澜沧江流域中长期径流预报方法研究  

Study of medium and long term runoff forecasting method for Lancang River Basin

作　　者：赵鹏雁 张利平[1] 王旭 胡振奎 吕双江 倪旺丹 ZHAO Pengyan;ZHANG Liping;WANG Xu;HU Zhenkui;LU Shuangjiang;NI Wangdan(State Key Laboratory of Water Resources and Hydropower Engineering Science,Wuhan University,Wuhan 430072,China;Centralized Control Center,Huaneng Lancang River Hydropower Co.,Ltd.,Kunming 650214,China)

机构地区：[1]武汉大学水资源与水电工程科学国家重点实验室,湖北武汉430072 [2]华能澜沧江水电股份有限公司集控中心,云南昆明650214

出　　处：《武汉大学学报（工学版）》2018年第7期565-569,595,共6页Engineering Journal of Wuhan University

基　　金：国家十二五科技支撑计划项目(编号:2013BAB06B04)

摘　　要：以澜沧江流域小湾水库为研究对象,分别采用基于时间序列的最近邻抽样回归模型、小波分析方法、混沌理论模型、周期均值叠加法、均生函数模型、前后期径流量相关法和基于水文气象因子的逐步多元回归模型、神经网络模型、支持向量机模型等10种方法对小湾水库逐月入库径流进行预报,分析比较了这10种模型的预报精度.从整体平均上来看,小波分析模型的预报效果最好,同时,各模型枯水期的预报精度比汛期的预报精度要高.结合各个模型的特点及预报精度,可以采用不同的预报方案对各月进行预报,从而为澜沧江流域小湾水库制定未来中长期调度计划提供技术支持.Taking Xiaowan Reservoir in Lancang River Basin for an example;and by using nearest neighbor bootstrapping regressive model（NNBR）;wavelet analysis model（WA）;chaos theory（CT）;periodic mean superposition method（PMSM）;mean generation function model（MGF）;the earlier and later runoff correlation method（RCM）;stepwise multiple regression model（SMR）;artificial neural network model（ANN）and support vector machine（SVM）model;etc.;their monthly runoffs are forecasted.The forecast accuracy of the ten models is analyzed to compare the disadvantage and advantage of three methods from different aspects.The results show wavelet analysis method has best effects on average.Meanwhile;the predication accuracy in the dry season looks much higher than in flood season.As the result;different forecast methods may be used at two different stages.Thus;the study results provide technical support for making the future medium and long-term power generation plan of Xiaowan Reservoir.

关 键 词：中长期径流预报 最近邻抽样回归 小波分析 混沌理论 小湾水库 

分 类 号：TV124[水利工程—水文学及水资源]