Renewable energies occupy a prominent place in global sustainable development strategies, aiming to meet growing energy needs while reducing the carbon footprint. This study focuses on the integration of the solar photovoltaic source in renewable energy systems (RES), aiming to improve its performance and maximize energy efficiency through optimization techniques and advanced predictive modeling. The main objective is to develop precise and efficient methods for predicting solar irradiance, one of the most important parameters influencing the performance of PV-wind systems. By exploiting experimental data of wind speed, solar irradiance and temperature in the Sahara region of Tunisia (North Africa), we developed and compared various machine learning (ML) models to evaluate their predictive performances. This evaluation will allow to manage the technical characteristics of the main components of the RES and its efficiency. The results show that the Gradient Boosting R Regression (GBR) model provides the most accurate predictions within the system.