滴灌土壤湿润体迁移计算的人工神经网络模型  

Quantification of soil wetting volume development under surface drip irrigation using artificial neural networks

作　　者：陈帅[1] 毛晓敏[1] CHEN Shuai;MAO Xiaomin(College of Water Resources&Civil Engineering,China Agricultural University,Beijing 100083,China)

机构地区：[1]中国农业大学水利与土木工程学院

出　　处：《排灌机械工程学报》2020年第2期206-211,共6页Journal of Drainage and Irrigation Machinery Engineering

基　　金：国家高技术研究发展计划资助项目(2016YFC040106-3);国家自然科学基金资助项目(51679234)

摘　　要：基于非饱和土壤水分运动理论和单点源滴灌中土壤水分迁移特征,应用HYDRUS-2D/3D模型对33种土壤质地(分属11类土质类型,美国制土壤质地分类系统)、不同滴灌流量(1,2,3 L/h)下的湿润体运动过程进行了数值模拟,然后根据不同土壤质地和滴灌流量下湿润体动态变化的HYDRUS模拟结果,以滴灌量和土壤饱和导水率与滴灌流量的比值作为输入变量,构建了描述滴灌湿润体在不同土质和滴灌流量下迁移变化的人工神经网络模型.该模型输入变量少、易于操作,且将模型计算结果与实测情况对比表明,计算的入渗过程与实测的入渗过程基本一致,相关系数的平方(R^2)均在0.82以上,因此该模型对不同土质中湿润体运移规律的预测效果较好.Based on the Richard equation of water movement in unsaturated soil and the characteristics of soil water movement with drip irrigation from single point source,the HYDRUS-2D/3D model was used to numerically simulate the soil wetting pattern under different dripper discharges(1,2,3 L/h)in 33 kinds of soil textures(seperately belonging to 11 kinds of soil types in USST).Based on the si-mulated results of the soil wetting volume development under different dripper discharges and soil textures,and using irrigation amount and the ratio of soil saturated water conductivity and dripper discharge as the input variables,the artificial neural network was built for predicting the soil wetting vo-lume development under different soil textures and dripper discharges.The results show that the infiltration processes,which are simulated by this artificial neural network with less input variables and more maneuverable,are consistent with the actual infiltration processes,and the correlation coefficient(R^2)is over 0.82.Therefore,the developed model can achieve good effect in predicting the migration of soil water for different kinds of soil textures compared with the experimental data.

关 键 词：地表滴灌 土质 滴灌流量 湿润体运移 人工神经网络 

分 类 号：S275.6[农业科学—农业水土工程]