基于神经网络的土壤重金属预测模型研究  

Study on Prediction Model of Soil Heavy Metal Based on Artificial Neural Network

作　　者：秦夕淳 袁琦[1] 周宝宣 QIN Xi-chun YuAN Qi ZHOU Bao-xuan(College of Mechanical and Electrical Engineering, Hainan University, Haikou Hainan 57022)

机构地区：[1]海南大学机电工程学院,海南海口570228

出　　处：《现代农业科技》2016年第19期178-180,共3页Modern Agricultural Science and Technology

基　　金：海南省自然科学基金项目(614223)

摘　　要：土壤中重金属含量变化具有非线性、大延时等特点,很难用传统方法建立土壤重金属预测的精确模型。BP神经网络具有良好的非线性函数逼近能力,非常适合处理土壤预测等复杂问题。利用神经网络模型,通过自适应的动态学习方法和模型优化,采用MATLAB神经网络工具箱建立了土壤重金属预测模型。在预测模型中输入测试样本,将预测结果与实测值进行比较,平均相对误差小于1%。结果表明,所构建的基于BP神经网络的土壤重金属预测模型具有良好的精确性和准确性,能有效预测土壤中重金属的状况。Changes of heavymetal content in the soil have the c^haracter^istics of nonlinear ,-large time delay, and it is difficult to set up the precise model of soil heavy metal in a traditional way. Artificial neural network has the advantage of approximating the nonlinear function, it is an ideal method for dealing with the complex problems such as prediction of soil heavy metal. A prediction model of soil heavy metal based on artificial neural network was established by applying a dynamic self-adaptive learning method and model optimization.This model was completed by programming with neural network toolbox of MATLAB.The results showed that the average relative error was less than 1% between the value of prediction and the measured value when the trained network was applied in prediction.Prediction model of soil heavy metal which constructed based on artificial neural network has good precision and accuracy, and it can effectively predict the status of heavy metals in soil.

关 键 词：土壤重金属 神经网络 预测模型 MATLAB工具箱 

分 类 号：X53[环境科学与工程—环境工程]