Improvement Of Particle Swarm Optimization And Application In A Dynamic Environment

Particle swarm optimization (PSO) is a random optimization algorithm, simulating foraging process of bird flocking. Due to its easy implementation and fast convergent speed, it has been successfully applied to many areas. This paper studied the PSO algorithm from the three aspects—structure optimization, parameter adjusting and application research.Structure optimization is one of the PSO research fields. Many experiments are conduct to verify the poor performance of standard PSO (SPSO)on the functions whose optimum values are not in the origin of coordinates or the center of search area. In order to overcome the advantage, space-dividing particle swarm optimization is proposed by introducing a factor called"0 search arithmetic operator"and two strategies named"Domain divided strategy"and"Coordinate dummy reorientation strategy". Simulation results of six test functions show the efficiency of above modification.Parameter adjustment is an important research direction of PSO. Many adjustment strategies of the two parameters (i.e. cognitive parameter and social learning factors parameter) proposed recently can improve the algorithm performance. However, those linear adjustment strategies may not work well on the complex optimization problems. Therefore, this paper proposed a time-varying accelerator coefficient adjustment strategy, in which that the above two parameters are adjusted by a predefined predicted velocity index dynamically, moreover, if the average velocity of particle is superior to the index, its two parameters will become convergent, and vice versa. Simulation results verify that the proposed algorithm is more effective and efficient than other three variants of particle swarm optimization when the algorithm is applied to solve multi-modal high-dimensional numerical problems. Application research of PSO is another hotspot. The problem of real world is seldom static due to the disturbance of many uncertainties. Therefore, the discussion for algorithm performance in dynamic environment is concerned as an application research. This paper applied the above-mentioned algorithms and SPSO algorithm in three different dynamic environments, and the conclusions show that the dynamic supper performance of all the selected algorithms.