A Local Search and Cutting Plane Approach for Security-Constrained Unit Commitment Problems

Unit commitment (UC) problems are usually modeled as mixed 0-1 linear programs and solved with branch-and-bound (B&B) solvers. The number of binary variables in a practical unit commitment problem could be very large, which poses great challenges to B&B algorithms. By providing good upper bounds, heuristics are used to accelerate B&B procedures in practice. In this paper, we propose a local search and cutting plane (LS&C) heuristic to produce high-quality feasible solutions to unit commitment problems. In this heuristic, a cutting-plane generation subproblem and a pseudo-boolean optimization subproblem are alternatively solved to approach feasible solutions. The experimental results of a 30-bus case, a 118-bus case, a 1354-bus case, and a 3375-bus case demonstrate that, compared to MindOpt with and without heuristic module, LS&C can find a higher-quality initial feasible solution in a shorter amount of time, which can be integrated into B&B solvers to accelerate the solution of SCUC problems.