The State-Of-The-Art in Short-Term Prediction of W

ProjectfundedbytheEuropeanCommissionunderthe5th(EC)RTDFrameworkProgramme(1998-2002)withinthethematicprogrammeEnergy,EnvironmentandSustainableDevelopmentProjectANEMOSContractNo.:ENK5-CT-2002-00665“DevelopmentofaNextGenerationWindResourceForecastingSystemfortheLarge-ScaleIntegrationofOnshoreandOffshoreWindFarms”TheState-Of-The-ArtinShort-TermPredictionofWindPowerALiteratureOverviewVersion1.1AUTHOR:GregorGiebelAFFILIATION:RisøNationalLaboratoryADDRESS:P.O.Box49,DK-4000RoskildeTEL.:[phone]EMAIL:[e-mail]:RichardBrownsword,RAL;GeorgeKariniotakis,ARMINESREVIEWER:ANEMOSWP-1membersAPPROVER:GregorGiebelDocumentInformationDOCUMENTTYPEDeliverablereportD1.1DOCUMENTNAME:ANEMOS_D1.1_StateOfTheArt_v1.1.pdfREVISION:5REV.DATE:2003.08.12CLASSIFICATION:PublicSTATUS:ApprovedAbstract:Basedonanappropriatequestionnaire(WP1.1)andsomeotherworksalreadyinprogress,thisreportdetailsthestate-of-the-artinshorttermpredictionofwindpower,mostlysummarisingnearlyallexistingliteratureonthetopic.ANEMOSANEMOS_D1.1_StateOfTheArt_v1.1.pdf,2003.08.12.DeliverableD1.12/36Contents1.Introduction.........................................................................................................31.1Thetypicalmodelchain...................................................................................31.2Typicalresults.................................................................................................62.Literatureoverview.............................................................................................72.1Timeseriesmodelsforuptoafewhours........................................................72.1.1Directtimeseriesmodels........................................................................................82.1.2Modellingwindspeedversuswindpower................................................................82.1.3Neuralnetworks.......................................................................................................92.1.4Anexplanationofthetimeseriesmodelimprovements..........................................92.2NumericalWeatherPrediction-basedmodels.................................................102.2.1Modelsnolongerorneverinaction.......................................................................102.2.2Researchmodels...................................................................................................112.2.3Modelscurrentlyinuse..........................................................................................132.2.4TheNorrköpingworkshop......................................................................................162.3Evaluationofforecastingmodels....................................................................172.4Uncertaintyofwindpowerpredictions............................................................182.5Ensembleforecasts........................................................................................192.6Thevalueofforecasting.................................................................................202.7Demandsonforecastingmodels....................................................................223.TheANEMOSproject........................................................................................234.Concludingremarks...........................................................................................255.Acknowledgements...........................................................................................266.Glossary............................................................................................................267.References........................................................................................................27NOTE:UpdatesofthisReportwillbemadeavailableinthefutureattheprojectwebsite().Thenextupdate(Version2.0)willfollowthereleaseofthe2003EWECConferenceProceedings.ANEMOSANEMOS_D1.1_StateOfTheArt_v1.1.pdf,2003.08.12.DeliverableD1.13/361.IntroductionThisreportwillgiveanoverviewoverpastandpresentattemptstopredictwindpowerforsingleturbinesorforwholeregions,forafewminutesorafewdaysahead.IthasbeenproducedfortheANEMOSproject[1],whichbringstogethermanygroupsfromEuropeinvolvedinthefield,withupto15yearsofexperienceinshort-termforecasting.Theliteraturesearchinvolvedhasbeenextensive,anditishopedthatthisreportcanserveasareferenceforallfurtherwork.Oneofthelargestproblemsofwindpower,ascomparedtoconventionallygeneratedelectricity,isitsdependenceonthevolatilityofthewind.Thisbehaviourhappensonalltimescales,buttwoofthemaremostrelevant:Oneisfortheturbinecontrolitself(frommillisecondstoseconds),andtheotheroneisimportantfortheintegrationofwindpowerintheelectricalgrid,andthereforedeterminedbythetimeconstantsinthegrid(fromminutestoweeks).Onecandistinguishthefollowingtypesofapplications:•Optimisationoftheschedulingoftheconventionalpowerplantsbyfunctionssuchaseconomicdispatchetc.Thepredictionhorizonscanvarybetween3-10hoursdependingonthesizeofthesystemandthetypeofconventionalunitsincluded(i