基于FOR-GRNN的寒地水稻产量预测  

Prediction of Rice Yield in Cold Region Based on FOR-GRNN

作　　者：林彦宇 衣淑娟[1] 张忠学 王孟雪[3] 聂堂哲 LIN Yan-yu;YI Shu-juan;ZHANG Zhong-xue;WANG Meng-xue;NIE Tang-zhe(College of Engineering,Heilongjiang Bayi Agricultural University,Daqina 163319,China;Key Laboratory of Efficient Use of Agricultural Water Resources,Ministry of Agriculture,Harbin 150030,China;College of Agriculture,Heilongjiang Bayi Agricultural University,Daqina 163319,China)

机构地区：[1]黑龙江八一农垦大学工程学院,黑龙江大庆163319 [2]农业部农业水资源高效利用重点实验室,黑龙江哈尔滨150030 [3]黑龙江八一农垦大学农学院,黑龙江大庆163319

出　　处：《数学的实践与认识》2019年第22期171-178,共8页Mathematics in Practice and Theory

基　　金：黑龙江省博士后资助经费项目;“十三五”国家重点研发计划项目子课题(2016YFC0400108-1);农业部农业水资源高效利用重点实验室开放基金项目(2017004);黑龙江八一农垦大学三横三纵计划(TDJH201803)

摘　　要：针对东北寒地作物水肥资源利用低下问题,以水稻为研究对象,采用氮肥、磷肥、钾肥、水作为模型输入,产量作为模型输出,运用果蝇优化算法对广义神经网络中的平滑因子进行寻优,利用MATLAB2014a神经网络工具箱对水稻不同水肥条件下产量数据进行分析和训练.结果表明,果蝇算法优化回归神经网络模型经过多次迭代寻优比较准确地确定了最佳的平滑因子,在非线性多因素系统建模方面有着很强的优势,其对产量的预测精度达86.5%,高于广义回归神经网络、BP神经网络及灰色神经网络.因此运用果蝇算法优化后的广义神经网络模型可以对水稻产量进行有效预测,为优化水稻水肥配施方案及灌溉制度的制定提供理论依据.This paper aiming at the problem of low utilization of water and fertilizer resources for crops in northeast cold region,rice was taken as study object,nitrogen,phosphate,potash fertilizer and water were used as model inputs,and yield was used as model output,the smoothing factors of generalized neural network were optimized by fruit fly optimization algorithm,the yield data of rice under different water and fertilizer conditions were analyzed and trained by MATLAB2014 a neural network toolbox..The results showed that the optimal regression factors could be accurately determined by multi-iterations and optimization of the fruit fly optimization regression neural network model.It had a strong advantage in the modeling of non-linear multi-factor system,and its prediction accuracy reached 86.5%in rice yield,which was higher than that of generalized regression neural network,BP neural network and grey neural network.Therefore,the generalized neural network model optimized by fly fruit algorithm could effectively predict rice yield,and provide a theoretical basis for optimizing water and fertilizer application schemes and formulating irrigation schemes.

关 键 词：寒地水稻 水肥 果蝇算法 广义神经网络 产量 

分 类 号：S51[农业科学—作物学]