基于小波分析的最近邻抽样回归预报模型研究  

The nearest neighbor bootstrap regressive model for forecast based on wavelet analysis

作　　者：汤成友 刘晔[2] 

机构地区：[1]长江水利委员会水文局长江上游水文水资源勘测局,重庆400014 [2]重庆交通大学,重庆400074

出　　处：《人民长江》2011年第6期53-56,共4页Yangtze River

摘　　要：为了提高水文中长期预报的精度,以屏山站1951～2005年年径流序列为例,应用小波分析技术对信号进行多尺度分析,将水文时间序列分解成若干个高频序列和低频序列,再将高频序列和低频序列分别应用最近邻抽样回归建立模型,然后应用小波重构技术将各序列进行重构,从而实现对原始序列的预测。结果表明:应用小波技术建立的组合模型,其精度明显高于单一的最近邻抽样回归模型,可以应用于生产实践中。In order to improve mid-long term hydrological forecast accuracy,taking annual runoff series from 1951 to 2005 of Pinshan hydrological station as a research subject,wavelet technology is applied in the analysis of multi-scale information.The hydrological data series are decomposed into high frequency data series and low frequency data series and the nearest neighbor bootstrap is used to establish regressive model for the 2 decomposed series,and then the series are reconstructed by wavelet technology.Therefore,the forecast for original data is realized.The analysis results show that the forecast accuracy by the regressive model integrated wavelet and the nearest neighbor bootstrap technology is higher than that by the nearest neighbor bootstrap technology alone.

关 键 词：水文中长期预报 小波分析 最近邻回归 组合模型 

分 类 号：P338[天文地球—水文科学]