資源描述:
《ann、anfis和ar模型在日徑流時(shí)間序列預(yù)測(cè)中的應(yīng)用比較》由會(huì)員上傳分享,免費(fèi)在線閱讀,更多相關(guān)內(nèi)容在教育資源-天天文庫(kù)。
1、ANN、ANFIS和AR模型在日徑流時(shí)間序列預(yù)測(cè)中的應(yīng)用比較 摘要:水文預(yù)測(cè)是水文學(xué)為經(jīng)濟(jì)和社會(huì)服務(wù)的重要方面。其預(yù)報(bào)結(jié)果不僅能為水庫(kù)優(yōu)化調(diào)度提供決策支持,而且對(duì)水電系統(tǒng)的經(jīng)濟(jì)運(yùn)行、航運(yùn)以及防洪等方面具有重大意義。自回歸模型(AR模型)、人工神經(jīng)網(wǎng)絡(luò)(ANN)和自適應(yīng)神經(jīng)模糊推理系統(tǒng)(ANFIS)在日徑流時(shí)間序列中應(yīng)用廣泛。將這三種模型應(yīng)用于桐子林的日徑流時(shí)間序列預(yù)測(cè)中,不僅采用納什系數(shù)(NS系數(shù))、均方根誤差(RMSE)和平均相對(duì)誤差(MARE)為評(píng)價(jià)指標(biāo),對(duì)三種模型的綜合性能進(jìn)行了比較。而且,在對(duì)三種模型預(yù)測(cè)結(jié)果的平均相對(duì)誤差的閾值統(tǒng)計(jì)基礎(chǔ)上,分析了三種模
2、型的預(yù)測(cè)誤差分布。同時(shí),通過研究模型性能指標(biāo)隨預(yù)見期的變化過程評(píng)價(jià)了三種模型不同預(yù)見期下的預(yù)測(cè)能力。結(jié)果表明ANFIS相對(duì)于ANN和AR模型不僅具有更好的模擬能力、泛化能力,而且在相同的預(yù)見期下具有更優(yōu)的模型性能,可以作為日徑流時(shí)間序列預(yù)測(cè)的推薦模型。 關(guān)鍵詞:自回歸模型;人工神經(jīng)網(wǎng)絡(luò);自適應(yīng)神經(jīng)模糊推理系統(tǒng);日徑流時(shí)間序列預(yù)測(cè) 中圖分類號(hào):P338文獻(xiàn)標(biāo)志碼:A文章編號(hào):16721683(2016)06001206 ComparativestudyofANN,ANFISandARmodelfordailyrunofftimeseriespredictio
3、n9 TANQiaofeng1,WANGXu2,WANGHao2,LEIXiaohui2 (1.CollegeofWaterResourceandHydropower,SichuanUniversity,Chengdu610065,China; 2.ChinaInstituteofHydropowerandWaterResourcesResearch,Beijing100038,China) Abstract:Hydrologicalpredictionisanimportantaspectofhydrology′sserviceforeconomica
4、ndsociety.Thepredictionresultnotonlyprovidesdecisionsupportforreservoirgenerationoperation,butalsoisofgreatsignificancetotheeconomicaloperationofhydropowersystems,navigation,floodcontrolandsoon.Theautoregressivemodel(ARmodel),artificialneuralnetwork(ANN)andadaptiveneuralfuzzyinferenc
5、esystem(ANFIS)havebeenwidelyappliedinthedailyrunofftimeseriesprediction.Inthispaper,thesethreemodelswereappliedindailyrunoffpredictionatTongzilinstation.NashSutcliffeefficiencycoefficient(NScoefficient),rootmeansquareerror(RMSE)andmeanabsoluterelativeerror(MARE)wereusedtoevaluatethep
6、erformancesofthreemodels.Thresholdstatisticsindexwasusedtoanalyzepredictionerrordistributionofthreemodels.Atthesametime,thepredictionabilityofthreemodelswasstudiedbygraduallyincreasingthepredictionperiod.TheresultsshowedthatANFIShadnotonlybettersimulationabilityandgeneralizationabili
7、ty,but9alsobettermodelperformanceinthesamepredictionperiodcomparedtoANNandARmodel.Asaresult,ANFIScanbearecommendedpredictionmodelfordailyrunofftimeseries. Keywords:autoregressivemodel;artificialneuralnetwork;adaptiveneuralfuzzyinferencesystem;dailyrunoffprediction 水文預(yù)測(cè)是防汛、抗旱和水資源利用等
8、重大決策的重要依據(jù),歷來