基于相空间重构小波神经网络的太阳逐日辐射能量预测

DAILY SOLAR RADIATION PREDICTION BASED ON PHASE SPACE RECONSTRUCTION OF WAVELET NEURAL NETWORK

将神经网络非线性学习能力与小波变换多尺度分解特性相结合,建立太阳逐日辐射能量小波神经网络预测模型。通过小波多尺度分解使太阳逐日辐射能量序列在一定尺度上表现出准平稳性,以太阳逐日辐射能量相空间重构数据作为模型的输入量对网络进行训练。仿真结果表明该方法可较好地用于太阳逐日辐射能量预测。

Solar radiation prediction is a non-liner and non-stationary process. It＇s hard to predict with high precision. A wavelet neural network model was set in this paper. The non-liner process of solar radiation was forecasted by neural network and the non-stationary process of solar radiation was decomposed into quasi-stationary at different scales by multi-scale characteristics of wavelet transform. The model was trained with phase space reconstruction radiation data. Simulation results indicate that the method is satisfactory for the prediction of solar radiation.