基于WOA-RBF的温室温度预测研究  

Research on greenhouse temperature prediction based on WOA-RBF

作　　者：李其操 董自健 Li Qicao;Dong Zijian(School of Electronic Engineering,Jiangsu Ocean University,Lianyungang 222005,China)

机构地区：[1]江苏海洋大学电子工程学院,连云港222005

出　　处：《现代计算机》2023年第11期43-46,51,共5页Modern Computer

摘　　要：温度是温室内作物生长的关键环境因素,准确预测出温室内温度的变化对温室管理有着重要意义。为了提高预测精度,提出了基于鲸鱼优化算法(WOA)优化RBF神经网络的温室温度预测模型,通过鲸鱼优化算法对RBF神经网络的输出层的神经元线性权重、隐藏层的神经元中心点和隐藏层的高斯核宽度进行寻优,得到最优参数,最终得到优化后的输出结果。实验证明,与未优化的RBF神经网络预测模型相比,WOA-RBF神经网络温室温度预测模型得到的结果误差更小,平均绝对误差(MAE),均方误差(MSE)和平均绝对百分比误差(MAPE)分别为0.478、0.346和0.588,与未优化的RBF神经网络预测模型相比分别提升了9.1%、32.7%和18.0%。Temperature is the key environmental factor for crop growth in greenhouse,and it is of great significance to accurately predict the temperature change in greenhouse for greenhouse management.In order to improve the prediction accuracy,a greenhouse temperature prediction model based on the whale optimization algorithm(WOA)optimized RBF neural network is proposed.The whale optimization algorithm is used to optimize the neural linear weight of the RBF neural network output layer,the neural center of the hidden layer and the Gaussian kernel width of the hidden layer to obtain the optimal parameters,and finally obtain the optimized output results.The experiment shows that compared with the unoptimized RBF neural network prediction model,the result error of the WOA-RBF neural network greenhouse temperature prediction model is smaller.The mean absolute error(MAE),mean square error(MSE)and mean absolute percentage error(MAPE)are 0.478,0.346 and 0.588,respectively,which are 9.1%,32.7%and 18.0%higher than the unoptimized RBF neural network prediction model.

关 键 词：鲸鱼优化算法 RBF神经网络 温室温度 预测模型 

分 类 号：S625.51[农业科学—园艺学] TP183[自动化与计算机技术—控制理论与控制工程]