基于主成分BP人工神经网络的参考作物腾发量预测  

Reference Evapotranspiration Prediction Based on the PCA-BP Artificial Neural Networks

作　　者：欧建锋[1] 程吉林[1] 

机构地区：[1]扬州大学水利科学与工程学院

出　　处：《灌溉排水学报》2008年第2期55-58,共4页Journal of Irrigation and Drainage

摘　　要：影响参考作物腾发量的气象因素众多,且相关程度较高。基于主成分分析原理,将影响ET0的7个主要气象因素以及旬序数进行特征提取,形成3个综合影响因子,既可保证气象信息的完整性,又可避免气象信息的交叉重叠。以江苏省无锡市某区作物腾发量预测为例,经主成分分析并简化的参考作物腾发量BP神经网络模型具有结构简单、收敛快、精度高的特点,可用于ET0预测。Reference evapotranspiration is affected by many high correlated meteorologic factors. Based on principal component analysis（PCA）, the ordinal of ten days and seven main meteorologic factors related to the ETo are synthesized into three synthetic factors. Therefore, not only the completeness of the meteorologic information can be achieved, but the overlapping information can be avoided. Exemplified with a case study in Wuxi, Jiangsu province, the reference evapotranspiration BP networks simplified with PCA have such advantages as more simplified structure, faster convergence and higher precision. Thus, the PCA-BP artificial neural networks can be applied in ETo prediction.

关 键 词：参考作物腾发量 主成分分析 BP神经网络 预测 

分 类 号：S311[农业科学—作物栽培与耕作技术]