In this paper a model devoted to the forecast of solar radiation and Photovoltaic (PV) power has been addressed. In particular, for what concerns the solar radiation prediction, the novelty of the approach stays in the use of the clear sky model proposed by Hottel fed by the output of a data driven algorithm. In this hybrid approach, key parameters are computed through the exploitation of a database of solar radiation values. The PV power is then forecasted through an Auto Associative Kernel Regression (AAKR) technique. The proposed model has been tested for a real application case, a PV plant located in South Africa, and the results have highlighted high effectiveness of the proposal.