Short-Term PV Power Forecasting Using Generalized Neural Network and Weather-Type Classification

Abstract

Generation of electricity from solar energy is gaining huge attention because of advancement in the solar photovoltaic technology. Power from solar energy is intermittent in nature and requires a good forecasting method for efficient and reliable operation of smart grid systems. A large number of forecasting approaches are available in the literature. Due to intermittent nature of power obtained from sun, the results obtained from mathematical models for solar PV power prediction are not found satisfactory. An intelligent approach based on generalized neural network (GNN) is proposed and applied for the short-term solar PV power forecasting. Short-term forecasting for an hour to day ahead has applications in energy storage optimization, electricity pricing, etc. Keeping in mind aforesaid, 15 min ahead, short-term solar energy forecasting has been done and presented in this work for smart grid applications. The developed model requires an input of historical data set for PV output power, i.e. solar irradiance, temperature and the relative humidity of the site where solar PV is installed. The performance of the developed PV power forecasting model is evaluated with respect to the accuracy of the developed model for a 1 kWp practical system. Further, the evaluation of proposed method has been performed on the basis of root mean square error (RMSE) and mean absolute error (MAE).