基于Elman神經(jīng)網(wǎng)絡(luò)的中期電力負(fù)荷預(yù)測(cè)模型研究.pdf

1、蘭州理工大學(xué)碩士學(xué)位論文基于Elman神經(jīng)網(wǎng)絡(luò)的中期電力負(fù)荷預(yù)測(cè)模型研究姓名：任麗娜申請(qǐng)學(xué)位級(jí)別：碩士專業(yè)：機(jī)械制造及其自動(dòng)化指導(dǎo)教師：芮執(zhí)元20070601AbstractLoadforecastingbelongstostratagemforecastingwhichisanimportantresearchcontentofpowersystemplanningandrunningandapremiseforreliablesup

2、plyingandeconomicrunningWiththedevelopmentofthestatepowersystem，theelectricnetworkmanagementmodernizesdaybydayandloadforecastingisoneofthemostimportanttasksofthemodempowersystemoperationreseal“cb_，whicharousesincreasingm

3、oreandmoreinterestsfromresearchersandisamajorfoundationfortheresearchofpowersystemplanningandpowersystemeconomicoperationandautomaticdispatchThereforefindingallappropriateloadforecastingmethodtoimprovetheaccuracyofprecis

4、ionhasimportantapplicationvalueThecoreproblemofloadforecastingistechnicalmattersorthemathematicalmodelConventionalmodelisdescribedbyclearmathematicrepresentation，whichhastheadvantagesoflittlecalculationandhighspeedbutwhi

5、chhaslotsoflimitationssuchasnonselflearningandnonselfadaptingatthesametimeEspeciallywiththedevelopingofeconomicofOUrcountrythestructureofpowersystemismoreandmorecomplicated，thenthefeaturesofnonlineartimevaryinganduncerta

6、intyofloadaremorevisible，SOasuitablemathematicalmodelwhichcanclearlyexpresstherelationshipbetweenloadandthevariableofaffectingloadisdifficulttobuiltButnon—mathematicalmodelbasedonneuralnetworkprovideanewwayforsolvingthep

7、roblemofmathematicalmodelInthispaper，mediumtermloadforecastingbasedonElmanneuralnetworkispresented，andamodifiedBPalgorithmoraback—propagationalgorithmwithadaptivelearningspeedandmomentumgradientfallingisusedThecontributi

8、onofthismodelistoofferanonlinearanddynamicbehaviorofelectricload，whichalsocanovercometheproblemsofslowrateofconvergenceandlocalminimumofstandardBPalgorithmHistoricalloadofdataGansu班dwhichisusedtoforecasteachmonthofnextye

9、arareinvestigatedindetailstodemonstratetheavailabilityandextensibilityofthisapproachThealgorithmbasedonmathematicalstatisticsandthreepointflatprincipleispresentedtoidentifyandcorrectanomalousdatawhichrealizesaccurateallo

10、cationandcorrectionofanomalousdataandprovidesapreparationforaccurateforecastingAtlast，anapplicationisgiventoprovethatthismethodhastheadvantageofsimplenesspracticalitysmallworkloadandlittlemallmadeinterferenceandshowsthat

11、themethodiseffectiveandpracticalThisthesisresearchesElmanneuralnetworkforecastingmodelwiththeabilityofadaptingtime—varyingincludingonehiddenlayerElmanneuralnetworkmodel，twohiddenlayerElmanneuralnetworkmodelandthreehidden