摘要
粒子群优化(Particle Swarm Optimization,PSO)算法是一类性能优越的寻优算法。但由于早熟问题,影响了算法性能的发挥。针对这一问题,通过获取粒子群的状态信息,来控制PSO进化过程,是一种有效的PSO改进方法。但现有的方法是从单一的角度来描述粒子群进化状态,使用时还具有一定的局限性。为了更进一步发挥PSO算法的优越性能,充分考虑了粒子群进化状态中的不同信息,根据证据融合理论,提出一种PSO算法(称为DS_PSO)。首先根据全局和局部搜索的要求,把算法分为不同的搜索模式;然后,在进化过程中,对描述粒子群的不同参数进行D-S融合。根据融合结果,确定粒子群状态,选择合适的搜索模式。对测试函数的仿真实验表明,与对比方法相比较,DS_PSO算法具有更好的收敛精度和更快的进化速度。
Particle Swarm Optimizer （PSO） is a probability algorithm with excellent performance. But the premature phenomenon limits the effect of PSO. An effective method for solving this problem is to obtain the information of PSO for controlling PSO evolution process. But the method introduced now describes the states of particle swarm evolution only from single sides, its effects are limited. In order to utilize the PSO algorithm, this paper considers all informations of PSO evolution process, and based on the D - S theory, introduces a new method of PSO（named DS - PSO）. First , different searching modes are introduced according to the requirement of global and detail searching. Secondly, different parameters for describing particle swarm evolution are fused. According to the fused solution, suitable searching mode is selected. The simulation test shows： compared with the contrast method, DS - PSO algorithm has better convergence accuracy and higher evolution velocity.
出处
《计算机仿真》
CSCD
2007年第2期162-164,182,共4页
Computer Simulation
基金
全国优秀博士学位论文作者专项资金资助
(200237)
关键词
粒子群优化
证据理论
收敛精度
进化速度
PSO
Theory of evidence
Convergence accuracy
Evolution velocity
