基于主成分分析单一参数入渗模型的BP神经网络模型

BP Artificial Neural Network Model of One-parameter Infiltration Model based on Principal Components Analysis

利用标定理论建立较大尺度上的单一参数入渗模型,在此基础上基于主成分分析建立其BP神经网络模型。结果表明:利用主成分分析可将研究区域土壤容重、有机质含量、砂粒含量、粗粉粒含量和粘粒含量综合成3个主成分;基于主成分分析建立的BP神经网络模型预测的标定因子的RMSE为0.4186,除偏大或偏小的标定因子,利用预测的标定因子预测的累积入渗量与实测值比较接近,可利用所建模型对较大尺度上的单一参数入渗模型进行预测。

The paper constructed one-parameter infiltration model with scaling theory at the large scale and established its BP artificial neural network model based on principal components analysis.The results showed that bulk density,organic matter content,sand content,silt content and clay content could be converted into three principal components;RMSE of scaled factors that were forecasted with established model was 0.4186,except too large or too small scaled factors,cumulative infiltration forecasted with scaled factors that were forecasted with above established model was close to its measured value,which indicated that established model could be used to forecast one-parameter infiltration model parameter at the large scale.