RÉSUMÉ / ABSTRACT :
Renewable energy sources have a variable nature and are greatly dependingon weather conditions. The load is also uncertain. Hence, it is necessary to use power reserve equipment to compensate unforeseen imbalances between production and load. However, this power reserve must be ideally minimized in order to reduce the system cost with a satisfying security level. The quantification of power reserve could be calculated through analysis of forecasting uncertainty errors of both generation and load. Therefore, in this paper, a back propagation artificial neural network approaches is derived to forecast solar radiations. Predictions have been analyzed according to weather classification. Some error indexes have been introduced to evaluate forecasting models performances and calculate the prediction accuracy. Forecasting results can be used for decision making of power reserve for renewable energy sources system with some probability or possibility methods.