基于主成分分析的土壤凋萎系数BP预测模型  

BP Prediction Model for Soil Wilting Coefficient Based on Principal Component Analysis

作　　者：于浕 樊贵盛[1] YU Jin FAN Gui-sheng(College of Water Resource Science and Engineering, Taiyuan University of Technology, Taiyuan 030024, Chin)

机构地区：[1]太原理工大学水利科学与工程学院,太原030024

出　　处：《节水灌溉》2016年第10期51-54,共4页Water Saving Irrigation

基　　金：国家自然科学基金项目(40671081)

摘　　要：基于黄土高原区农田耕作层土壤凋萎含水率的测试资料,建立了主成分分析与BP神经网络相结合土壤凋萎系数预测模型。通过主成分分析法减少了输入层神经元个数,优化了网络结构,提高了工作效率。预测值和实测值的相对误差平均值控制在5%以内,在可接受的范围,表明利用土壤基本理化参数预报农田耕作土壤的凋萎含水率是可行的。研究结果在提高传统神经网络的预测精度和收敛速度的同时,可为黄土高原区耕作农田作物用水管理以及促进土壤生产潜力的发挥提供强有力的理论支撑。Based on the test data of farming soil wilting percentage of farmland in the Loess Plateau region, a prediction model for soil wilting coefficient is established based on the combination of principal component analysis and BP neural network. Through principal component analysis, the number of neurons in the input layer is reduced, the network structure is optimized and the work efficiency is enhanced. The mean relative error between the predicted value and the measured value is within 5 % and in an acceptable range, which indicates that using the soil basic physicochemieal parameters to forecast wilting percentage of farming soil is feasible. The study results not only improve the forecasting accuracy and convergence speed of traditional neural networks, but also provide a strong theoretical support for the crop water management and promoting soil production potential in the Loess Plateau.

关 键 词：主成分分析 凋萎系数 BP模型 土壤理化参数 误差分析 

分 类 号：S152[农业科学—土壤学] TV93[农业科学—农业基础科学]