Based On The Dynamic Characteristics Of The Particle Swarm Optimization Algorithm

Particle swarm optimization (PSO) is a kind of swarm intelligence algorithm and a new optimization tool which was proposed on the simulation of birds flocking predation social behavior. PSO has drawn wide attention since it was proposed due to its fast convergence speed, convenient parameter adjustment and stable optimization character. It has been widely applied to various fields and has gradually become the hotspot of swarm intelligence researches.After the study of PSO algorithm theories and applications at home and abroad, two improved points of this algorithm were proposed based on optimization properties and shortcomings of this algorithm, the detailed contents were as follows:(1) In order to raise its convergence speed and prevent premature convergence, an improved PSO algorithm was proposed based on inertia weight method in the third chapter of this paper. As the traditional linearly decreasing weight (LDW) strategy cannot reflect the actual complex situation of PSO optimization process, dispersion degree factor and optimization speed factor were introduced as two affecting factors to demonstrate the optimization statement. Furthermore, the function of these two factors can be expressed by the inertia weight, and then the extended spatial ability of the particle can be dynamically adjusted according to the changes of inertia weight. Finally, the simulation comparison has been made using several testing functions with improved algorithms, and the results show that the improved algorithm has better optimization performance.(2) In view of the disadvantages of PSO, such as the prior information deficiency of optimization problems and the blindness of particle searching performance, an improved particle swarm optimization algorithm was proposed in the fourth chapter of this paper based on the neighborhood model and rough set theory. As the prior information of particle swarm searching space can be easily obtained using neighborhood method, the improved algorithm has better fitness value of the problem spatial regions. Additionally, the optimization region of particle swarm can be dynamically narrowed by ignoring those regions with poor fitness based on knowledge reduction property of the rough set theory. The particle swarm optimization with more efficiency was achieved by continuous adjusting its search space of optimization. In the end, the optimization of ten classical testing functions and the comparison between it and the LDW-PSO optimization process have been carried out, and the result proved the effectiveness and feasibility of the improved algorithm.