Boosting Wind Power Forecasting using a Dual-Stage Cascaded Support Vector Regression-Gaussian Process Regression Model

Le résumé fourni par la source

Accurate wind power forecasting is essential for maintaining power system stability, optimizing dispatch strategies, and managing costs to ensure a balanced electricity supply and demand. However, existing forecasting models often fail to achieve the required accuracy due to the inherent volatility and uncertainty of wind power generation. To address this challenge, a dual-stage Support Vector Regression– Gaussian Process Regression (SVR-GPR) model is proposed. This hybrid approach leverages the strengths of both kernel functions, enhancing predictive performance. In the first stage, the SVR kernel maps nonlinear input features into a high-dimensional space, while the GPR kernel subsequently refines deterministic predictions by incorporating uncertainty through probabilistic corrections. This cascaded framework improves both accuracy and efficiency in wind power forecasting. Performance evaluation using four quarters of Supervisory Control and Data Acquisition (SCADA) data demonstrates that the SVR-GPR model outperforms standalone machine learning models, offering valuable insights for power system dispatch optimization.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.