Wind power is an increasingly viable source of renewable energy, and accurate forecasting of wind power generation is essential for maintaining a balance between supply and demand in smart grids. However, the inherent fluctuations and intermittent nature of wind pose significant challenges to precise forecasting. As the integration of renewable energy sources into power systems intensifies, the demand for accurate short-term forecasts of wind power generation becomes even more critical. While several single-model approaches exist for forecasting wind power, including ARIMA, support vector machine (SVM), exponential smoothing (ETS), Persistence Model, Wavelet Transform, and Least Absolute Shrinkage and Selection Operator (LASSO), the hybridization of these models has emerged as a promising alternative for improving forecasting accuracy. This study explores two hybrid forecasting methods: a combination of random forest and XGBoost, and the integration of SVM with ARIMA. The performance of these models is rigorously assessed using statistical indicators such as the Durbin-Watson statistic and the Jarque–Bera test, along with key evaluation metrics, including the coefficient of determination (R2), root mean square error (RMSE), mean absolute error (MAE), and mean square error (MSE). The SVM-ARIMA model demonstrates superior accuracy and reliability compared to the RF-XGBoost model, with an MSE that is 7.32% lower, an MAE that is 10.43% lower, and an RMSE that is 7.03% lower. Additionally, SVM-ARIMA achieves a 4.55% improvement in R2, highlighting its enhanced pattern recognition capabilities. Consequently, SVM-ARIMA is identified as the preferred model for accurate wind power forecasting, particularly in addressing the challenges associated with the intermittent nature of wind.