Relative Velocity Updating in Parallel Particle Swarm Optimization Based Lagrangian Relaxation for Large-scale Unit Commitment Problem

Abstract

This paper presents an effectiveness of combined parallel relative particle swarm optimization (PRPSO) and Lagrangian relaxation (LR) for a large-scale constrained unit commitment (UC) problem in electric power system. The proposed algorithm incorporates PRPSO with a new relative velocity updating (RVU) approach to tradeoff the solution of each slave processing unit. The parallel algorithm based on the synchronous parallel implementation is developed to consider the neighborhoods decomposition of multiple particle swarm optimizers. The proposed PRPSO divides the neighborhood into sub-neighborhood so that computational effort is reduced and UC solutions are remarkably improved. The proposed method is performed on a test system up to 100 generating units with a scheduling time horizon of 24 hours. The numerical results show an economical saving in the total operating cost when compared to the previous literature results. Moreover, the proposed PRPSO based RVU scheme can considerably speed up the computation time of a traditional PSO, which is favorable for a large-scale UC problem implementation.