Artificial Neural Network based Model for Short Term Solar Radiation Forecasting considering Aerosol Index

Solar photovoltaic (PV) power estimation is difficult due to intermittent meteorological parameters. Thus, it is utmost important to predict solar PV power for proper operation and control of renewable energy based power system. Before predicting solar PV system power, forecasting of solar irradiation is essential. In this paper an intelligent model based on ANN is developed to forecast the solar radiation using aerosol index as input data. Based on daily weather classification, the artificial neural network (ANN) using the Nonlinear Auto-Regressive for Exogenous inputs (NARX) approach is implemented to forecast the next day 3-hour solar radiation outputs. To find a precise prediction for 3-hour ahead PV power generation, earlier models were designed by considering temperature, humidity, wind direction and wind speed data. The predicted results of the proposed PV forecasting model are evaluated on the basis of statistical indicators Mean Square Error (MSE). The percentage MSE of the proposed model using Aerosol Index (AI) is 4.67%. Thus, showing that the proposed model is found to be accurate and may be used for PV Power forecasting and also in smart energy management system.