Hourly Global Radiation Prediction from Geostationary Satellite Data

The possibility to generate 2-D predictions of solar radiation one hour ahead and over a large geographic area is the focus of this study. In that view, hourly satellite maps of irradiations extracted from the Heliosat-3 have been used as training data and inputs of artificial neural networks (ANN) and as inputs for apersistence model. These models were computed for each point of the map, allowing generating solar radiation predictions maps. Comparisons between the computed maps and satellite estimations ones were proceed to evaluate the performances. The first results show that ANN and persistence have similar performances in a majority of pixel.