This paper proposes a method for evaluating the available solar energy according to different type of sky (clear sky, moderately covered sky and covered sky). Four semi-empirical models for estimating solar radiation (Link Cloudiness Factor (LCF), Perrin-Brichambaut (PB), Ghouard (GD), BirdHulstrom (BH)) have been used to predict solar radiation accoording to the sky conditions. An analysis of the influence of seasonal variation on the quality of available solar energy potential is carried out. The study was done for two seasons (dry seasons and rainy season) identified for the Nouakchott site. The proposed approach is applied and validated for the case of the Nouakchott site by considering the solar radiation data measured over the period of one year . The results showed that for the rainy season, the model of the LCF is the best for the clear sky case with NMSE (0.073) and R (0.975). However, for the dry season, the model of BH is most suitable with NMSE (0.069) and R (0.973). For a moderately covourt sky, the GD model is best indicated with NMSE (0.186) and R (0.933) during the dry season. For the rainy season, BH gives better results with NMSE (0.242) and R (0.911). Finally for a covered sky, the most adapted models are, that of GD for the dry season with NMSE (0.880) and R (0.831), and the model of PB for the rainy season with NMSE (0.615) and R (0.831).

Keywords: Solar energy, prediction model, solar radiation, data