A Short\|term Power Forecasting Model of Photovoltaic System Based on Improved EMD and Elman Neural Network

A short term power forecasting model of photovoltaic system based on the improved EMD and Elman neural network is proposed. First of all, Historical data is dealt by cluster analysis according to time period and intensity of radiation, and the category of predicted day and relative predicting period of radiation intensity are determined. Then hourly sequences of similar days from the category of the predicted day are built according to principal environmental factors of solar radiation, and the median filtering is carried out using improved EMD algorithm, which is decomposed into different channels according to the fluctuation degree, and channels with similar characteristic are classified as a group. In the end, the radiation intensity of each mode is predicted by using Elman model, and the hourly power output is obtained. This method aims to increase the power forecasting accuracy of photovoltaic system under weak solar radiation circumstance, which can predict the radiation intensity of different days and improve predicting accuracy in certain degree.