Daily Direct Solar Radiation Exposure Prediction Based On Phase Space Reconstruction Of Wavelet Neural Network

Daily direct solar exposure prediction is very important for the outputsituation analysis of grid connected photovoltaic system. It is widely researched bydomestic and foreign scholars. The nonlinear, large scale intermittentnon-stationary and chaotic characteristics make it difficult to establish its accurateprediction model. Trial and error method which usually used to construct thetraditional prediction neural network model is lack of theoretical guidance. Waveletneural network and phase space reconstruction were combined to build theprediction model of daily direct solar exposure. This method provides a new way tothe construction of neural network and improves the prediction accuracy of dailydirect solar exposure.The main works in this paper are as follows:1. Analyzed the non-linear and non-stationary characteristics of daily directsolar exposure sequence. Power spectrum method, principal component analysisand Lyapunov exponent method were used to determine its chaotic properties.2. Phase space reconstruction embedding dimension of daily direct solarexposure was used to determine the input layer neurons. Wavelet analysis theorywas used to determine the hidden layer neurons. This method get ride of theshackles of trail and error method and provide a way of building neural network.3. Phase space reconstruction and wavelet neural network were used to buildthe predictive model of daily direct solar exposure. Parameter adjustment algorithmwas derived. The reconstructed daily direct solar exposure data was used to trainthe model and made prediction of daily direct solar exposure.Daily direct solar exposure data from2000to2004that form U.S. NationalAeronautics and Space Administration longitude117o, latitude31o(Hefei region)was used in this paper. Matlab R2009b simulation software was selected to makeverification experiment of daily direct solar exposure wavelet neural networkprediction model. Results show that the method is feasible and effective.