Performance estimation technique for solar-wind hybrid systems: A machine learning approach

Abstract

This study introduces a novel method to assess solar-wind hybrid energy systems, utilizing data analysis for key metrics (Performance Yields YR, Capacity Factor CF, and Performance Ratio PR), focused on their uncertainties estimation. Employing a bootstrapped-based K-means algorithm on features from a local weather station in Universidad de La Salle, Bogotá, the approach evaluates the technical feasibility of a photovoltaic system on the campus. It discusses the clustering algorithm advantages, emphasizing its suitability for modeling continuous processes in renewable energy systems. The hypothesis assumes independence and identically distributed centroid coordinates, aligning with large-scale energy system characteristics. Methods include data normalization, feature selection, statistical calculations, bootstrapping, and Probability Density Functions (PDF) estimation through a Maximum Likelihood Estimator (MLE) to assess uncertainties. Results suggest the unfeasibility of a wind energy system due to technological constraints. The study evaluates system flexibility, and design considerations, providing valuable decision-making insights. The conclusion highlights the potential use for accurate estimations, enhancing the system reliability.

Keywords

• Bootstrap method
• Clustering analysis
• Hybrid power systems
• Machine learning
• Probability density function
• Solar power generation