Modeling random fluctuations of the solar irradiance has already been the object of several studies published in the literature. These are based mostly on the random processes. The Markovian approaches in particular, contributed extensively to this modeling. One can see for example, the works of Brinkworth (1977), Bartoli et al. (1981), Lestienne et al. (1979), Aguiar et al. (1988) and Maafi (1991). This last reference treated the problem of the classification of the insolation and the daily irradiation indirectly by joining them to the states of the sky: clear sky, covered sky, etc. (Maafi 1991, 1998).
Other statistical methods were used for classification of typical meteorological days such as automatic classification (Bouroubi 1998), the analysis of the correlations (Louche and al. 1991) and the Ward’s method (Muselli et al. 1991).
More recent studies are interested by the modeling of the random character of the solar radiation using neural networks (Guessoum et al. 1998; Sfetsos and Coonick 2000). In addition to the originality of these new approaches, these studies aim to value the contribution of their formalisms in the description of the solar radiation fluctuating character.
However, very few works treating the classification of the solar radiation signals using the fractal analysis were published (Maafi and Harrouni 2000, 2003; Harrouni and Guessoum 2003; Harrouni and Maafi 2002; Louche et al. 1991). In this section the contribution of the fractal analysis to the classification of the solar irradiance signals is given. This examination leads to the determination of different sky types in a given time interval as: clear sky, partially covered sky, covered sky etc. which is useful for planning and analyzing solar energy systems. Hence, a classification method is proposed which allows the categorization of the solar radiation fluctuations based on the fractal dimension (Harrouni et al. 2005).