD1 schedules areset to zero and the algorithm waitsuntil next cycle.

If diversity exists, positive and negative ACEs areseparated into two groups, d y ne smaller and onelarger. Zero or ignored raw ACE values are set asideand each of their participant's schedules set to zero.Schedules are set for the Small Group (SG). Since theLarge Group LG)will be able to supply all thediversity needed to zero the SG, all the SG's schedulesare set to their raw ACE value, which will zero theirACEs. Note: by convention, a schedule of the samesign and magnitude as ACE willzero ACE.SG's AD1 schedules are checked for violations ofrate-ofchange or maximum size. Any that require itarereduced.Schedulesare then set for the Large Group:Begin with the smallest raw ACE and prorate AD1schedule(s) to each (ith) member of the LG:

(each amount is rounded until thelastLG participant)Set last (largest raw ACE in the LG) AD1 schedule tothe remaining unused diversity amount to avoidrounding prelems.Check LG schedules for violations of rate-ofxhange ormaximum size. Any that require it are r e d d .If reductions in LG schedules were made, scan othermembers of the LG for any additional diversity whichthey could provide above the amount calculated inabove. If additionalMW are available, utilize them inprorated fashion up to the point where they areexhausted or the total AD1 requirement for the SG issatisfied.If, after this search, reductions in the LG schedulesleave the SG short of needed AD1 MW return to theSG and reduce interchange using the same techniqueas n (2) above to balance the two groups.

Transmit schedules to participants.Increment MWh accumulators (for end-of-hour energycalculations). MWh schedules are transmitted toparticipants at the end of each hour for their energyaccounting.


3/6

1073IV.NTERFACING WITH PARTICIPANTS

ENEREX uses a sma l l PC-based SCADA system fordata interchange. To interface with ENEREX ADI, eachparticipant hasdefined a data point for raw ACE to be sent toENEREX and a schedule to be received from ENEREX.Hourly database points allow transmission of the resultingintegrated received and delivered MWh amounts.

Time delay skew) exists in every SCADA system.Participants' ACEs are the products of sequential samplingsystems which have inherent time delays in theircomponents. Participantsmust be aware that while the AD1schedule sent to them wil lproperly correcta raw ACE that isknown at a given point in time, delay in data transmissioncould occasionally cause the schedule to make ACE Iookdifferent than it ideally should. Tests have shown the actualdelay in turnaround (the time between a participant'stransmission of raw ACE and receipt of a correspondingAD1 schedule) to be 8-12 seconds,which has an ect on theappearance of ACE but has almost no effect on the overallimmovement in CPC ~erformanceDercentages. Because oftime delay, it has been advised that a softwarecheck be madein the participant's AGC software,where necessary, o allowgenerating unit controls to ignore the effect of the AD1schedule if it momentarily makes ACE worse (this shouldalready exist in some form in most AGC systems).Deadband values must be chosen to minimize this effectwhile allowing AD1 to do itsjob.

V. GENERATING UNlT CONTROL EFFECTSA. ADI Schedule Usage

AD1 scheduling either reduca raw ACE or leaves itunaffected. ADIcorrected raw ACE should be used toSURR~VAGC. If this is made so, generating units will nothave to chase ACE as much, since ACE movementswillbe reduced. If the schedule is ignored by AGC, ACE (on thechart) will be improved along with NERC performancecriteria, but units will still chase aw ACE as if AD1 werenot there. Note, in Figure 1, that the ACE fed to AGC maycome from the summation calculation (raw ACE plus AD1schedule, or AD1 ACE) or fromrawACE.

PARTICIPANTS

IIigure 1 AD1 InformationFlow and UsageB.UndesirableSchedule Settling

Settling is a term used to describe a condition whereAD1 participants a l l use AD1 schedules to a f f e c t control andas a result offsetting schedules may zero participant ACEs,causing their AGCs to settle their generating unitsat levelswhich do not change the schedules. In AGC systems thatcalculate Economic Dispatch based upon total generationrather than generation reauirement, pairs or groups ofparticipants may be held offeconomics or indefinite periods.Experience has shown this to occur although infrequently.A reactive approach may be a permissive AD1 algorithmwhich monitors duration of the schedule@)may be installedto taper the schedule to zero if its duration is consideredexcessive (usually in excess of 10 minutes and 10 MW).Another approach suggested has been for participant AGC torecognize the duration of the schedule (integrate it) and limitthe permissible integration value. The latter solution isdifficult to implement across a group of participants due todelays and difficulty in changing AGC software The extentof and constraints on this phenomenon will vary with thesizeof the participant system(s).

VI. EFFECTSON THEINTERCONNECTED YSTEMSThe AD1 schedules exchanged every 4 seconds net tozero. The fect on participants is in reducing raw ACE tothe extent possible, and decreasl n generating unit COfTectiveaction. The immediate net effect on the interconnection iseither nil, in the case of participants who only use raw ACEto drive AGC, to sma l l in the case of groups usingADIarrected ACE in AGC. The smaU effect results fromthe net reduction in units chasing ACE within participantcontrol areas. This net &ect may c w the otal momentary

net inadvertent of the group to be either larger or smallerthan it othemise would have been, but taken as a whole the


4/6

1074effect must be small since CPC percentages must bemaintained at acceptable levels.

Under AD1 a portion of what would otherwise have beeninadvertent energy for a control area with the interconnectionasa whole isdiverted to reduce other companied inadvertent,in return for which the inadvertent of the company inquestion is reduced (and performance mproved). Since hetotal absolute inadvertent of the group is reduced, inadvertentenergy payback at future times outside the group iscorrespondingly reduced.

After-the-fact repayment schedules are made in such away as to ensure that each hour they net to zero, also. Apossible benefit to the interconnection is in that theinadvertent account of each AD1 participant is held to aminimum, which also minimizes the necessity for thatcontrol area to unilaterally balance their account. This, inturn may reduce the accumulated inadvertent of othernon-participant control areas in the interconnection, as wellas the perfahations of frequency that at times accompanypayback.

VII. ENERGYACCOUNTINGThe interchangeof periodic energy schedules to correctraw ACE and the repayment of accumulations of such energyis classified as inadvertent repayment, as defined in theNERC Operating Manual [l]. This type of payback ismultilateral rather than bilateral or unilateral, but may be

prorated if necessazy to determine bilateral schedules. TheAD1 method is in the same mirit as bilateral inadvertentremvment. but encornmisses more Darticiwts.The integration (accumulation) of the AD1 schedules to

calculate hourly energy involves totaling 4-second-basedMWh values, received and delivered, during an hour. At theend of the hour, the totals are divided by the number of4-second intervals in the hour (900) and the result rounded tothe nearest whole number. Each hour's schedules are assuredof Mance by testing for same, then aiecting the rounding ofaccumulated schedules as necessary to balance receipts anddeliveries (to net zero). The remainders from the integrationprocess (fractionalMWh) are carried forward and added tothe next hour's beginning accumulations.

Integrations of AD1 schedules produce hourly deliveredand received totals for each participant. Each day, theirsums produce on-peak hours ending 0700-2200) andoff-peak (hours ending 0100-06oO and 2300-2400 netenergy accumulations as if between each participant andENEREX. These accumulations are reconciled (verified)between ENEREX and each participant daily. Energy

balances are accumulated (as if to ENEREX) each hour andare repaid weekly as needed. Balances larger thanpredetermined limits (HIGH THRESHOLD) are scheduledwith the participants by ENEREX accoUnting to bring themback within limits LOW THRESHOLD). Since AD1 isbidirectional and somewhat random, large energycci unul t i ons in any given direction do not occur for anyone participant in a week. During the first 6 montbs ofpermanent use of t h i s procedure in ENEREX, ourly energyamulat ions from AD1 averaged just over 1MWh/hr or atypical participant.

WI. I N G PARAMETERSELECTIONIn order to gain maximum benefit from ADI,participants must not overanstrain the AD1 system by useof restrictive tuning parameters. Each parameter and itsintention and effects s mentioned below:Maximum Allowable raw ACEThe raw ACE which is used by the AD1 calculation is

limited to a selected maximum value. It is a reasonabilitycheck on values which may be of questionable accuracy.Initially,thisvalue was left large (100 MW for the ENEREXpartners) to allow for large ACE values to be recognized, yetreject anomalouslyhighACES. It remains at 100MW for allparticipantsa year after inception of ADI.Maximum Allowable Rate-of-Change of raw ACE

Limits the raw ACE recognized during the current4-second interval to the raw ACE in the last intervalplus-or-minus the allowable rate-ofchange. It is both areasonability and a filtering parameter. It, too has been leftlarge (100 MW/cycle).

Maximum Schedule Returned to ParticipantSpecifies the largest schedule allowable for a givenparticipant in ADI. Most are set to 100 MW, although oneparticipant was reduced to 30 MW temporarily duringsolution of the ''settling phenomenon.

Maximum Rateaf-Change of ScheduleRestricts the change in schedule from one cycle to thenext.

Deadband for Acting on raw ACE Value ReceivedThis is the most imwrtant tuning ~ a r a mter relative totime delay. It allows only raw ACE values which fall outside


5/6

1075

A1 YoRaw ACE

its boundaries o be recognized. It is thus possible to ignoresmall raw ACE valw...the ones most likely to be smal lenough tobe adversely affected by time delay. The value fordeadbandinghas been kept small...defhitely less that Ld [l](a NERC control criterion parameter unique to each controlarea , or ADI'sbeneficial effectswill be hampered. A typicalvalue for deadband is 3 MW in the ENEREX control areaswhile Lds average approximately 10 MW. Figure 2illustrates what effect deadbanding has on AD1 and ACE,where the narrow trace represents raw ACE and the thicktrace shows correctedACE:

IELP IIGE MPSI ISU83.6% 77.4% 62.7% 81.2%

Figure 2 AD1 Deadband Effects

With AD1

The larger a participant's deadband, the less opportunitythere is to correct raw ACE when diversity is present. In thefigure above it has been assumed, for simplicity, that enoughdiversitywas always available to correct raw ACE to zero. Ifless is available, ACE will not move completely to zerowhenAD1 is applied. 94.0% 90.5 81.7% 92.4%

E FNEFITSOF AD1A. Eflciency Improvement

Regulation for control costs money. ADI, when aDDliedto a Darticimt's control algorithm, reduces the cost ofregulation.The ENEREX partnership began AD1 on a permanentbasis January 1, 1993. It has run continuouSy since thattime. It is not precisely known what the value of thereduction in regulation through AD1 is worth in dollars. Its

worth includes both efficiency gains and wear-and-tear (alsocalled Operation and Maintenance, or O&M') cost reductions.

The value of AD1 to the ENEREX partnership has beenapproximated at 4OO,OOO per year. This figure is based onefficiency gains, only. Wear-and-tear costs are much harderto quane andwere not included in the estimate.B. NERC Control Performance Criteria Compliance Eflects

During the first six months of ENEREXs use of ADI,improvement was seen in CPC percentages relative to whathad been experienced in the past. This mprovement ranged2-7% across what was at that time four participants. Thisimprovement is real and can be said to represent an actualincrease in perfolllunce.In looking at a comparison of raw ACE performanceversus AD1 ACE performance, a merent picture emerges.Each ENEREX partner used AD1 ACE to feed AGC, whichallowed generation to rdax and not respond to raw ACE.It is important to be aware that when AD1 is used in thisway, CPC compliance for the AD1 ACE is very real but thepedormance of the raw ACE is no longer representativeofwhat the participant would have done without ADI. Rather,

it shows the extent to which AGC and the generating unitswere able to relax and reduce heir pursuit of the raw ACE.This caveat in no way dilutes the effectiveness of ADI, itjustcautions the reader to interpret the d t s n the proper way.

Figures 3a and 3b show the CPC percentage di f f erencesduringthe first 6 months of 1993 for the ENEREX partners.These CPC percentages represent statistics compiledcontinuously (every four seconds for the 181days January 1to May 31.

L

loo I--

IELP IIGE MPSI ISUFigure 3a A 1 Per5ormance Comparison


6/6

1076This problem canbe looked for by software and corrected orsimply left alone. The ~ ~ ~ u r r e n ~ ef such violations was

WithAD1 .9? 84.1% 83.8 94.7% (representing about 104,000 total ten-minute int ends forRaw ACE 90.8 72.1% 72.7% 89.1% about five in six months of operation

four participants).There remains no question in the ENEREX partnership

that AD1 is a worthwhile, lowcost way to achieve some100 ,908070650

benefit of combined operation while upholding the autonomyof the separate controlareas.

XI. ACKNOWLEDGMENTSThe author respectfully acknowledges the philosophicalcontributions of Loren R Walker and the support of theENEREX Executive Committee, Operating Committee andprogramming stafin bringing t h i s project to reality.

IELP IIGE MPSI ISU W . EFERENCES

Figure 3b A2 Performance ComparisonX.CONCLUSIONS

The ENEREX experience with AD1 has been a hitfidone. Now that the process is running on a permanent basis,it needs almost no attention to operate properly and provideongoing benefits to participants. For those control rmwhich m y be willing to undertake implementation of AJX,the process only requires two or more participants, thecommunications interface to exchange the data and acomputer on which to effect the necessarycalculations.Potential problems to be aware of are:

Time delaysin ~~mmunicatiomSettling of schedules when ACEs are zeroedData loss due to hardware or software malfunction(Enough data shouldbe saved to reconstruct)handled algorithmically. A2 performance violation phenomenon ..can be

[l] North American Electric Reliability Council (NERC)OperatingManUal.

WI IOGRAPHYAlan R h e a l is a native ofcentral Iowa, graduating with aBSEE from Iowa State Universityin 1971. After a tour in the USArmy (1972-74) he returned toprivate business in electronicdesign and maintenance. In 1978he began employment with IowaPower Light Co. in DesMoines,Iowa in their System Operationsdepartment. Three years later hemoved to planning with IowaPower. After threemore yearsin planning, he became involved with the development of theENEREX project working toward a fivecompanypartnership for a combined control area. In 1984 be joinedENEREX as Superintendent of Operations and was promoted

to Manager in 1989. He is currently serving on the NERCThe last problem needs additional treatment. When perfom= subcommittee as coordinator for theinpartners through data exchange, bulk energy interchange

Other participantsn a goup need schedules in a given Midanhent Area power pool (MAPP) and is a ParticipantWen are backand forth zero, the program. E m X to the remaining threedirection to coffect their ACEs and one or more otherother partners will experience a zeroing or nearly so) ofraw ACE on one side of zero and no effecton the other side.This causes a sort of half-wave rectification of their raw

Emergency Electric power

and ADI.ACE which will shift its average value. If the average isshifted far enough during a ten-minute interval, an A2violation may occur where none existed in the raw ACE. December 22 993

ieee - improving the ace

Documents