Hybrid Particle Swarm Optimization Algorithm And Its Application

The particle swarm optimization proposed by Kennedy and Eberhart aims to study animal behavior. The algorithm has the advantages of simple principle, esay implement and fewer parameters. Now, the particle swarm optimization has become a new hotspot in of optimization and successfully applied in many areas. But, the algotithm exists slow convergence and easily fall into local optima. So, we minly discuss the improvement of the algorithm.In this thesis. We analyze the basic principle swarm optimization and summary the various aspects of the improved algorithm, then propose three modified algorithms: First, to overcome the particle swarm optimization algorithm to the local optima, an improved algorithm is proposed by introducing the swarm behavior and disturbances. The new algorithm improves the ability to avoid the local optima and speeds up the convergence; Second, in order to overcome the premature convergence of particle swarm optimization algorithm, other improved algorithm is proposed by introducing mutation and crossover operators. This algorithm not only effectively solves the premature convergence problem, but also significantly speeds up the convergence; Third, to overcome the slower convergence or lower accuracy or easy to the local optima in the later evolution period, a new algorithm is proposed by introducing multiple population and differential evolutio, which divides the whole population into many subpopulations and handles the current optimum positions with mutation and crossover. Several benchmark functions are tested and the experimental results show that the new algorithm significantly speeds up the convergence and the accuracy of conergence.