This paper introduces a dynamic prediction of solar radiation and ambient temperature. Medium term prediction is based on climatic parameters behaviours during the day before and on time distribution models. As for short term prediction, it is ensured by an ARMA predictor using Kalman filter. Herein, we describe our method and present prediction results. Validation is based on measures taken during the year 2005 in the north of Tunisia. The work effectiveness is illustrated by a short term and medium term prediction of the electric energy produced by a 1 KWp photovoltaic panel (PVP) installed at the Energy and Thermal Research Centre (CRTEn) in the north of Tunisia. Since our work delivers accurate climatic parameters prediction, the obtained results can be easily adapted to predict any other solar conversion system output. (c) 2008 Elsevier Ltd. All rights reserved.