基于改进的BP神经网络对蓄水坑灌冬季果园土壤温度预测  

Prediction of Winter Soil Temperature of Apple Orchard Under Water Storage Pit Irrigation Based on Improved BP Neural Networks

作　　者：贺琦琦 郭向红[1] 雷涛[1] 王晓磊[1] 孙西欢[1,2] 马娟娟[1] 张少文[1] 刘艳武 HE Qi qi;GUO Xiang hong;LEI Tao;WANG Xiao lei;SUN Xi huan;MA Juan juan;ZHANG Shao wen;LIU Yan wu(School of Hydraulic Science and Engineering, Taiyuan University of Technology, Taiyuan 030024, China;Jinzhong University, Jinzhong 030600, Shanxi Province, China)

机构地区：[1]太原理工大学水利科学与工程学院,太原030024 [2]晋中学院,山西晋中030600

出　　处：《节水灌溉》2019年第7期16-20,共5页Water Saving Irrigation

基　　金：国家自然科学基金项目(51109154,51579168);山西省自然科学基金资助项目(201601D011053)

摘　　要：为准确预测冬季果园土壤温度,建立了蓄水坑灌条件下BP神经网络土壤温度预测模型(BP-WSPI-T)、遗传算法优化的BP神经网络土壤温度预测模型(GA-WSPI-T)和增量逆传播学习算法优化的BP神经网络土壤温度预测模型(IBP-WSPI-T),采用坑内平均气温、地表温度、沿相邻两蓄水坑中心连线距坑壁的距离和距坑壁5cm处分层土壤最低温度为模型输入,对距坑壁15、25和35cm处分层土壤最低温度进行预测,并通过与田间实测数据的统计学分析来判定预测效果。结果表明:BP-WSPI-T、GA-WSPI-T和IBP-WSPI-T模型的平均相对误差分别为8.19%、4.41%和7.57%,GA-WSPI-T模型的预测效果最好,较BP神经网络预测精度得到了很大的提高,建议采用GA-WSPI-T模型对蓄水坑灌冬季果园土壤温度进行预测。In order to accurately predict the winter soil temperature of apple orchard, this paper established a BP neural network soil temperature prediction model (BP-WSPI-T), a BP neural network soil temperature prediction model optimized by genetic algorithm (GA-WSPI-T) and a BP neural network soil temperature prediction model optimized by incremental back propagation algorithm (IBP-WSPI-T) under the condition of water storage pit irrigation. The average temperature in the pit, the surface temperature, the distance from the center of the adjacent two water storage pits to the pit wall, and the lowest temperatures of the layered soil 5 cm away from the pit wall were taken as input. The models were used to predict the lowest temperatures of layered soil at 15, 25 and 35 cm from the pit wall, and the prediction effects were determined through the statistical analysis with the field measured data. The results show that the mean absolute percentage errors of BP-WSPI-T model, GA-WSPI-T model and IBP-WSPI-T model are 8.19%, 4.41% and 7.57%, respectively;the GA-WSPI-T model has the best prediction effect compared with BP-WSPI-T, and the prediction accuracy has been greatly improved. Thus, GA-WSPI-T model is suggested to predict the winter soil temperature of apple orchard under water storage pit irrigation.

关 键 词：土壤温度 增量逆传播 遗传算法 BP神经网络 蓄水坑灌 

分 类 号：S275.9[农业科学—农业水土工程]