“This research investigates the application of Random Forest (RF) with automatic hyperparameter optimization for wind speed forecasting and wind energy potential assessment, focusing on a case study in Sfax, Tunisia. The study aims to enhance the accuracy and reliability of wind speed predictions, which are crucial for optimizing wind energy production and making informed decisions on turbine sizing. The methodology involves utilizing the Random Forest algorithm due to its robustness in handling nonlinear relationships in meteorological data. To maximize predictive performance, Bayesian hyperparameter optimization was applied to tune critical RF parameters. Results denote that the optimized RF model mostly follows the trends of measured wind speed and provides a general indication of wind power fluctuations. The final model achieved a Root Mean Square Error (RMSE) of 0.486 m/s, Mean Absolute Error (MAE) of 0.314 m/s and a coefficient of determination (R2) of 0.556. The estimated wind energy potential was 16.72 kWh/m2, and to meet the annual energy demand of a typical refrigerator, a turbine with a swept area of around 77m2 would be required. These findings validate the feasibility of using machine learning-enhanced forecasts for site-specific wind energy estimation and micro-scale energy planning.”