An innovative two-stage machine learning-based adaptive robust unit commitment strategy for addressing uncertainty in renewable energy systems

Abstract

Confronting the challenge of intermittent renewables, current unit commitment practices falter, urging the development of novel short-term generation scheduling techniques for enhanced microgrid stability. This study presents an adaptive robust unit commitment approach using machine learning techniques for renewable power systems, computing the Calinski-Harabasz index to identify prediction inaccuracies related to intermittent sources. The uncertainties are subsequently grouped together using the spatial clustering tool, and the average density of the K-means distribution is calculated. The clustering of points in space, considering noise, discrete uncertainty in renewable energy sources, and outliers within the comprehensive uncertainty set, is addressed via a nonparametric algorithm. The implementation of established methodologies and frameworks, in conjunction with density-based spatial clustering of applications with noise, introduces an innovative method for vulnerability clustering. This methodology guarantees that every cluster aligns with data pertaining to vulnerabilities of renewable energy sources. The performance of the suggested method is showcased by conducting experiments on modified IEEE 39-bus and 118-bus test systems that use intermittent wind power. The results demonstrate that the proposed framework may lower the cost of robustness by 8–48% compared to traditional robust optimization techniques. The results of stochastic programming showed that the optimized system with a stable economic organization would have 75 % faster calculations.

Keywords

• Renewable energy
• Robust unit commitment
• Machine learning
• Disjunctive data uncertainty
• Optimization