資源描述:
《新豐江流域徑流量的分析及預(yù)測(cè)》由會(huì)員上傳分享,免費(fèi)在線閱讀,更多相關(guān)內(nèi)容在學(xué)術(shù)論文-天天文庫(kù)。
1、中山大學(xué)碩士學(xué)位論文新豐江流域徑流量的分析及預(yù)測(cè)姓名:申云申請(qǐng)學(xué)位級(jí)別:碩士專業(yè):水文學(xué)及水資源指導(dǎo)教師:陳俊合20070606中山大學(xué)碩士學(xué)位論文新豐江流域徑流量的分析及預(yù)測(cè)專業(yè):水文學(xué)及水資源碩士生:申云指導(dǎo)教師:陳俊合教授摘要中長(zhǎng)期水文預(yù)報(bào)在抗洪救災(zāi)、水利工程的興建、管理、水量的合理調(diào)配等的重要依據(jù),本文通過分析中長(zhǎng)期水文預(yù)報(bào)的發(fā)展歷程,說明中長(zhǎng)期水文預(yù)報(bào)已經(jīng)取得的成就,以及普遍還存在的一些問題:預(yù)報(bào)的精度偏低,結(jié)果可靠性不高,同時(shí)指出這也是今后研究中亟待得到解決的問題。本文利用方差分析周期外推方法,提出了改進(jìn)的正規(guī)化周期回歸模型。文中首先
2、對(duì)模型的基本結(jié)構(gòu)原理迸行了詳細(xì)的介紹,在此基礎(chǔ)上將模型應(yīng)用到新豐江流域的月徑流序列,進(jìn)行模擬和預(yù)測(cè),計(jì)算結(jié)果為:新豐江流域月徑流序列年內(nèi)分布呈單峰型,序列存在較為明顯的長(zhǎng)期趨勢(shì)項(xiàng)k-一1599641.375—0.02606t一0.28211t24-0.1688t3,經(jīng)過計(jì)算篩選,共找到3個(gè)穩(wěn)定的周期:長(zhǎng)度分別位:151個(gè)月、129個(gè)月、87個(gè)月.長(zhǎng)期趨勢(shì)項(xiàng)和周期項(xiàng)擬合后基本上能夠反映出實(shí)測(cè)徑流序列的特征和變化,但預(yù)報(bào)的精度還不夠高,為此,本文對(duì)周期回歸模型分離出來的隨機(jī)項(xiàng)時(shí)間序列進(jìn)行BP神經(jīng)網(wǎng)絡(luò)的分析、計(jì)算預(yù)報(bào),并且把預(yù)報(bào)的結(jié)果擬合到周期回歸模型
3、的預(yù)報(bào)結(jié)果中去,從而達(dá)到了從整體上提高精度的目的。關(guān)鍵詞:中長(zhǎng)期水文預(yù)報(bào)周期回歸神經(jīng)網(wǎng)絡(luò)徑流新豐江流域中山大學(xué)碩士學(xué)位論文AnaIysiSandForecastoftheRunoffinXinfengjiangBasigMajor:hydrologyandwaterresourcesName:ShenYunSupervisor=ProfessorChenJun-heABSTRACTThemidandlong-termhydrologicforecastistheimportantbasistothefloodmitigation,theconst
4、ruct、managementofhydraulicengineering,thereasonablyuseofwaterandsoon.Throughtheanalysisonthedevelopmentcourseofthemidandlong-termhydrologicforecast,thearticleshowstheachievementwehavegotten,andtheproblemswhichexistuniversally·——--··-thepredictionprecisionislowrelatively,there
5、sultisnotsoreliable.a(chǎn)ndtheseproblemsareexpectedtobesettledinthefuture.usingthevarianceandperiodanalysismethod,thearticleputsforwardtheimprovednormalizedperiodregressionmodelThefirstofall,theauthorintroducesthestructureandtheprincipleofthemodelindetaiLOnthebasisofit,toaptlythe
6、modeltothemonthlyrunofftimeseriesofthexinfengjiangbasin,theoutcomeshows:Thedistributionofthemeanrunoffisthesingle-peaktypeduringtheyear,Thereisthegradualrisingtrendinthetimeseries,andtheformulaofthetrenditemis:K-一1599641·375—0·02606t一0·28211t2+0·1688t3,thmugh鋤lysis,find3stabl
7、eperiods=151months、129months、87months.Thesumofthelong-termtrenditemandtheperiodsitembasicallycanreflectthecharacteristicandthechangeoftheactualrunoffsequence,bettheprecisionisⅡ中山大學(xué)碩士學(xué)位論文insufficientlyhigh,forthis,usingBPNeuralNetworks,analyses,computes,forecaststhestochastict
8、imeserieswhichisseparatedfromthePeriodicRegressionmodel,andfitsthefo