新增课文1EMS.docx

新增课文1EMS.docx

《新增课文1EMS.docx》由会员分享，可在线阅读，更多相关《新增课文1EMS.docx（13页珍藏版）》请在冰点文库上搜索。

新增课文1EMS

EnergyManagemantSystem

SurveyofArchitechtures

ThehistoryofEnergyManagementSystemsforthepast10yearshasbeenacontinualprogresstowarddistributedsystems.AnEMSiseasilydividedalongfunctionallines,apropertythatmanysuppliersandutilitieshavetakenadvantageofindistributingsystemsamongseveralcomputers．ThetechniqueofusingmultipleidenticalcomputerstohandleanEMSwasacceptableaslongastheloadwasevenlybalancedbetweenthedistributedparts．Theexpansionthattookplacefrom1980to1985，however,hasbeenpredominatelyinthenetworkanalysisarea.Thissegmentofthesystemsofaroutweighstheothersthatabalancedsystemapproachisnolongerfeasible．Thedualor“quad”minicomputersystemscannotsupportamodernEMSbutremainanexcellentapproachforevenverylargesupervisorycontrolanddata-acquisition（SCADA）systemsandSCADAsystemswithgenerationcontrolfunctions.

Certainproblemsinherentindistributedsystemsareamplifiedinsystemsthatcombinediversetypesofcomputers.Thetaskofmaintainingthesystemcanbeverydifficultifeachcomputerhasaseparatedatabase.Programminglanguagesandtechniquesthataredifferentalsocomplicatethemaintenanceproblem.Themethodsusedbydifferentsupplierstoaddresstheneedforadistributedarchitectureandtheproblemsassociatedwithsuchanarchitecturearepresentedinthefollowingsections.

ConceptualModelofanEMS

TheinformationusedtodevelopthemodelofanEMSpresentedherewastakenfromastudyperformedbyFerrantiInternationalControlsCorporationin1983anddiscussedintheAugust1985issueofIEEESpectrum.Conceptually,anEMSconsistsoffourparts：

1.Networkanalysis：

stateestimation,contingencyselectionandanalysis，optimalloadflow,voltagecontrol,dispatcher'sloadflow,anddispatchertrainingsimulator.

2.Generationschedulingandcontrol：

automaticgenerationcontrol，economicdispatch，andinterchangescheduling．

3.Database：

datastorageandretrievalservicesforallareas.

4.SCADA:

dataacquisition,alarmprocessing,andman/machine．

Figure1showstherelationshipbetweenthevariouspartsoftheconceptualmodel．Themostnotablecharacteristicsofthismodelare：

1.Otherthannetworkanalysis，floating－pointprocessingisminimal.

2.Thetotalprocessingrequirementismuchhigherfornetworkanalysisthanforallotherpartscombined.

3.Allareasrequireaveryhightransactionratewiththedatabase（whereatransactionisdefinedasacompleteaccessofthedatabase）．

4.TheSCADAfunctionmustrespondtoahighinterruptload.

Itisclearthatthediverserequirementsofthesedifferentareascannotbesatisfiedbyasinglecomputerorbymultiplecopiesofthesamecomputer.Acomputationalloadofover9MIPScannotbesustainedinconjunctionwith800interruptspersecond.Computersthatareverygoodfloating-pointprocessorsdonothandleinterruptswell．Ingeneral，everythingthatcanbedonetoimprovetheperformanceofacomputer（pipelining,caching,etc.）makesitlesseffectiveatprocessingcontextswitchesandinterrupts.Forexample，apipelinedmachinegetsitsspeedbyprocessingseveralinstructionsatonce，eachoneclockcyclebehindtheother.Ifaparticularinstructiontakeseightcycles，bythetimeitmakesitswaythroughthepipeline，sevenotherinstructionsarepartwayintothepipeline.Aninterruptorcontextswitchattheendofthisinstructioncausesthecontentsofthepipelinetobelost.Inthisexample,partsofseveninstructionsarelost.Memorycachingpresentsadifferentproblem.Thespeedofthecomputerdependsonthecachehitrate（thatis,thepercentageofthetotalmemoryaccessesthatcanbesatisfiedoutofcache）．Thedifferenceinperformancewithandwithoutcachecanbeasmuchas4to1，butisusually2to1．Theeffectofaninterruptoncacheisthattheinterruptroutineanditsdatareplacethecontentsofcache．Inareal-timesystem,aninterruptusuallycausesataskswitch.Whenthetaskisinvoked,itanditsdatawillagainreplacethecontentsofcache.Whentheinterruptandthetaskitcausestoexecutearecompleteandthesystemreturnstotheinterruptedtask,thecachemustbereloadedagain.Sincetheinterruptroutineandthesubsequenttaskareeachexecutedonlyonce，cachedoesnothelpthemmuch.Evenworse，theinterruptedtask（thatprobablywashelpedbycache）mustreestablishthecachecontentsaftereveryinterrupt.

InanEMS,theSCADAloadischaracteristicallyalargenumberofsmall，quicktasks.Thistypeofloaddefeatsapipelinedorcachedcomputerbycausingthepipelinetobelostandthecachetobereloadedoften.Ontheotherhand，networkanalysisischaracteristicallyasmallnumberoflarge,long-runningtasks.Thesetypesofloadsaregreatlyaidedbypipeliningandcaching.So,clearly,differenttypesofcomputersareneededfornetworkanalysisandSCADA.

GrowthinSCADAtakestheformofhorizontalgrowth[replicatingcomponentstoaddmoreCRTsandremoteterminalunits（RTUs）]andgrowthinnetworkanalysistakestheformofverticalgrowth（expandingthesizeofthecomputerusedfornetworkanalysis）．Toaccommodatethistypeofgrowth,theSCADAportionshouldbehorizontallyexpandableandthenetworkanalysisportionshouldbeverticallyexpandable．Thatis,theSCADAcomputer（s）shouldbeincrementallyexpandableasnewenddevicesareadded,andthenetworkanalysiscomputer（s）mustbeupgradeableorreplaceabletoaccommodateadditionalfunctionsandbiggerpowersystemmodels.Ifthegrowthratefrom1980to1985ismaintaineduntil1990,networkanalysiswillrequirealmost160MIPSofprocessingpower!

Generationschedulingandcontrolissmallenoughtobeplacedwhereveritisconvenientandisnotsignificantintermsofexpansion.

Thefollowingsectionswillconsiderthethreemajorarchitecturesinusetodayandwillcontrastthemtotheconceptualmodel.Thefollowingdiscussionshouldbetreatedasatheoreticaldiscussion,notadescriptionofanyexistingsystems．Certainly,manyexamplesofperfectlyviablesystemsthataresimilartooneofthefollowingarchitecturescanbefound．Inreality,mostsystemsaresomewherebetweentwoofthesearchitectures.Forexample，theOntarioHydrosystemdescribedearlierhasattributesofboththecentralizedanddistributedarchitectures.

CentralizedArchitecture

ThecentralizedarchitectureshowninFig．2istypicalofseveralofferingsavailabletoday.Thissystemconsistsofapairofmainframecomputersandsomenumberofmicroprocessorsconnectedtothemainframe．

Themicroprocessorsperformlow-leveldata-acquisitionanddisplaysupportfunctions,leavingthebulkofSCADAinthemainframesalongwithnetworkanalysisandgenerationschedulingandcontrol．

TherelationshipbetweenthissystemandtheconceptualmodelisshowninFig.3.Thisapproachpartiallyaddressestheproblemofhandlingalargenumberofinterrupts;themicroprocessorsabsorbsomeofthem.ItprovidessomelevelofhorizontalexpansionwithinthemicroprocessorsbutdoesnotallowforexpansionofSCADAprocessingcapabilityindependentofnetworkanalysis．Thissystemprovidesexcellentverticalexpansion；typicalmainframefamilieshaveprocessorsfromsmalltoverylarge，butthelackofhorizontalexpansioncapabilitycouldresultinbeingforcedintoamainframemodelupdatetoaddRTUsorCRTs.

Themaintenanceofthissystemisextremelysimple．Infact,somesuppliersappeartohaveabandonedthedistributedarchitecturebecauseoftheirexperienceintryingtosolvethemaintenanceproblemsthatarepresentinearliersystemsthatlackedacentralizeddatabasemanagerandcommonprogrammingtools.

Theresponsivenessofthistypeofsystemwillvarygreatlywithsystemconditions.Duringupsetconditions，averyhighnumberofalarmsmaybereceived．Processingthesealarmswilltakeresourcesawayfromnetworkanalysisanddisplaysupport.Settingtheprioritiesofthesystemsuchthatprocessingalarmsissecondarytonetworkanalysiscancausealargealarmsbacklog，hidingserioussystemconditionsfromtheoperator.Whileitmaybereasonabletoallownetworkanalysisfunctionstobesloweddownbyalarmprocessingwiththesetoffunctionsavailabletoday,asclosed-loopfunctionslikevoltagecontrolbecomemoreprevalent,thismethodofresolvingresourcecontentionwillbecomelessacceptable．

Inaddition,whatisreasonabletoasystemdesigner（slowingdownnetworkanalysis）maybeunreasonabletothesystemoperator.Inthefuture，utilitiesmayhaveengineersonduty,studyingthesystemasitgoesdown.Ifthesystemisnotresponsive，includingnetworkanalysis，duringupsetconditions,theengineerandtheoperatorwillbeunhappywithit.

ClusteredArchitectures

TheuseoftheVAXCluster[fromDigitalEquipmentCorporation（DEC）]toimplementadistributedEMSisarecentdevelopmentthathasbeenadoptedinvaryingdegreesbyseveralsuppliers．ThesystemdepictedinFig．4istypicalofthesesystems.ThistypeofsystemusuallyconsistsofsomenumberofVAX8000seriescomputers,performingnetworkanalysisandgenerationcontrolandscheduling,somenumberofsmallerVAXcomputersperformingSCADA,andtwoHSC（hierarchicalstoragecontroller,anintelligentstoragedevice）50sand70sactingasacommondatabase．MostsuppliersusemicroprocessorstohelphandletheSCADAload，particularlyfordataacquisition.Eachcomputerhasadedicatedbackupinsomesystemswhereasothersprovidedegradedoperationwhenoneofthenetworkanalysiscomputersfails.

TherelationshipbetweenthistypeofsystemandtheconceptualmodelisshowninFig．5．Thisarchitecturedoesnothandletheproblemsassociatedwithinterruptsverywell.TheVAXitselfdoesno