Solar Radiation Estimate and Forecasting by Neural Networks-based Approach

This paper proposes the use of two dynamic Artificial Neural Networks (ANNs) to obtain the estimation and forecast of daily solar radiation. In particular, the Focused Time-Delay Neural Network (FTDNN) and the nonlinear autoregressive network with exogenous inputs (NARX Network) are used. The proposed models are implemented in Matlab® and experimentally validated on the basis of observed data. Both the models provided by the two considered ANNs give good performance. The NARX network gives the further advantage to allow both missing data in times series of solar radiation to be retrieved and future trend of the same quantity to be forecast.