Improved Prediction Approach on Solar Irradiance of Photovoltaic Power Station

Prediction of solar irradiance has great significance to photovoltaic power forecasting and the scheduling plan of power generation. Aim at unsatisfactory prediction accuracy of traditional forecasting methods, this paper presents an approach to predict solar irradiance of photovoltaic power station based on wavelet decomposition and extreme learning machine. With historical irradiance sequence and relative meteorological data as inputs, solar irradiance at intervals 15 minutes is predicted one day ahead. In this approach, historical solar irradiance data is divided through the wavelet decomposition of three layers to obtain detail component and tendency component. Then the prediction models in terms of each component are built based on the extreme learning machine respectively. Finally each prediction value of the component gets final prediction result through wavelet reconstruction. The simulation result coming from the actual measured data of a photovoltaic power station indicates that the proposed model is of higher accuracy in comparison with the traditional ones.