水文时间序列逐步回归随机组合预测模型及其应用  

Application of prediction model for stochastic combination of stepwise regression of hydrologic time series

作　　者：汤成友 郭丽娟 王瑞 

机构地区：[1]长江上游水文水资源勘测局,重庆400014

出　　处：《水利水电技术》2007年第6期1-4,共4页Water Resources and Hydropower Engineering

基　　金：国家重点基础研究发展计划"973"项目(2003CB415205)

摘　　要：由于气候因素和下垫面因素的综合影响,水文时间序列表现出复杂的非线性,包括确定性成分和随机成分,如月径流序列、日均流量序列等。这些预报对象如果不加处理直接用AR(p)建模进行预测误差较大。文中介绍了通过对非平稳序列提取周期项和趋势项后的余差序列建立AR(p)模型进行水文中长期预报的组合预测方法,并以嘉陵江北碚站7月最大洪峰流量序列为例对组合模型进行验证,结果比较满意。Because of the overall influence of meteorology factors and underlaying surface factors, the hydrologic time series are high nonlinear, including the definitiveness and randomicity i.e. the monthly runoff series and daily average flow series etc. If these series are not to be processed and are predicted with AR（p） model directly, the errors will be quite larger. The integrated model established by residual series, formed by taking out the cyclical component and the trend component from the non-steady series, is introduced herein. It is verified with the max-flow series of July from Beibei Station, and the result is satisfactory as well.

关 键 词：逐步回归 周期分析 趋势分析 自回归模型 组合模型 水文时间序列 水文中长期预报 

分 类 号：P33[天文地球—水文科学]