摘要： 粒子群优化算法在搜索全局最优过程中，粒子可能超出界限。针对该情况，提出5种新的受限制的边界条件，将出界粒子随机置于搜索空间内。通过基准函数将这5种边界条件与原有的6种边界条件进行对比测试，并从全局最优和收敛速度两方面对仿真结果进行分析，结果表明，新提出的随机重置的边界条件其性能明显优于置于边界的情况，无形/吸收的边界条件也稍优于其他不受限制的边界条件。

关键词: 粒子群优化算法, 适应度评估, 边界条件, 搜索空间

Abstract: In Particle Swarm Optimization(PSO) algorithm, many particles may exceed the limit of positions when searching the global best. In order to overcome the problem, a new group of restricted boundary conditions, which relocate the arrant particles randomly in the solution space are proposed. The performances of the five new boundary conditions and six existed boundary conditions are tested based on two benchmark functions. Simulation results are examined from both the global best and convergence rate of the algorithm. Comparisons show the performance of the new restricted boundary conditions are much better than that of the boundary conditions relocated in the boundary, and the invisible/absorbing boundary condition edges out other unrestricted boundary conditions.

Key words: Particle Swarm Optimization(PSO) algorithm, fitness evaluation, boundary condition, search space

中图分类号: 

• TP312