A HYBRID BAYESIAN REGRESSION AND KALMAN FILTERING APPROACH FOR ROBUST WIND POWER FORECASTING

Wind power is a critical renewable energy source for sustainable energy systems, but its variability and intermittency create challenges for accurate forecasting, essential for reliable grid operations and cost management. This study presents a novel hybrid model integrating Bayesian Linear Regression (BLR) with the Kalman Filter (KF) for probabilistic wind power prediction. BLR provides an uncertainty-aware baseline forecast, refined by KF for real-time adaptability to wind variations. Key contributions include combining Bayesian inference for uncertainty quantification with Kalman filtering for dynamic adjustments. Hyperparameter tuning via grid search and crossvalidation enhances model performance. The model's effectiveness is assessed using metrics such as Mean Squared Error (MSE), Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and the Coefficient of Determination (R²). Residual validation through the Durbin-Watson statistic and ANOVA underscores model reliability. Comparative results indicate the hybrid model achieves an RMSE of 7.5%, outperforming ARIMA (10.5%) and neural networks (8.2%). These findings highlight the model's robustness and accuracy, addressing critical challenges in wind power forecasting. By leveraging probabilistic forecasting and real-time adaptability, the hybrid model offers a scalable solution for smart grid integration. Its superior performance makes it a promising tool for renewable energy management. Future work will explore expanding the model to other renewable domains and incorporating additional meteorological features to further enhance prediction accuracy and robustness.