A Improved Particle Swarm Optimization and its Application in the Parameter Estimation

Abstract

For the issues of inferior local search ability and premature convergence in the later evolution stage of the traditional particle swarm optimization, an improved particle swarm optimization is proposed and applied to the parameter estimation. Firstly, in the evolutionary process of particle swarm optimization, the particles which have crossed the border are buffered according to the speed. Then, each particle is performed mutation in different probability according to the evolutionary generations, which can keep the diversity of the particle swarm and avoid the premature convergence effectively. Thirdly, a crossover operation is conducted between the current best particle and the particle which is selected from the particle swarm with a certain probability, which can lead particles gradually approaching to the extreme point and hence, the local search ability of the algorithm will be improved. The advanced particle swarm optimization is applied to the parameter estimation of the kinetic model in the Hg oxidation process and the application result show the effectiveness of the suggested algorithm.