基于遗传神经网络的黑龙江浅表地层水分预测  

Prediction of Near-surface Moisture Based on GNN in Heilongjiang Province

作　　者：许秀英[1] 黄操军[1] 杨秋梦[1] 

机构地区：[1]黑龙江八一农垦大学,黑龙江大庆163319

出　　处：《水土保持研究》2013年第1期269-272,共4页Research of Soil and Water Conservation

基　　金：黑龙江教育厅科学技术研究项目(12521382);黑龙江农垦总局攻关项目(HNK10A-07-01-03)

摘　　要：针对BP神经网络预测土壤墒情容易出现较大空间内存在局部极值点的问题,采用GA算法对BP网络进行优化,根据大豆作物在不同生长阶段的根系分布及吸水情况,划分3个不同发育阶段,5个地层深度,建立3种对应的土壤含水量遗传神经网络预测模型,并应用于黑龙江垦区红星农场大豆田间土壤水分预测,分别对3种模型的整体预测误差进行了分析,2009年大豆播种前期及其全生育期土壤体积含水量预测的平均绝对误差为1.83%,能较好地反映大豆田间土壤水分具体情况,为大豆节水灌溉与管理提供可靠的科学依据,该预测方法亦可为寒地大豆或其他农作物田间土壤水分预测提供借鉴。According to BP neural network for prediction of soil moisture in the presence of larger space existing in the local extremum, BP neural network was optimized by using GA algorithm. In accordance with the crop root growth and distribution of water situation, the different growth period of soybean, 5 strata depths, genetic neural network models were built for prediction of soil moisture. They were applied in Heiiongjiang Province reclamation area farm red star soybean field. The results showed that the prediction model of soybean sowing in early 2009 and the whole growth period soil volumetric water content prediction Of the mean absolute error was 1.83%, better indicating soybean field soil moisture in specific cases. GA-BP forecast method provided reliable scientific basis for soybean water-saving irrigation and management, and couldlbe used to provide reference of cold soybean or other crops field soil moisture forecast.

关 键 词：土壤水分 预测 遗传神经网络 

分 类 号：S152.7[农业科学—土壤学]