基于BP神经网络草原土壤Cu含量预测  

Prediction of copper contents in the grassland soil based on BPNN

作　　者：史红飞 侯建伟 卢志宏 杨勇 SHI Hongfei;HOU Jianwei;LU Zhihong;YANG Yong(Tongren University, Tongren 554300, Guizhou, China;Inner Mongolia Institute of Grassland Survey and Planning, Hohhot 010051, Inner Mongolia, China)

机构地区：[1]铜仁学院,贵州铜仁554300 [2]内蒙古自治区草原勘察规划院,内蒙古呼和浩特010051

出　　处：《草业科学》2019年第10期2492-2498,共7页Pratacultural Science

基　　金：内蒙古自治区自然科学基金“基于高光谱的草原矿区土壤光谱特征与重金属定量反演研究(2017BS0310)”;内蒙古自治区自然科学基金“典型草原退化植被的光谱特征研究(2015BS0321)”;内蒙古自治区科技创新引导奖励资金“天空地协同草原生态监测及预警体系构建与示范(20170519)”

摘　　要：草原矿区开采破坏了草地生态系统的平衡,Cu元素随粉尘、径流扩散和迁移,并通过食物链富集而影响正常生产活动,因此建立简单又快速的预测模型对草原土壤Cu含量监测具有重要意义。以内蒙古锡林郭勒草原矿区为研究对象,耦合土壤有机质和pH,建立预测Cu含量的BP神经网络模型。结果表明:1)建模数据和测试数据的拟合优度随隐含层层数的增加呈先增加后降低的规律,4层时拟合优度最大;2)隐含层层数为3、4、5和7时,隐含层节点数相近;3)4层隐含层BP神经网络模型具有较高的拟合精度和适当的规模,可以用于草原表层土壤Cu含量的快速预测。Mining has dramatically damaged the ecological balance of grassland, and the elemental copper is diffusing through dust and surface runoff;This has affected normal production and the enrichment of food chains. Therefore, establishing a fast and easy prediction model for monitoring the content of copper in grassland soil is of great significance. This study was conducted in a mining area in Xilingol grassland and aimed to build a BPNN (BP neural network) to predict the copper content by coupling soil organic matter and pH. The results were: 1) the fitting of the training and test data was first increased and then decreased with the increase in the number of hidden layers, and the fitting reached the maximum at layer 4;2) the number of nodes in the hidden layers was similar in layers 3, 4, 5, and 7;3) The BPNN achieved a relatively high fitting and optimal scale when the number of hidden layers was four. Therefore, this method was able to rapidly predict the Cu contents rapidly in the surface soil of grassland.

关 键 词：BP神经网络 多层网络 CU含量 

分 类 号：S812.2[农业科学—草业科学]