A Modified Particle Swarm Optimization Algorithm for Global Optimizations of Inverse Problems

Abstract

Particle swarm optimization (PSO) is a population-based stochastic search algorithm inspired from the natural behavior of bird flocking or fish schooling for searching their foods. Due to its easiness in numerical implantations, PSO has been widely applied to solve a variety of inverse and optimization problems. However, a PSO is often trapped into local optima while dealing with complex and real world design problems. In this regard, a new modified PSO is proposed by introducing a mutation mechanism and using dynamic algorithm parameters. According to the proposed mutation mechanism, one particle is randomly selected from all personal best ones to generate some trial particles to preserve the diversity of the algorithm in the final searching stage of the evolution process. Moreover, the random variations in the two learning factors will help the particles to reach an optimal solution. In addition, the dynamic variation in the inertia weight will facilitate the algorithm to keep a good balance between exploration and exploitation searches. The numerical experiments on different case studies have shown that the proposed PSO obtains the best results among the tested algorithms.