
Veryshort-termwindspeedprediction:Anewarti cialneuralnetwork–Markovchainmodel
S.A.PourmousaviKania,*,M.M.Ardehalib
ab
ElectricalandComputerEngineeringDepartment,627CobleighHall,MontanaStateUniversity,Bozeman,MT59717,USA
EnergyResearchCenter,DepartmentofElectricalEngineering,AmirkabirUniversityofTechnology(TehranPolytechnic),424HafezAve.,Tehran15914,Iran
articleinfoabstract
Astheobjectiveofthisstudy,arti cialneuralnetwork(ANN)andMarkovchain(MC)areusedtodevelopanewANN–MCmodelforforecastingwindspeedinveryshort-termtimescale.Forpredictionofveryshort-termwindspeedinafewsecondsinthefuture,datapatternsforshort-term(aboutanhour)andveryshort-term(aboutminutesorseconds)recordedpriortocurrenttimeareconsidered.Inthisstudy,theshort-termpatternsinwindspeeddataarecapturedbyANNandthelong-termpatternsareconsideredutilizingMCapproachandfourneighborhoodindices.TheresultsarevalidatedandtheeffectivenessofthenewANN–MCmodelisdemonstrated.Itisfoundthatthepredictionerrorscanbedecreased,whiletheuncertaintyofthepredictionsandcalculationtimearereduced.
Ó2010ElsevierLtd.Allrightsreserved.
Articlehistory:
Received8December2009Accepted30July2010
Availableonline21August2010Keywords:
Arti cialneuralnetworkMarkovchainapproachVeryshort-termpredictionWindspeedprediction
1.Introduction
Thecontributionofwindpowerinmarket-drivenpowersystemstogetherwiththeuncertainnatureofwindhaveledtoresearchonmodelstopredictfuturewindspeed/powerproduction.Accurateveryshort-termwindspeed/powerpredictioncanreducevoltageandfrequency uctuationsduetovariationinwindpowerandunacceptableshocksintheconventionalpowerunitscausedbyasuddencut-offofwindpowerresultingfromexcessivewindspeeds[1,2].Also,incontrolsystemsofawindturbine,theimpor-tanceofshort-termdecisionssuchasconnectingaload,changingthepitchofthebladesandothercontrolactionsinvolvingdelaysprompttheneedforveryshort-termwindspeedprediction[3–5].Further,inlargemodernwindturbineswithhighinvestmentcosts,itisimportantforagenerationcompanytoextractmaxi-mumavailableenergyfromwindspeedandapplicationofdiffer-entcontrolalgorithmsfordifferentoperationalregimesoftenbecomenecessary.Ithasbeenshownthatthemodelpredictivecontrol(MPC)isaneffectivealgorithmthatcanbeusedtocontrolwindturbinesinalltheoperationalregimesanditisconstructedbasedonpastmeasurementsandcomputedfutureinputs[6].Therefore,itisanticipatedthatveryshort-termwindspeedpredic-tionwill ndawiderangeofapplications,whenMPCisutilized.Inastudy,apitchcontrolledtechniqueisproposedforgrid-con-nectedwindturbineinasmallpowersystem[7].Theproposed
*Correspondingauthor.Tel.:+14069945965,mobile:+14065481969.
E-mailaddresses:s.pourmousavikani@msu.montana.edu(S.A.PourmousaviKani),ardehali@aut.ac.ir(M.M.Ardehali).

0196-8904/$-seefrontmatterÓ2010ElsevierLtd.Allrightsreserved.doi:10.1016/j.enconman.2010.07.053
pitchcontrollerinthatstudy,incorporatedwithveryshort-termwindspeedprediction,isappliedtominimizetheeffectsonthepowersystem,whileproducingoptimumwind-generatedpower.Inanotherstudy,anoperationaloptimizationstrategyisintro-ducedatthewindparkcontrollevel[8],whereveryshort-termwindspeedforecastsareknownforindividualwindturbines.Thenotedoperationalstrategyisdevelopedwithaconcerntotheminimizationoftheconnection/disconnectionchangesoftheindividualwindgenerators,foragiventimehorizon.
Theshort-termwindspeedpredictionmodelscanbegenerallysubdividedintotwomaingroups,dependingontheunderlyingpredictionalgorithmused.The rstcategoryisbasedonnumericalweatherprediction(NWP)models,similartothoseusedbyna-tionalmeteorologicalagencies,orotheralternativeapproaches[9,10].Thesecondcategoryencompasses,amongothers,arti cialintelligence(AI)(fuzzylogic[11–13],arti cialneuralnetworks(ANN)[2,5,14–23])andstatisticalmodels[1,4,12,19,24–27].Also,somepost-processingapproachesbasedonstatisticalmodelsareintroducedasahybridalgorithminconjunctionwithNWPmodels[9,10,12,21].WhiletheNWP-basedmodelsbeingusuallythebestoptionforlongertermpredictions(overdays,weeksandmonthsahead),AIandstatisticalapproachesarethemostpromisingmod-els[19,28],whenshort-termanditssubclassestimehorizons(overaroundseconds,minutesandfewhoursahead)predictionsareused[19,28].Forstatisticalapproaches,ithasbeendeterminedthattheyhavede cienciesinveryshort-termwindspeedpredic-tion[1,12].First,mostofstatisticalapproachesassumethatthewindspeeddataisnormallydistributed.However,itisawellknowncharacteristicofgeneralwindspeedseriesthatitsvariation
Short-Term Prediction of Wind Farm Power_能源/...a recurrent neural network (NN) based on the ...Negnevitsky, “Very short-term wind forecasting ...
Innovative Short-Term Wind Generation Prediction ...Neural Fuzzy Inference System (ANFIS) to ...Index Terms — Windpower, Very Short-Term ...
one hour ahead prediction of wind speed based ...[5], artificial neural network [6], fuzzy ...In short-term wind speed forecasting, the data ...
Artificial neural network prediction of mechanical properties of hot rolled low carbon steel strip_天文/地理_自然科学_专业资料。...
A Prediction Methods of ShortTerm Wind Power in Wind Power Plant CHEN ... windspeed prediction; artificial neural network; kalman filter; time series ...
The State-Of-The-Art in Short-Term Prediction of Wind Power-A Literature Overview 隐藏>> Project funded by the European Commission under the 5th (EC...
and Michael Negnevitsky, Member, IEEE Abstract—This paper describes very short-term wind prediction for power generation, utilizing a case study from ...
method of traffic flow parameters prediction -- short-term prediction of traffic flow parameters based on multi-scale analysis and artificial neural network....
In the second stage, the wind speed to output ... transmission operators, network managers, and ...Short-term wind power prediction contributes to ...
A Prediction Methods of ShortTerm Wind Power in Wind Power Plant CHEN ... windspeed prediction; artificial neural network; kalman filter; time series ...
热门文档