Searching and selecting an adequate methodology for daylight modeling is essential in the design of energy efficient buildings that guarantee the visual, physical and psychological comfort of their occupants. The first step in determining the indoor building illuminance lies in knowing the outdoor illuminance. This dissertation addresses this key aspect through different strategies such as luminous efficacy models and the determination of the angular distribution of the sky's luminance. Daylight is strongly determined by sky conditions. The CIE/ISO standard provides a good general framework to represent the real conditions of the sky, covering the entire probable spectrum of skies, and has been used as a reference throughout this work. The characterization of the skies according to the CIE standard requires experimental measurements of the luminance distribution of the sky, scarcely recorded in terrestrial meteorological facilities. The thesis proposes, as alternatives for the classification of skies according to the CIE taxonomy, the use of meteorological indices, sky images and algorithms based on artificial intelligence. The structure and efficiency of the machine learning algorithms used, both neural networks and decision trees, have been optimized through feature selection procedures in the case of the use of meteorological indices and through image pre-processing techniques, as a step prior to using the classification algorithm. The thesis has also developed a new locally calibrated luminous efficacy model, with excellent results both when used for all-sky types and for clear, overcast and partially overcast sky conditions.