This document describes the path I followed to develop a model for the evaluation of the sky luminance distribution on the basis of meteosat information. Considering the available data, there are two possible methods: the single pixel and the multiple pixels as input data. The first method will provide a model that cannot take into account the cloud distribution in the sky vault and the output data will therefore be symmetrical to the sun azimuth. The multiple pixel method, even taking into account the clouds' position, needs higher resolution input data and IR data to evaluate the altitude of the clouds; it gives bad results. I first made the model versus measurements tests for the heliosat models with the Genave's data. The results induced discussions, and I needed a zenith luminance model, I then decided to derive my own models and to base all my work on them.