A High Performance Computing Based Market Economics Driven Neighborhood Search and Polishing Algorithm for Security Constrained Unit Commitment

This paper introduces a market economics based neighborhood search and polishing algorithm to solve security constrained unit commitment (SCUC). The algorithm adaptively fixes binary and continuous variables and chooses lazy constraints based on hints from an initial solution and its associated neighborhood. A concurrent computing framework is developed to enable parallel neighborhood search and to start the algorithm from multiple initial solutions simultaneously. The initial solutions can come from historical commitments, relaxation or incumbent solutions from a MIP solver (obtained through callbacks), or any other algorithms. Testing on a large set of cases from Midcontinent Independent System Operator (MISO) (including both hourly interval and 15-min interval day ahead cases) on a high performance computing cluster with the concurrent neighborhood search and the polishing algorithm shows significant performance improvements compared to a MIP solver alone.