摘要
为提高温室温度模型预测的准确率,提出一种基于PSO-RBF神经网络的温室温度预测模型。针对梯度下降法收敛速度慢的问题,利用PSO算法优化RBF神经网络参数;为验证该模型有效性,以农场实测数据建立样本,对温室温度进行预测,验证了其比梯度下降法优化的RBF神经网络模型具有更好的预测效果;为给温室内调控设备的提前控制提供依据,根据1月-5月温度数据,利用时间序列法预测相关温室参数,作为该模型输入,利用其预测6月份温度,预测结果表明该月温度呈上升趋势。
To improve the prediction precision of greenhouse temperature module,a PSO-RBF neural network based greenhouse temperature prediction module was proposed.To accelerate the speed of gradient descent,RBF neural network parameters were optimized using PSO algorithm.To verify the effectiveness of this module,it was used to forecast greenhouse temperature by taking the real measured data of the farm as sample.The assumption that using this module generates better forecast result than using gradient descent optimized RBF neural network module was proved.To provide prerequisite data for greenhouse control system,related parameters,which were forecasted by time series forecasting using temperature data from January to May,were used as input of this module for predicting the temperature in June.The results indicate that the temperature is rising continuously in June.
出处
《计算机工程与设计》
北大核心
2017年第3期744-748,共5页
Computer Engineering and Design
关键词
RBF神经网络
PSO算法
温室温度
模型
预测
RBF neural network
PSO algorithm
greenhouse temperature
model
prediction
