PowersystemloadforecastingmethodsandcharacteristicsofAbstract：Theloadforecastinginpowersystemplanningandoperationplayanimportantrole，withobviouseconomicbenefits，inessence,theelectricityloadforecastingmarketdemandforecast。Inthispaper,asystematicdescriptionandanalysisofavarietyofloadforecastingmethodsandcharacteristicsandthatgoodloadforecastingforpowersystemhasbecomeanimportantmeansofmodernmanagement。Keywords：powersystemloadforecastingelectricitymarketconstructionPlanning1。IntroductionLoadforecastingdemandforelectricityfromaknownstartingtoconsiderthepolitical，economic,climateandotherrelatedfactors，thefuturedemandforelectricitytomakepredictions.Loadforecastincludestwoaspects：onthefuturedemand（power）projectionsandfutureelectricityconsumption（energy）forecast.Electricitydemandprojectionsdecisiongeneration,transmissionanddistributionsystem，thesicofnewCapacity;powergeneratingequipmentdeterminethetypeofprediction（.suchaspeakingunits,baseloadunits,etc｝。Loadforecastingpurposesistoprovideloadconditionsandthelevelofdevelopment，whileidentifyingthevarioussupplyareas,eachyearplanningforthepowerconsumptionformaximumpowerloadandtheloadofplanningtheoveralllevelofdevelopmentofeachplanyeartodeterminetheloadcomposition.loadforecastingmethodsandcharacteristicsof2。1UnitConsumptionActOutputofproductsinaccordancewithnationalarrangements,planningandelectricityintensityvaluetodetermineelectricitydemand.Sub—UnitConsumptionAct;ProductUnitConsumption;andthevalueofUnitConsumptionAct；two。Theprojectionofloadbeforethekeyistodeterminetheappropriatevalueoftheproductunitconsumptionorunitconsumption.JudgingfromChina'sactualsituation，thegeneralruleistheproductunitconsumptionincreasedyearbyyear,theoutputvalueunitconsumptionisdeclining。Unitconsumptionmethodadvantagesarc:Themethodissimple，short-tornloadforecastingeffective.Disadvantagesarc:needtodoalotofpainstakingresearchwork,moregeneral，itisdifficulttoreflectmoderneconomic，politicalandclimateconditions。2。2TrendextrapolationWhenthepowerloadinaccordancewithtime—varyingpresentsamekindofupwardordownwardtrend,andnoobviousseasonalfluctuations，butalsotofindasuitablefunctioncurvetoreflectthischangeintrend,youcanusethetimetasindependentvariables,timingvalueofyforthedependentvariabletoestablishthetrendmodely=f（t）。Whenthereasontobelievethatthistrendwillextendtothefuture,weassignedthevalueofthevariabletneedto，youcangetthecorrespondingtuneseriesofthefuturevalueofthemoment。Thisisthetrendextrapolation。Applicationofthetrendextrapolationmethodhastwoassumptions:（1）assumingthereisnostepChangeinload；（2）assumethatthedevelopmentofloadfactorsalsodeterminethefuturedevelopmentofloadanditsconditionisunchangedorchangedlittle.Selecttheappropriatetrendmodelistheapplicationofthetrendextrapolationanimportantpartofpatternrecognitionmethodandfinitedifferencemethodistoselectthetrendmodelarctwobasicways。Alineartrendextrapolationforecastingmethod,thelogarithmictrendforecastingmethod，quadraticcurvetrendforecastingmethod,exponentialcurvetrendforecastingmethod,growthcurveofthetrendpredictionmethod.Trendextrapolationmethod'sadvantagesarc:onlyneedtohistoricaldata,theamountofdatarequiredforless。Thedisadvantageisthat:Ifachangeinloadwillcauselargeerrors。2。3ElasticCoefficientMethodElasticitycoefficientistheaveragegrowthrateofelectricityconsumptiontoGDPratioofbetween,accordingtothegrossdomesticproductgrowthrateofcoefficientofelasticitytobeplanningwiththeendofthetotalelectricityconsumption。Modulesofelasticitylawisdeterminedonpowerdevelopmentfromamacrowiththerelativespeedofnationaleconomicdevelopment，whichisameasureofnationaleconomicdevelopmentandanimportantparameterinelectricitydemand.Theadvantagesofthismethodarc：Themethodissimple,easytocalculate。Disadvantagesarc：needtodoalotofdetailedresearchwork.2.4RegressionAnalysisMethodRegressionestimateisbasedonpasthistoryofloaddata,buildupamathematicalanalysisofthemathematicalmodel.Ofmathematicalstatisticsregressionanalysisofthevariablesinstatisticalanalysisofobservationaldatainordertoachieveloadtopredictthefuture.Regressionmodelwithalinearregression,multiplelinearregression，nonlinearregressionandotherregressionpredictionmodels.Amongthem，linearregressionforthemedium—torntoadforecast.Advantagesarc：ahigherpredictionaccuracyforthemediumandtheuseofshort-termforecasts.Thedisadvantageisthat:（1）planninglevelitisdifficultyearsofindustrialandagriculturaloutputstatistics；（2）regressionanalysiscanonlybemeasuredoutthelevelofdevelopmentofanintegratedelectricityloadcannotbemeasuredoutthepowersupplyareaoftheloadinglevelofdevelopment,thuscannotbethespecificgridconstructionplan。2.5TimeSeriesAnalysisTheloadisonthebasisofhistoricaldata,tryingtobuildamathematicalmodel，usingthismathematicalmodeltodescribethepowerloadontheonehandthisrandomvariableofstatisticalregularityofthechangeprocess;theotherhand，themathematicalmodelbasedonthere-establishmentofthemathematicalexpressionofloadforecastingtype，topredictthefutureload.TimeseriesaremainlyautoregressiveAR（p）,movingaverageMA（q）andself—regressionandn3ovingaverageARMA（p，q）andsoon.Theadvantagesofthesemethodsarc：thehistoricaldatarequiredforless，workless。Thedisadvantageisthat：Thereisnochangeinloadfactortoconsider，onlydedicatedtothedatafitting,thelackofregularityoftreatmentisonlyapplicabletorelativelyuniformchangesintheshort—termloadforecastingsituation.2。6GraymodelmethodGraypredictionisakindofasystemcontaininguncertainfactorstopredictapproach.Graysystemtheorybasedonthegrayforecastingtechniquesmaybelimitedcircumstancesinthedatatoidentifytheroleoflawwithinacertainperiod，theestablishmentofloadforecastingmodels。Isdividedintoordinarygraysystemmodelandoptimizationmodelfortwokindsofgray.Ordinarygraypredictionmodelisanexponentialgrowthmodel,whentheelectricloadinstrictaccordancewithexponentiallygrowing，thismethodhashighaccuracyandrequiredlesssampledatatocalculatesimpleandtestableetc。;drawbackisthatforachangeinvolatilityThepowerload,thepredictionerrorlargo，doesnotmeetactualneeds。Andthegraymodeloptimizationcanhaveupsanddownsoftheoriginaldatasequencetransformedintoincreasedexponentiallyincreasingregularitychangesinsequence，greatlyimprovingpredictionaccuracyandthegraymodelmethodofapplication。GrayModelLawappliestoshort-tornloadforecast。Graypredictedadvantages：smallerloaddatarequirements,withoutregardtothedistributionoflawsanddonottakeintoaccounttrends,computingconvenient,short—termforecastsofhighprecision，easytotest.Drawbacks：First，whenthedatathegreaterthedegreeofdispersion,namely,thegreaterthegrayleveldata，predictionaccuracyisworse;2isnotverysuitableforthelong-termpowersystemtopushanumberofyearsaftertheforecast.2。7DelphiMethodTheDelphimethodisbasedonthespecialknowledgeofdirectexperience，researchproblemsofjudgment，amethodforpredictionof,alsocalledexpertsinvestigation.Delphimethodhasfeedback,anonymityandstatisticalcharacteristics.Delphimethodadvantageis：(1)canacceleratepredictionspeedandsavepredictionCost;(2)cangetdifferentbutvaluableideasandopinions；(3)suitableforlong-termforecastsinhistoricaldata,insufficientorunpredictablefactorsisparticularlyapplicablemore。Detectis:(1)theloadforecastingfarpointsareamaynotreliable；(2)theexpertopinionssometimesmaynotcompleteorimpractical.ExpertSystemApproachExpertsystempredictionisstoredinthedatabaseoverthepasttowyears，evendecades,theHourlyloadandweatherdataanalysis，whichbringstogetherexperiencedstaffknowledgeloadforecasting,extracttherelevantrules,accordingtocertainrules,loadprediction。Practicehasprovedthataccurateloadforecastingrequiresnotonlyhigh-techsupport,butalsoneedtoreconciletheexperienceandwisdomofmankinditself:Therefore,youneedexpertsystemssuchtechnologies.Expertsystemsapproachisanon-quantifiablehumanexperiencetranslatedintoabetterwayButexpertssystemsanalysisitselfisatime—consumingprocess,andsomecomplexfactors(suchasweatherfactors),eventhoughawareofitsloadimpact,ht}ttoaccuratelyandquantitativelydeterminetheirinfluenceontheloadareaisalsoverydifficult.Expertsystemforforecastingmethodsuitableformediumandlong—termloadforecast。Theadvantagesofthismethod:(1)canbringtogethermultipleexpertknowledgeandexperiencetomaximizetheabilityofexperts；(2)possessionofdata,informationandmortfactorstoconsideramorecomprehensiveandbeneficialtoarriveatmartaccurateconclusions。Thedisadvantageisthat：(1)donothavetheself-learningability,subjecttotheknowledgestoredinthedatabaselimitsthetotal；(2)pairsofunexpectedincidentsandpooradaptabilitytochangingconditionsNeuralNetworkMethodNeuralnetwork(ANN,ArtificialNeuralNetwork)forecastingtechniquestomimicthehumanbraintodointelligentprocessing，alargenumberofnon—structural.non-deterministiclawsofadaptivefunction.ANNusedinshort-termloadforecastingandlong-termloadforecastthanthatappliedtobemartappropriate.Becauseshort—termloadchangescanberegardedasastationaryrandomprocess.Andlong—termloadforecastingmaybeduetopolitical，economicandothermajorfumingpointleadingtoamathematicalmodel-baseddamage.Advantagesarc:(1)tomimicthehumanbrain，intelligenceprocessing;(2}alargenumberofnon—structural.non—adaptivefunctionoftheaccuracyofthelaw；(3)withtheinformationmemory,self—learning,knowledge，reasoningandoptimizationofcomputingfeatures.Thedisadvantageisthat:(1)thedeterminationoftheinitialvaluecannottakeadvantageofexistingsysteminformation，easilytrappedinlocalminimumofthestate；(2)neuralnetworklearningprocessisusuallyslow,pooradaptabilitytosuddenevents.OptimumCombinationForecastingMethodOptimalcombinationhastwomeanings：First,severalforecastingmethodsfromtheresultsobtainedbyselectingtheappropriatea0cightintheweightedaverage；2referstothecomparisonofseveralpredictionmethods，choosethebestorthedegreeofpreparationandthestandarddeviationofthesmallestpredictionmodelforecast。Forthecombinedforecastingmethodmustalsonotedthatthecombinedforecastisasingleforecastingmodelcannotcompletelycorrecttodescribethechangesoftheamountpredictedtoplayarole.Onecanfullyreflecttheactuallawofdevelopmentofthemodelpredictionsagreewellwiththecombinationforecastingmethodthanpredictedgoodresults。Thismethodhastheadvantage：Tooptimizethecombinationofawiderangeofinformationonasinglepredictionmodel，considertheimpactofinformationisalsomartcomprehensive，soitcaneffectivelyimprovetheprediction.Thedisadvantageisthat:(1)theweightisdifficulttodetermine;(2)allpossiblefactorsthatplayaroleinthefuture，allincludedinthemodel，toacertainextent，limitthepredictionaccuracyimproved。2。11WaveletanalysisandforecastingtechniquesWaveletanalysisisatime—domain-frequencydomainanalysismethod,itisinthetimedomainandfrequencydomainatthesametimehasgoodlocalizationproperties，andcanautomaticallyadjustaccordingtothesignalsamplingfrequencyofhighandlowdensity,itiscast'tocaptureandanalysisofweaksignalsandsignal，imagesofanysmallparts.Theadvantageis：Canthedifferentfrequencycomponentsgraduallyrefinedusingasamplingstep，whichcanbegatheredinanyofthedetailsofthesignal,especiallyforsingularsignalisverysensitivetothetreatmentwellormutationweaksignals,theirgoalistoasignalinformationintowaveletcoefficients，whichcaneasilybedealtwith，storage，transmission，analysisorforthereconstructionoftheoriginalsignal.Theseadvantagesdeterminethewaveletanalysescanbeeffectivelyappliedtoloadforecastingissues。ConclusionLoadforecastingistheelectricpowersystemscheduling，real—timecontrol，operationplananddevelopmentplanning,thepremiseisagriddispatchingdepartmentsandplanningdepartmentsmusthavethebasicinformation.Improveloadforecastingtechnologylevel，behelpfulforprogrammanagement,reasonablearrangementoftheelectricitygridoperationmodeforthemaintenanceplanandthecrew，tosectioncoal，fuel—efficientandreducegeneratingcost,behelpfulforformulaterationalpowerconstructionplanningofthepowersystem,improvetheeconomicbenefitandsocialbenefit.Therefore，theloadforecasthasbecomeapowersystemmanagementmodernizationrealizationoftheimportantcontent。電力系統(tǒng)負荷預(yù)測及方法摘要：負荷預(yù)測在電力系統(tǒng)規(guī)劃和運行方面發(fā)揮的重要作用,具有明顯的經(jīng)濟效益，負荷預(yù)測實質(zhì)上是對電力市場需求的預(yù)測.該文系統(tǒng)地介紹和分析了各種負荷預(yù)測的方法及特點，并指出做好負荷預(yù)測己成為實現(xiàn)電力系統(tǒng)管理現(xiàn)代化的重要手段。關(guān)鍵詞：電力系統(tǒng)負荷預(yù)測電力市場建設(shè)規(guī)劃引言負荷預(yù)測是從已知的用電需求出發(fā)，考慮政治、經(jīng)濟、氣候等相關(guān)因素，對未來的用電需求做出的頂測。負荷預(yù)測包括兩方而的含義:對未來需求量（功率）的頂測和未來用電量（能量）的頂測。電力需求量的預(yù)測決定發(fā)電、輸電、配電系統(tǒng)新增容量的大?。浑娔茴A(yù)測決定發(fā)電設(shè)備的類型（如調(diào)峰機組、基荷機組等）。負荷預(yù)測的日的就是提供負荷發(fā)展狀況及水平，同時確定各供電區(qū)、各規(guī)劃年供用電量、供用電最大負荷和規(guī)劃地區(qū)總的負荷發(fā)展水平,確定各規(guī)劃年用電負荷構(gòu)成.2。負荷預(yù)測的方法及特點2。1單耗法按照國家女排的產(chǎn)品產(chǎn)量、產(chǎn)值計劃和用電單耗確定需電量.單耗法分“產(chǎn)品單耗法”和“產(chǎn)值單耗法”兩種.采用“單耗法”預(yù)測負荷前的關(guān)鍵是確定適當?shù)漠a(chǎn)品單耗或產(chǎn)值單耗。從我國的實際情況來看，一般規(guī)律是產(chǎn)品單耗逐年上升，產(chǎn)值單耗逐年卜降。單耗法的優(yōu)點是:方法簡單，對短期負荷預(yù)測效果較好。缺點是：需做大量細致的調(diào)研工作，比較籠統(tǒng)，很難反映現(xiàn)代經(jīng)濟、政治、氣候等條件的影響。2.2趨勢外推法當電力負荷依時間變化呈現(xiàn)某種上升或下降的趨勢，并且無明顯的季節(jié)波動，又能找到一條合適的函數(shù)曲線反映這種變化趨勢時，就可以用時間t為自變量，時序數(shù)值y為因變量,建立趨勢模型Y=f（t）。當有理由相信這種趨勢能夠延伸到未來時，賦予變量t所需要的值,可以得到相應(yīng)時刻的時間序列米來值。這就是趨勢外推法。應(yīng)用趨勢外推法有兩個假設(shè)條件：（1）假設(shè)負荷沒有跳躍式變化；（2）假定負荷的發(fā)展因素也決定負荷未來的發(fā)展，其條件是不變或變化不大。選擇合適的趨勢模型是應(yīng)用趨勢外推法的重要環(huán)節(jié),圖形識別法和差分法是選擇趨勢模型的兩種基本方法.外推法有線性趨勢預(yù)測法、對數(shù)趨勢頂測法、二次曲線趨勢頂測法、指數(shù)曲線趨勢預(yù)測法、生長曲線趨勢預(yù)測法。趨勢外推法的優(yōu)點是:只需要歷史數(shù)據(jù)、所需的數(shù)據(jù)量較少。缺點是：如果負荷出現(xiàn)變動,會引起較大的誤差.2。3彈性系數(shù)法彈性系數(shù)是電量平均增長率與國內(nèi)生產(chǎn)總值之間的比值，根據(jù)國內(nèi)生產(chǎn)總值的增長速度結(jié)合彈性系數(shù)得到規(guī)劃期末的總用電量。彈性系數(shù)法是從宏觀上確定電力發(fā)展同國民經(jīng)濟發(fā)展的相對速度，它是衡量國民經(jīng)濟發(fā)展和用電需求的重要參數(shù).該方法的優(yōu)點是：方法簡單，易于計算。缺點是：需做大量細致的調(diào)研工作。2。4回歸分析法回歸預(yù)測是根據(jù)負荷過去的歷史資料,建立可以進行數(shù)學分析的數(shù)學模型.用數(shù)理統(tǒng)計中的回歸分析方法對變量的觀測數(shù)據(jù)統(tǒng)計分析，從而實現(xiàn)對未來的負荷進行預(yù)測。回歸模型有一元線性回歸、多元線性回歸、非線性回歸等回歸預(yù)測模型.其中，線性回歸用于中期負荷預(yù)測。優(yōu)點是：頂測精度較高，適用于在中、短期頂測使用。缺點是:（1）規(guī)劃水平年的工農(nóng)業(yè)總產(chǎn)值很難詳細統(tǒng)計;（2）用回歸分析法只能測算出綜合用電負荷的發(fā)展水平，無法測算出各供電區(qū)的負荷發(fā)展水平,也就無法進行具體的電網(wǎng)建設(shè)規(guī)劃.2。5時間序列法就是根據(jù)負荷的歷史資料，設(shè)法建立一個數(shù)學模型,用這個數(shù)學模型一方面來描述電力負荷這個隨機變量變化過程的統(tǒng)計規(guī)律性;另一方而在該數(shù)學模型的基礎(chǔ)上再確立負荷預(yù)測的數(shù)學表達式,對未來的負荷進行預(yù)測。時間序列法主要有自回歸AR(p).滑動平均ma(q)和自回歸與滑動平均ARMA(p,q)等。這些方法的優(yōu)點是:所需歷史數(shù)據(jù)少、工作量少.缺點是：沒有考慮負荷變化的因素，只致力于數(shù)據(jù)的擬合，對規(guī)律性的處理不足，只適用于負荷變化比較均勻的短期頂測的情況。2。6灰色模型法灰色預(yù)測是一種對含有不確定因素的系統(tǒng)進行預(yù)測的方法.以灰色系統(tǒng)理論為基礎(chǔ)的灰色頂測技術(shù)，可在數(shù)據(jù)不多的情況下找出某個時期內(nèi)起作用的規(guī)律，建立負荷頂測的模型，分為普通灰色系統(tǒng)模型和最優(yōu)化灰色模型兩種。普通灰色預(yù)測模型是一種指數(shù)增長模型,當電力負荷嚴格按指數(shù)規(guī)律持續(xù)增長時，此法有預(yù)測精度高、所需樣本數(shù)據(jù)少、計算簡便、可檢驗等優(yōu)點；缺點是對于具有波動性變化的電力負荷，其預(yù)測誤差較大，不符合實際需要。向最優(yōu)化灰色模型可以把有起伏的原始數(shù)據(jù)序列變換成規(guī)律性增強的成指數(shù)遞增變化的序列，大大提高頂測精度和灰色模型法的適用范圍?；疑Ｐ头ㄟm用于短期負荷預(yù)測?；疑A(yù)測的優(yōu)點：要求負荷數(shù)據(jù)少、不考慮分布規(guī)律、不考慮變化趨勢、運算方便、短期頂測精度高易于檢驗.缺點：一是當數(shù)據(jù)離散程度越大，即數(shù)據(jù)灰度越大，預(yù)測精度越差；二是不太適合于電力系統(tǒng)的長期后推若干年的預(yù)測。2.7德爾菲法德爾菲法是根據(jù)有專門知識的人的直接經(jīng)驗，對研究的問題進行判斷、預(yù)測的一種方法,也稱專家調(diào)查法。德爾菲法其有反饋性、展名性和統(tǒng)計性的特點。德爾菲法的優(yōu)點是:(1)可以加快頂測速度和節(jié)