基于SO-BP神经网络的温室环境预测模型研究
Research on greenhouse environment prediction model based on SO-BP neural network
张万帆 1任力生 1王芳1
作者信息
- 1. 河北农业大学信息科学与技术学院,河北保定,071001;河北农业大学河北省农业大数据重点实验室,河北保定,071001
- 折叠
摘要
由于温室环境中温湿度的调控过程存在滞后响应特性,预测温室环境变化趋势是构建温室精准控制系统中不可或缺的一部分.针对传统神经网络算法在温室预测方面精度不足等问题,提出一种基于蛇优化算法(snake optimizer,SO)优化BP神经网络的温室环境预测方法.试验结果表明,该方法预测15 min内温度的决定系数R2为0.956 4,比BP模型、HHO-BP模型分别提高14.87%、2.19%,平均绝对误差MAE、平均绝对百分比误差MAPE、均方根误差RMSE值分别为0.481 3、2.237 8、0.672 9;预测15 min内湿度的R2为0.982 1,比BP模型、HHO-BP模型分别提高13.12%、2.37%,预测指标MAE、MAPE、RMSE值分别为1.709 0、2.584 2、2.283 8.该模型的预测结果较理想,可用于温室温湿度预测.
Abstract
Predicting change trends in greenhouse environment is an essential part of the construction of an accurate greenhouse control system,due to the hysteretic response characteristic of temperature and humidity regulation process in greenhouse environment.In response to the issue of insufficient accuracy of traditional neural network algorithms in greenhouse prediction,a method for predicting greenhouse environment based on snake optimizer(SO)optimized BP neural network is proposed.Experimental results show that the R2 value of predicting the temperature within 15 minutes by this method is 0.956 4,which is 14.87%and 2.19%higher than that of the BP model and the HHO-BP model,respectively.The prediction indicators of MAE,MAPE,and RMSE values for temperature are 0.481 3,2.237 8,and 0.672 9,respectively.The R2 value of predicting the humidity within 15 minutes is 0.982 1,which is 13.12%and 2.37%higher than that of the BP model and the HHO-BP model,respectively.The prediction indicators of MAE,MAPE,and RMSE values for humidity are 1.709 0,2.584 2,and 2.283 8,respectively.The prediction results of this model are quite ideal and can be used for greenhouse temperature and humidity prediction,providing technical support for the establishment of an accurate greenhouse control system.
关键词
温室环境/温湿度预测/精准控制系统/蛇优化算法/神经网络Key words
greenhouse environment/temperature and humidity prediction/precise control system/snake optimizer algorithm/BP neural network引用本文复制引用
基金项目
河北省省级科技计划项目(19220119D)
出版年
2024
国家科技期刊平台
NSTL
万方数据