Dynamic zoning based supervisory control for elevators

《Dynamic zoning based supervisory control for elevators》由会员上传分享，免费在线阅读，更多相关内容在学术论文-天天文库。

Proceedingsofthe1999IEEEInternationalConferenceonControlApplicationsWp2-66:OOKohalaCoast-IslandofHawai’i,Hawai’i,USA*August22-21,1999DynamicZoningBasedSupervisoryControlforElevatorsAlbertT.P.SoJansonK.L.YuW.L.ChanCityUniversityofHongKongCityUniversityofHongKongHongKongPolytechnicUniversityAbstract-Zoningisoneconventionalwaytomakeanscheduled.Staticzoningreferstothepermanentassignmentexistingelevatorsystemadapttodirerenttrafficpatterns.ofagroupofcarstoserviceanumberoffloorsusuallyHowever,zoningisgenerallyeitherfixedpermanentlyoradjacentinthebuilding.Time-scheduledzoningreferstostaticbasedontimeschedulingwherethefloorsbeingtemporaryzoningofthebuildingduringapre-scheduledgroupedintozonesarepredeterminedduringthedesignperiodoftimeduringtheday.Theperiodusuallycoincidesstage.Inthispaper,anewconceptisintroducedthatcanwiththepeaktrafficsituationorrushhoursandtheprovidethefeatureofdynamiczoningforaelevatorsystemelevatorsserviceallfloorsoutsidethisperiod.basedontherealtimetrafficpatterns.TheobjectivesareThemeritofzoningisnotcontroversial.Actually,ittoachievemuximumhandlingcapacityofthesystemandhasbeenimplementedinlotsofcommercialanddomesticminimumwaiting/travellingtimeforpassengersduringup-buildings.However,theexistingcontrolpatternsofzoningpeakanddown-peaktraf/icconditions.Twonewareeitherpre-deteminedorfixedonatime-scheduleshasis,algorithms.namely“UniformRTT“and“Demandorinotherwords,theyarenotadaptabletotherealtimeMatching’:havebeendevelopedforintelligentcontrol.trafficpatterns.ThisshortfallinitiatesourresearchintotheInformationrelatedtorealtimezoningcanbedeliveredtoideaofdynamiczoning.TheconceptofprimitivedynamicpassengersbyLEDdisplayboardsinstalledateachlandingzoningforup-peakonlyhasbeendeveloped[3]sothatabovethelandingdoors.Acomputersimulationbasedonelevatorsmakelessstopsperroundtripandcarsretumtoasoftwarepackageonmarket,Elevator2.0,hasrevealedlobbyfaster.However,thatsystemwasnotadaptabletothethatthetwonewalgorithmscanimprovethepassengerchangingtrafficflowpatternsandtheassignmentoffloorswaiting/travellingtimeunderthetwomajortrafficforchannellingwasnottotallydynamic.Inthispaper,theconditions.Theoriginalconceptofdynamiczoningwasmathematicalalgorithmsfortotaldynamiczoningispublishedinanelevatorjournal[I].presentedforbothup-peakanddown-peakconditions.Floorassignmentistotallyflexibleanditiseasily1lntroductionimplementedtomodemelevatorsystems.AcomputerItisgenerallyawarethatelevatortrafficpatternsaresimulationusingastandardsoftwarepackage,Elevate2.0,changingsignificantlyduringthedailyoperations,fromup-onmarketverifiedtheclaimthatdynamiczoningwaspeak,down-peaktopeakinterfloorandevenoff-peaketc.superiortotheconventionalsupervisorycontrolwithoutConventionalelevatorsystemdesignhashadmuchemphasiszoningatall.onup-peaktraffic.Hence,thedesignofagoodelevatortrafficcontrollerbecomesaveryimportantjobifthetarget2KnowledgeofTrafficPatternsofoptimalcontrolisrequiredwholedaylong.Here,Beforeadecisiononthezoningdetailsisarrivedat,optimalcontrolreferstohighesthandlingcapacityandinformationrelatedtotherealtimeelevatortrafficpatternsshortestwaitingtimeandtravellingtimeofpassengers.isabsolutelynecessary.Thatconsistsoftwoaspects.ZoningisaclassicalwaytoachievingpartoftheaimstoFirstofall,thetrafficmodehastobedetermined,up-improvetheperformanceofanelevatorsystem.peak,down-peak,interfloororoff-peak.AsithasbeenInamodeml.ighrisebuilding,eachelevatorcarisnotmentionedintheintroduction,zoningwillheeffectiveusuallyrequiredtoserviceeverylevel,asthiswouldimplyduringtwomodes,i.e.up-peakanddown-peak.Therearealargenumberofstopsduringeachcartrip,anddifferentwaystoidentifyup-peakanddown-peaktrafficconsequentlylongjourneytimesforthepassengersinthemodes.TwoparameterswereproposedbyBeehe[4]thatcar[2].Morecommonly,aelevatorcarservesonlyacouldidentifytheoccurrenceofup-peaktrafficpattern,numberoffloorsclusteredtogethertoformazone.ThenamelytheUp-DownStoprelationshipandtheUp-StopAJp-floorsservedareusuallyadjacent,althoughsomebuildingslandingCallrelationship.Similarly,therearetwomayhavesplitsubzoneswheretheoccupantsofeachparameters[4]thatcanidentifytheoccurrenceofdown-subzoneareassociatedwitheachotherandcanbeexpectedpeaktrafficpattern,namelytheDown-UpStoprelationshiptogeneratesomeinterfloormovements.AgroupofcarsandtheStops-LandingCallrelationship.Thedetailshaveservingahighrisezonealsoprovidesservicetoandfrombeensummarisedinapaperpresentedbytheauthorsinthemainterminal,travellingexpressbetweenthisfloorandElevcon’93[SI.Anotherwayistheemploymentofthelowestfloorinthezone,knownastheexpresszoneartificialneuralnetworks(ANN)thatcontinuouslymonitorterminal.Amajoradvantageofzoningistheincreaseoftheelevatorsystem[6].Allrelevantdataareretrievedfromtheelevatorsystemhandlingcapacity.Atpresent,zoningthecomputerisedsupervisorycontrolsystemandnormalisedinahighrisebuildingmayeitherhestaticortime-intoarange[0,11andfedintotheinputnodesoftheANN.0-7803-5446-X/99$10.0001999IEEE15911 AwelltrainedANNcanidentifythepossibletrafficmodesmathematicalproblemthenbecomestheminimisationofbasedontherealtimedata.eitherofthefollowingtwocostfunctions:Secondly,thetrafficdemandofeachfloorneedstobeidentified.Inconventionalelevatordesign[7],thedemandofeachfloorcorrelateswiththepopulationofthatfloorandtheassumptionisparticularlyvalidforup-peaktraffic.Inourcase,thisassumptioncanstillbeused.Foramoreadvancedapproach,thechangeofcarweightand/orthecardooropeningtimeassociatedwitheachstopateachlandingcanheusedtoestimatetheinstantaneoustrafficdemandofthatlanding,orevenacomputervisionbasedpassengercountingsystemcanheemployed18-101.3AlgorithmsofDynamicZoningI”Itisassumedthatthetrafficdemandofeachflooris=-Rnjrnj-lknown.AcommercialbuildingwithNnumberoffloors,excludingmainterminal(MT=O/F),isconsidered.Within“Ithe5-minutedurationofup-peak,totaldemandoftheMin(Uj-HC,)’buildingisUandthedemandofthekthfloorisU,.It11,,....L1j-1300shouldhenotedthatUandU,candeviatemuchfromthewhereHC.A-(0.8CCj)I-R7Tjactualpopulationofthebuildingandtheycanchangesignificantlywithinaworkingday.ThebuildingisservedCC,isthecontractcapacityofthejthcar.Here,p,isthebyagroupofmnumbersofelevatorcars.AllelevatorcarsnumberofpassengersineachroundtripanditisnotsimplycanservicealltheN+Ifloors.Forsimplicity,itisassumedequalto80%ofthecontractcapacitybecauseitverymuchthateveryzonedoesnotoverlapwithanotherzone,dependsonthenumberoffloorsbeingservicedbythejthalthoughazonemay,besidestheMT,consistofoneuppercar.Ifthearrivalrateexceedsthehandlingcapacity,theflooronly.Also,noduplicatezonewhichmeanstwocarsnumberofpassengersinsideeachcarwillberestrictedtoservicethesamezoneisallowed.Everycarservicesthetheoverallcontractcapacityofthecar.Thereasonofusingmainterminalbecausethezoningisdesignedforup-peakthestandarddeviationforallthemRTT’sinequation(2)anddown-peakmodesinsteadofinterfloormodes.istoensureauniformdistributionofRTT’ssothattheNaturally,mnumberofcarscandividethebuildingintomwaitingtimeofmostpassengerscanbemoreorlessnumberofzones.Whenthetrafficisneitherinup-peaknoridentical.Itwillbefairtoeachpassengerwaitingatadown-peakmode,zoningiscancelledandeverycarserviceselevatorlobbyandaskingforservices.Thecostfunctionthewholebuilding.Aclearerpictureofthezoninginequation(2)thusconcentratesonuniformRTT.arrangementisshownbelow:Equation(3)optimisesonthedifferencebetweenhandling1stcarservingO/F(MT),1/F,...,n,/F;capacityanddemandandthusitconcentratesondemandmatching.Theremainingthreeparametersinequations(2)jthcarservingO/F(MT),(n,.,+I)/F,...,and(3),i.e.H,(highestreversalfloorofjthzone),SI6=2,...,m-I);(expectednumberofstopsofjthzone)andp,,arevariablesthatneedtobedeterminedonareal-timebasis.FormthcarservingO/F(MT),(n,,.i+l)/F,...,N/F.simplicity,allcarsareofidenticalcontractcapacityandAstricttuleis:0dynamiczoning).Theaveragetransittimeis79.8s(59.5sFigure2(b)PassengerTransitTimeProfile,Up-peaK,fordynamiczoning)andthelongesttransittimeis149.1sDynamicZoning(123.9sfordynamiczoning).5.2Down-peakSimulationThesimulationcommencesat12:30p.m.whenpeocLrushdowntothemainterminaltoleavethebuildingfo-lunch.Foreveryfiveminutes,100passengersgettotheliftlobbyofeveryfloorandtheyhavegotonec0m.i)~destination,i.e.themainterminal.ThedistributionofY-L2-Lm40passengersisexactlyidenticaltothatoftheup-pea!,situation,theonlydifferencebeingthedirectionchangesfromuptodown.Again,thesimulationwithnozoningisfirstconducted.Figure3(a)showsthepercentageofcaliIrr<*Iansweredwithrespecttotime.Figure3(h)showstheFigurel(a)PassengerWaitingTimeProfile,Up-peak,Nopercentageofcallscompletedwithrespecttotime.Next,Zoningdynamiczoningiscamedout.OwingtothefactthattheItcanbeseenfromFiguresl(a)and2(a)thattherateofpatternofpassengerarrivalfollowsexactlythatoftheup-risingofthecurveismuchhigherwithdynamiczoning.peakconditionandtheequationsarequitesimilar,theThatindicatespassengersneedtowaitforacomparativelyoptimalzoningbasedondemandmatchingreturnsthesameshorterperiodoftime.FromFiguresl(h)and2(b),itcanfloordistribution,i.e.1/Fto7/Fforzone1;8lFto13/Fforheseenthatalmostallpassengershavebeenservedaflerzone2;14/Fto18/Fforzone3andfinally19/Fto20)Fior100swithdynamiczoningversusthe150swithoutzoning.zone4.Thecurvesaredisplayedzonebyzone.1594 respecttotimeforzone3underdown-peakcondition.Figure7(a)showsthepercentageofcallsansweredwithrespecttotimeforzone4underdown-peakcondition.Figure7(b)showsthepercentageofcallscompletedwithrespecttotimeforzone4underdown-peakcondition.IomeI10IWgmIMISIWmFigure3(a)PassengerWaitingTimeProfile,Down-peak,NoZoning0lalaMloIrn1Y1-ISIDmk.(.>FigureS(a)PassengerWaitingTimeProfile,Down-peak,DynamicZoning,Zone20m(0Maam-<.>IMIM,lomFigure3(b)PassengerTransitTimeProfile,Down-peak,NoZoningIC.:I/0joryMYJIrn,mIMIMIrnm&,I>FigureS(b)PassengerTransitTimeProfile,Down-peak,DynamicZoning,Zone2*,I>Figure4(a)PassengerWaitingTimeProfile,Down-peak,DynamicZoning,ZoneIFigure4(a)showsthepercentageofcallsansweredwithrespecttotimeforzone1underdown-peakcondition.Figure4(b)showsthepercentageofcallscompletedwith19.wFigure6(a)PassengerWaitingTimeProfile,Down-peak,respecttotimeforzone1underdown-peakcondition.DynamicZoning,Zone3FigureS(a)showsthepercentageofcallsansweredwithrespecttotimeforzone2underdown-peakcondition.Figure5(b)showsthepercentageofcallscompletedwithrespecttotimeforzone2underdown-peakcondition.I10mryS10Irnm,MaIM110m*