Wrapper-based feature selection methods outperform simpler approaches for renewable energy prediction, and the proposed CSFS method achieves comparable accuracy to existing wrappers with significantly lower computational overhead.
This paper addresses feature selection for renewable energy prediction by proposing CSFS, a clustering-based wrapper method that automatically identifies the most important variables for wind and solar power forecasting. The approach matches the performance of traditional methods while reducing computational costs by 21%, making it practical for real-world energy systems.