小波分析在径流时间序列预测的应用  

On Wavelet Analysis in Runoff Time Series Prediction

作　　者：丁红[1,2] 武招云 龚若愚 廖文凯 

机构地区：[1]武汉理工大学信息工程学院,湖北武汉430070 [2]柳州师范高等专科学校物理与信息科学系,广西柳州545005 [3]柳州市水文水资源局,广西柳州545001

出　　处：《柳州师专学报》2012年第3期122-128,共7页Journal of Liuzhou Teachers College

基　　金：国家自然科学基金项目"混沌分形理论研究及其在华南降水系统中的应用"(11161029);广西自然科学基金北部湾重大专项"北部湾台风;暴雨灾害的流形学习数据挖掘和智能计算客观集合预报建模理论方法研究与应用"(2011GXNSFE018006);柳州师范高等专科学校科研基金项目"基于流形学习降维方法研究"(LSZ2011B007)

摘　　要：利用dmey小波变换、db4小波变换将柳江年径流总量及年最大径流量时间序列分别分解为不同尺度的频率信号,再使用传统时间序列分析方法对径流分解的频率信号分别设置参数并进行预测,最后利用小波逆变换对预测结果进行重构,以此建立年径流总量、年最大径流量预测模型.与只采用传统的时间序列预测模型对比,联合了小波变换的方法均获得更精确的预报结果.计算结果表明将小波变换运用于径流时间序列的预测能有效提高径流变化趋势预测的准确率,预报的稳定性以及精确度均能明显改善.Using dmey wavelet transformation and db4 wavelet transformation, two time series of the annual total runoff and the annual maximum runoff of Liujiang River are decomposed into several parts with different scales of frequency respectively. Then the future values of these sub - series are forecasted by using traditional forecasting models, in which the process, the parameters are set separately according to the features of each subpart of the runoff time series to obtain better predicting results. In the final stage, through the inverse dmey wave- let transformation and the inverse db4 wavelet transformation, using the predicted results of the subparts, future behavior of the runoff series is reconstructed. The calculation results show that the presented methods are superior to the traditional models in the accuracy of runoff change trend forecast, stability and the prediction precision of the models to use wavelet transform in runoff time series prediction.

关 键 词：小波变换 时间序列 尺度分析 径流预测 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]