摘要
基于BP神经网络建立云南省粮食产量预测模型,分析有关文献,最终选择农业机械总动力、有效灌溉面积、农用化肥施用折纯量、农村用电量、农药使用量、粮食作物播种面积、农用柴油使用量和受灾面积等8个指标作为输入变量,粮食产量为输出变量。首先以云南省1993—2016年的粮食产量及8个粮食产量影响因素等数据,搭建BP神经网络预测模型,预测2017年、2018年和2019年的粮食产量。试验结果表明,基于BP神经网络预测模型在训练阶段,相对误差绝对值基本小于1%;在验证阶段,预测2017年、2018年和2019年的相对误差分别为1.84%、3.25%和2.86%,误差率均控制在5%以为,说明该模型具有很好的预测效果,能够有效地对粮食产量进行预测,并为粮食产量的预测提供了一种新的方法。
The prediction model of grain yield in Yunnan Province was established based on BP neural network. After studying and analyzing the relevant literature, 8 indexes such as total power of agricultural machinery, effective irrigation area, net amount of agricultural chemical fertilizer application, rural power consumption, pesticide use, grain crop sowing area, agricultural diesel use and disaster area are finally selected as input variables, grain yield is the output variable. Firstly, based on the data of grain output and 8influencing factors of grain output in Yunnan Province from 1993 to 2016, a BP neural network prediction model is established to predict the grain output in 2017, 2018 and 2019. The test results show that the absolute value of the relative error of the prediction model based on BP neural network is basically less than 1% in the training stage. In the verification stage, the relative errors in 2017, 2018 and 2019are 1.84%, 3.25% and 2.86% respectively, and the error rates are controlled at 5%, which shows that the model has good prediction effect and can effectively predict grain output. It also provides a new method for the prediction of grain yield.
作者
路思恒
尹红
Lu Siheng;Yin Hong(School of Mechanical and Electrical Engineering,Kunming University of Technology,Kunming 650500,Yunnan,China)
出处
《农业装备与车辆工程》
2023年第1期39-43,共5页
Agricultural Equipment & Vehicle Engineering
基金
云南智能化自动化产业发展研究(云府发研(2017)32号—YNDR2017G1C06)。
关键词
BP神经网络
粮食产量
归一化
梯度下降法
预测模型
BP neural network
grain yield
normalization
gradient descent method
prediction model
