The prediction of daily solar irradiance is a fundamental and very important task in solar power plants operation and control. Various models and solutions, both linear and nonlinear, have been developed for this purpose. In this paper we propose a new hybrid approach including estimating the normalized profile of the daily irradiance and the estimating the peak value of daily irradiance. For daily profiles estimation, a extended fuzzy self-organizing network will be used, and for daily peak irradiance estimation, a Support Vector Regression model will be used with average relative error less than 3.3%. The proposed solution was test on real data taken from Ninh Thuan province.