Estimation of hourly global solar irradiation on tilted planes from horizontal one using artificial neural networks

Calculating global solar irradiation from global horizontal irradiation only is a difficult task, especially when the time step is small and the data are not averaged. We used an Artificial Neural Network (ANN) to realize this conversion. The ANN is optimized and tested on the basis of five years of solar data; the accuracy of the optimal configuration is around 6% for the RRMSE (relative root mean square error) and around 3.5% for the RMAE (relative mean absolute value) i.e. a better performance than the empirical correlations available in the literature.

► ANN (Artificial Neural Network) methodology applied to hourly global solar irradiation in order to estimate tilted irradiations. ► Model validation with more than 23,000 data. ► Comparison with “conventional” models. ► The precision in the results is better than with empirical correlations. ► 6% for the RMSE (root means square error) and around 3.5% for the RMAE (Relative Mean Absolute Value).