基于小波变换的BP神经网络参考作物腾发量预测模型  

Reference crop evapotranspiration forecasting model for BP neural networks based on wavelet transform

作　　者：刘丙军[1] 邵东国[1] 沈新平[1] 

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

出　　处：《武汉大学学报（工学版）》2007年第1期69-73,共5页Engineering Journal of Wuhan University

基　　金：国家重点"973"资助项目(编号:2003CB415206)

摘　　要：受气温、日照、风速、水汽压等因子随机变化的影响,参考作物腾发量时序过程具有非线性、多时间尺度变化等特性.为研究参考作物腾发量在时间尺度上的分布规律,提出了一种基于小波变换与人工神经网络相结合的参考作物腾发量预测模型.该模型吸取了小波分析的多分辨分析功能和人工神经网络的非线性逼近能力,具有较高的预测精度.以韶山灌区参考作物腾发量时间序列为样本,论述了上述模型的优越性.Reference crop evapotranspiration（RCE） characterized by its nonlinearity and multi-time scale feature, may vary with the change of time under the influence of stochastic variation of meteorological factors such as temperature, sunlight, wind speed, vapor pressure and so on. In order to reveal the evo- lutionary law of RCE time series process, a forecasting model with the wavelet transform and the BP neural network combined together is established. This model has super advantage with its absorbing some merits of wavelet transform and artificial neural network. The prediction of Shaoshan Irrigation District series is researched; the results show that the model is satisfactory.

关 键 词：参考作物腾发量 小波变换 人工神经网络 

分 类 号：TV211.1[水利工程—水文学及水资源]