Elitist mutated particle swarm optimisation algorithms: application to reservoir operation problems

The particle swarm optimisation algorithm is a global optimisation method proposed and mainly used for continuous optimisation problems. The method shows interesting features, in particular fast convergence characteristics. It does, however, lack sufficient exploration, especially when the grouping of the swarm starts leading to sub-optimal solutions when solving difficult problems. This paper introduces two mutation mechanisms to balance the exploitative characteristics of the algorithm. The timing of the proposed mutations is designed such that the inherent exploration of the method is not disturbed. The proposed mutated algorithms cannot, therefore, produce inferior results to that of the original method. Furthermore, mutation is only carried out on those particles that are already converged to the global best position and, in effect, are not of any particular use to the collective intelligence of the swarm. The performance of the proposed mutated algorithms is tested against two reservoir operation problems and the results are presented and compared with those of the standard algorithm. The mutated algorithms show improved performance for the examples considered.