Metaheuristic search based methods for unit commitment

This paper presents two new solution approaches capable of finding optimal solutions for the thermal unit commitment problem in power generation planning. The approaches explore the concept of matheuristics, a term usually used to refer to an optimization algorithm that hybridizes (meta)heuristics with mixed integer programming solvers, in order to speed up convergence to optimality for large scale instances. Two algorithms are proposed: local branching, and an hybridization of particle swarm optimization with a mixed integer programming solver. From extensive computational tests on a broad set of benchmarks, the algorithms were found to be able to solve large instances. Optimal solutions were obtained for several well-known situations with dramatic reductions in CPU time for the larger cases, when compared to previously proposed exact methods. (C) 2014 Elsevier Ltd. All rights reserved.