Wind power generation’s inherent volatility and uncertainty present considerable power system scheduling and management challenges. To enhance forecasting accuracy, this work leverages Variational Mode Decomposition (VMD) to analyze wind farm data, facilitating a clearer understanding of the fluctuations in wind power generation. Subsequently, we utilize a Random Forest algorithm to learn from time series data for short-term predictions, thereby improving model stability. We introduce a Random Forest model enhanced by the African Vulture Optimization Algorithm (AVOA-RF) to further optimize the forecasting process. This method emulates the foraging and competitive behaviors of African vultures to explore global optimal conditions effectively. By refining the hyperparameters of the Random Forest model, VMD-AVOA-RF enhances both accuracy and stability. Additionally, it demonstrates strong adaptability to the changing environmental parameters that influence wind power generation, resulting in shorter training times and improved predictive accuracy. Overall, this research addresses the critical need for reliable forecasting in the increasingly variable landscape of renewable energy, facilitating better integration of wind power into power systems.