An Effective Two-Stage Column Generation Strategy for Solving Unit Commitment

Unit commitment (UC) has been extensively studied in power systems literature and is typically formulated as a mixed-integer programming (MIP) problem. However, existing methods and commercial MIP solvers often struggle to efficiently solve large-scale UC instances. This paper proposes a novel MIP model for addressing UC by employing an innovative column generation-based two-stage approach to efficiently compute high-quality solutions. Our method leverages parallel computing to solve pricing problems and generate columns effectively. Following column generation, we solve the master problem as a MIP model to obtain high-quality feasible solutions for the UC problem. Comprehensive numerical experiments demonstrate that our proposed method can effectively tackle large-scale UC problems involving over 1,000 units, significantly outperforming state-of-the-art MIP solvers in both solution quality and computational efficiency. The results underscore the potential of our approach for practical application in large-scale power system scheduling.