基于小波变换的支持向量机水文过程预测  

Prediction of hydrological series based on wavelet transform and support vector machine

作　　者：王红瑞[1] 刘晓红[2] 唐奇[1] 何麟[3] 

机构地区：[1]北京师范大学水科学研究院,水沙科学教育部重点实验室,北京100875 [2]河北理工大学轻工学院,唐山063000 [3]北京师范大学数学科学学院,北京100875

出　　处：《清华大学学报（自然科学版）》2010年第9期1378-1382,共5页Journal of Tsinghua University(Science and Technology)

基　　金：“十一五”国家科技支撑计划项目(2006BAD20A06)

摘　　要：针对支持向量机在水文过程应用分析中存在的问题,该文将小波变换和支持向量机相结合建立水文时序趋势分析模型。首先对水文序列通过小波变换进行预处理,把处理后序列分解成不同时间尺度下的子序列,然后用支持向量机对各子序列分别进行模拟和预测,将这些支持向量机的预测结果通过小波逆变换重构水文时间序列,建立基于小波变换的支持向量机水文过程趋势分析模型,以三门峡水文站天然月径流时序为例进行应用验证。研究结果表明:与传统的支持向量机、神经网络等预测模型相比,本文模型在预测精度和时间长度上均优于前二者。Support vector machines can not accurately predict time series related to hydrological processes due to the different time series in the data. This study combines wavelet transforms with a support vector machine to predict hydrological time series. The wavelet transform is used to decompose the hydrological time series into subseries with different time scales. Thus, the support vector machine （SVM） is used to simulate and predict the future behavior of each subseries. The results from the SVM are then reconstructed using the inverse wavelet transform. The wavelet SVM model was used to analyze monthly natural runoff rates at the Sanmenxia hydrological station. The prediction accuracy and the time variations of the model are both excellent, better than the SVM model or an artificial neural network model. This wavelet-SVM model is shown to be universally useful in a range of applications.

关 键 词：水文过程 支持向量机 小波变换 回归 径流 三门峡水文站 

分 类 号：TV124[水利工程—水文学及水资源] P338[天文地球—水文科学]