BP神经网络在小麦赤霉病气象预测中的应用  

Application Research on Wheat Scab Forecasting Based on BP Neural Network

作　　者：刘志红[1] 张雷[2] 燕亚菲[2] 周洁[2] 张正健[1] 张涵斌[2] 

机构地区：[1]成都信息工程学院资源环境学院,四川成都610225 [2]成都信息工程学院大气科学学院,四川成都610225

出　　处：《云南农业大学学报（自然科学版）》2010年第5期680-685,共6页Journal of Yunnan Agricultural University:Natural Science

基　　金：四川省教育厅重点项目(08ZA099)

摘　　要：应用BP神经网络的方法建立四川资中小麦赤霉病发病的预报模型,为预防小麦赤霉病发病提供科学依据。根据四川资中小麦赤霉病发生发展的气象生理指标及历年该病发生的统计资料,借助人工神经网络强大的函数映射能力,采用Fletcher-Reeves算法的变梯度反向传播算法——Traincgf,建立了小麦赤霉病发病的气象预报模型。该模型不需要事先确定数学模型,拟合与预测的平均绝对偏差分别为0.01,0.05,优于多元线性回归模型的0.17,0.29。BP神经网络预报模型的拟合精度和预报精度都较高,优于多元线性回归模型,能很好地实现预期效果,对小麦赤霉病发病的预测预防工作具有一定的指导意义。Zizhong wheat scab forecasting model was established using the method of BP neural network.The model＇s main purpose was to provide scientific basis for preventing wheat scab from happening.According to the occurrence and development of wheat scab＇s physiological indicators and the weather factors that occurred over the years in Sichuan Province,as well as the strong ability of the artificial neural network to function mapping.The wheat scab weather forecasting model was set up by using the Traincgf algorithm,the alternating gradient back-propagation algorithm of Fletcher-Reeves algorithm.The model had overcome the shortcoming of the multiple linear regression model which needed to know the pre-determined gray theory.The average absolute deviations of the model＇s fitting and prediction were 0.01 and 0.05,respectively,which were superior to those of the multiple linear regression model（0.17 and 0.29,respectively）.The forecasting model built by using the BP neural network will be useful in achieving positive results in forecasting.

关 键 词：BP神经网络 小麦赤霉病 预测模型 

分 类 号：S165.28[农业科学—农业气象学]