附錄SolarEnergyCopyrightElsevierScienceDATASAMPLINGSPEEDVERSUSENERGETICMEASUREMENTERRORSOFIRRADIATIONMONITORINGINPHOTOVOLTAICTokyoUniversityofAgricultureandTechnology,2-24-16,Nakamachi,Koganei-shi,Tokyo184,Japan.(CommunicatedbyGERARDWRIXON.)Abstract:Tomeasuresolarirradiationandphotovoltaicarrayoutputenergyameasuringaccuracycannotbeguaranteedunlessthedatasamplingintervalisappropriatelyselected.Fromthisviewpoint,actualirradiancehasbeenmeasuredbycomparativelyhighspeedsamplingofl-4sfor44monthsandthedailyerrorsofthenumericalintegralhavebeenestimatedforvariousstepsizes.Approximationformulaeoftheerrorversusthestepsizehavebeenstatisticallyobtainedaswellastheirprobabilitydensityfunctioncovering+4c.Finally,anomographispresentedtodecideanappropriatesamplinginterval.Aconcludingexampleshowsthatthedeviatingcomponentoftheerrorexceeding+1%canhappenonceforevery1monthor6monthsifthestepsizeisselectedas105or65.5s,ineachofwhichthetotalerrorbecomes-0.0485k1%or-0.0336&1%includinganaverageerrorcomponentaccordingtothedatameasuredatTsukuba.Copyright1996ElsevierScienceLtd.Introduction:Inaphotovoltaicsystemtheinputsolarenergyhasabasicday-and-nightcycleandanuncertain,timevaryingfactorcausedbymeteorologicalconditions.Whenameasurementisperformedtoobtainlong-termperformanceparameterssuchasirradiation,theamountofgeneratedenergy,etc.,theaccuracyofthemeasurementmaynotbemaintainedifthesamplingspeedofdataacquisitionisdecidedregardlessofthefluctuatingrateofsolarradiation.Sincesimplifiedmeasurementsmayhavetobeadoptedinthefuturewiththespreadofphotovoltaicsystems,someinstructionsmustbeclearlypreparedasasystemmonitoringstandard.Theauthorpresentsananalysisofthisneed.Forthispurpose,actualirradianceprofilehasbeenmeasuredbycomparativelyhighspeedsamplingfor44monthsattheElectro-technicalLaboratoryinTsukubaScienceCity(KurokawaandMine,1989；Kurokawa,1994).Byusingthesedata,thedailyerrorsofnumericalintegrationhavebeenevaluatedforvarioustimestepsizes.Then,theerrorsforeachstepsizehavebeenanalyzedforacertaintermtoobtainstatisticalparameterssuchasstandarddeviation0andaveragevalue.Approximationformulaehavebeenalsogiventoexpressrelationshipsbetweenthesevaluesandstepsizes.Inaddition,theprobabilitydensityfunctionprofileoftheerrorshasbeenalsoestimated.Finally,byapplyingtheapproximationformulaeandtheprobabilitydensityfunction,anomographispresentedfordeterminingamaximumallowablestepsizetoassuremonitoringprecision.DataAcquisition:Inordertoknowthetrueintegralofirradianceprofiles,originalirradiancedataweretakenatacomparativelyhighspeedsamplingratefor44monthsfromSeptember1986toApril1990atTsukubaScienceCity.Thecityislocated60kmnorthofTokyoanditsclimaticconditionsseemtobeordinaryasfarasJapanisconcerned.MeasuringequipmentusedisillustratedinFig.1.Noises,whichmaybegeneratedinsignalconditioners,werealsosuppressedverycarefullybyusinganisolationamplifierandelectronicfilter.Inaddition,adigitalaveragingtechniquesynchronizedwiththeutilitygridfrequencywasadoptedanddrasticallydecreasedthenoiselevel.Morethan100dataweresampledandaveragedforjust40ms,whichcorrespondsto2cyclesofthefrequency.Thisbecomesonesampleofrawdataforthisstudy.TotaldataacquisitionflowisillustratedinFig.2.Thenormalrawdatasamplingperiodwas1sbutitwasmadelonger,upto4s,incaseofaslowerfluctuatingspeedoftheirradiancetoreducedatavolume.Dailydatawerestoredina2HDdiskforeveryday.Byapplyingthetrapezoidalruletothemeasuredirradiancedata,typicalexamplesofwhichareshowninFigs3(a)and4(a),theintegratedvalueofirradiance,i.e.irradiation,siisgivenby.Theintegratingintervalisindicatedbyh.Iterationofeqn(1)givesasdailyintegralofSiuntilti=sunset.Thevariousintervalsaredenotedashjforj=l-60.Thesamplingintervaloftheoriginaldataisho.Therangeofhicorrespondsto10sto2hlogarithmicallyasshowninTable1.S,andSjaregivenasSforh=hoandh=hi,respectively.Then,percentintegrationerror,ejisdefinedby.ThestatisticalvaluesobtainedforacomparativelyshorttermareapttoshowsomeirregularityinthetrendsofstandarddeviationsaspresentedinFig.6.However,itisobservedthattheoverallresultsforthewhole44monthsgaveenoughsmoothnessasshowninFig.9.Therefore,the44monthresultsaretobestudiedasdescribedbelow.ByusingthedatainFig.9,approximationformulaewereformed.AsshowninFigs10and11,therelationshipshowingandcsversushjisdeterminedby.Toobtainenoughprecisionforbothsmallerhregionsandlargerones,two-regionapproximationwasadopted.Forexample,itgivesm=0.0315andd=0.220forh=60andtheirratiostostatisticaldataare1.07and0.980,respectively.Althoughtheformerlooksslightlylarge,thedataofmcontainlargerirregularities.Theyseemtobeareasonableapproximationfordatasmoothing.Probability:Asatentativeconclusionoftheprecedingstatistics,itisunderstoodthatthestandarddeviationofintegrationerrorsismostlyanoteworthyfactorfordetermininganappropriateintegrationinterval.However,itsprobabilisticmeaningisnotpreciselyknownsofar.Thissectionstudiesit.Thefrequencyoftheerrorclassescovering+40atrespectiveintervalshasbeenplottedashistogramsasshowninFigs12-14.Figure12isgivenforh=60s,Fig.13forh=600sandFig.14forh=3600s.Fromthesefigures,itissupposedthattheirdistributionshapesarefairlysimilarinallintervals.Therefore,eachhistogramissuperimposedbynormalizingtheintegrationerrorswitheachstandarddeviation6.ThisresultisgivenasahistograminFig.15.TheGaussiandistributionisalsodrawninthesamefigurebyadottedlineasareference.在光伏誤差輻射監(jiān)測能量測量的數(shù)據(jù)與采樣速度耿介黑川東京大學(xué)農(nóng)業(yè)和科技，16年2月24日Nakamachi小金井市，東京184，日本摘要：為了測量太陽輻射和光伏陣列輸出能量，除非數(shù)據(jù)采樣間隔是適當(dāng)選擇，否則測量精度無法保證。從這個角度看，實際輻照已被1-4秒的比較高的速度采樣測量44個月，用數(shù)值積分的每日誤差估計各種步長。用逼近公式的誤差與步長以及統(tǒng)計學(xué)獲得其概率密度函數(shù)覆蓋4。最后，諾模圖是決定一個適當(dāng)?shù)牟蓸娱g隔。總結(jié)的例子表明，偏離誤差超過1的組件可以為每1個月或6個月發(fā)生一次，如果選擇步長為105或65.5s，則總誤差變?yōu)?0.04851，或-0.03361，其中包括根據(jù)筑波測得的數(shù)據(jù)平均誤差分量。版權(quán)1996愛思唯爾網(wǎng)絡(luò)科技有限公司。引言:在太陽能光