基于BP神经网络和小波神经网络的太阳辐射强度预测

Prediction of Solar Radiation Intensity based on BP and Wavelet Neural Network

针对太阳辐射的波动对太阳能相关系统的不利影响,提出了一种基于BP神经网络和小波神经网络的混合预测模型,对太阳逐时辐射强度进行预测。首先,利用相关性分析确定对太阳辐射强度影响较大的气象因素;然后,分别对BP和小波神经网络进行优化,确定两种神经网络的最优结构;再次,利用小波分解对太阳逐时辐射强度进行小波三层分解,对分解后的分量分别用优化后的神经网络进行预测;最后,将各神经网络输出结果叠加,得到太阳辐射强度的逐时预测值。通过MATLAB软件进行仿真,并与BP神经网络、小波神经网络及国内相关模型进行对比,结果表明:所提出的混合预测模型能有效减小太阳辐射预测的误差。

In view of the adverse effect of solar radiation fluctuations on solar energy related systems,this paper proposes a hybrid prediction model based on BP(Back Propagation)and wavelet neural network,in order to predict solar radiation intensity.Firstly,the meteorological factors which have great influence on solar radiation intensity are determined by correlation analysis.Then,BP and wavelet neural networks are optimized to determine the optimal structure of the two neural networks.Next,wavelet decomposition is used to decompose the hourly solar radiation intensity in three layers,and the decomposed components are predicted by the optimized neural network.Finally,the hourly predicted value of solar radiation intensity is obtained by superimposing the output results of each neural network.The simulation is carried out by MATLAB software and the proposed model is compared with BP neural network,wavelet neural network and domestic related models.The results show that the proposed hybrid prediction model can effectively reduce the error of solar radiation prediction.