黄河主要产沙区典型支流径流泥沙预测深度学习方法探究
打开文本图片集
关键词：径流泥沙；预测；深度学习；Mann-Kendall检验；无定河；窟野河中图分类号：TP39；P338 文献标志码：A doi：10.3969/j.issn.1000-1379.2026.02.021川用格式：李睿，吴丹，刘启兴.黄河主要产沙区典型支流径流泥沙预测深度学习方法探究[J].人民黄河，2026，48（2）：136-141.
Exploration of Deep Learning Methods for Predicting Runoff and Sediment in Typical Tributaries of the Yellow River's Main Sand Producing Areas
LI Rui¹，WU Dan2，3，4，LIU Qixing2，3，4 （1.Hydrological Bureau，YRCC，Zhengzhou 450004， China;2.Yelow River Institute of Hydraulic Research，YRCC， Zhengzhou ，China；3.Yellow River Laboratory，Zhengzhou ，China； 4.Henan Smart Water Conservancy Engineering Technology Research Center，Zhengzhou 45OoO3，China） Abstract：Teblisetfdiioaldrolgicalodelselothuadgeralidssuptiosfaldolicalp nomena，wichrentlsufefroparaerucaddiltisindacallyacteinguderlsufcehaLe agingdeeplearingtechniquesbasedobigdata，thisstudyconductedanin-depthinvestigationintothetheoryandmethodologofdeep learningmodelfodtpilgst，ves，gqtialnaallea poralatributesofYellowRiverwaterandsedimentdata.SelectingthemainsedimentproductingtrbutarsofteYelowRiver，teKuye RiverandWudingRiverolctedulti-souedataodrologegeationdterfactors，perfoedfatureextractio，dentifedte timeofbruptagsiwaterndsdimentelementsusigtea-Kendalltest，antifddrivingforstouhultipleg earanalysis，andonstructedunofndsdimentitellgentpredictiomodelsfortyicalbasinsinthensdment-producinggiosf theYellwRiverbasedonDeepBelief Network（DBN），Long ShortTerm Memory（LSTM），andExtremeGradientBosting（XGBost） models.Teresultssowthattheaualrunofandsdimentdischargeofthetwowatershedsshoadereasingtrendaftertearuptchange year.Thecontributionvalueofnonrainfalfctors toannualsedimentdischargeisnegative，andteyplayaroleinreducingsedmentThe LSTMmodelandodelcasicallyflttheprocessoffdfuctuatiosdthepfoancofLSTinsiulatingpakvalusis slightlybetertantatofodeleaveagelatiepereeerosofteXGBoostodelontetstdatasesofaijcanStation and Wenjiachuan Station are 44.0% and 13.4% ， respectively.
Key words：runoff and sediment;prediction；deep learning；Mann-Kendall test；Wuding River； Kuye River
0 引言
黄河流域大部分地区属干旱半干旱地区，水环境条件相对复杂，其径流泥沙的形成受气候、下垫面条件、人类活动以及社会经济等多种因素的综合影响[]，径流泥沙预报呈现出非线性、强关联性、复杂性。（剩余8420字）
