Accurate prediction of photovoltaic module temperature is essential for optimizing energy production, especially in hot climates where thermal effects significantly reduce panel efficiency. This study compares empirical models and AI-based approaches to estimate module temperature under varying environmental conditions in Marrakech, Morocco. Three physical models of increasing complexity are evaluated, taking into account radiation, air temperature and wind speed. In addition, a deep learning model based on an artificial neural network (ANN) is developed, using meteorological inputs such as global horizontal irradiance (GHI), direct natural irradiance (DNI), air temperature, and wind speed. The models are evaluated using standard error metrics: RMSE, MAPE, and R2. The results show that the ANN model outperforms the empirical models, achieving an RMSE of 1.29 and MAPE of 0.978. The study highlights the critical role of wind speed in improving model accuracy and emphasizes the potential of AI-based solutions to support PV system optimization in the context of Morocco's renewable energy transition.