粒子群算法中粒子轨迹特性研究  

Characteristics of Particle Trajectory in Particle Swarm Optimization

作　　者：韩璞[1] 孟丽[1] 王彪[1] 王东风[1] 

机构地区：[1]华北电力大学控制与计算机工程学院,河北保定071003

出　　处：《计算机仿真》2015年第12期235-240,共6页Computer Simulation

摘　　要：在粒子群优化算法中,惯性权重和学习因子的选取对粒子轨迹的特性有很大的影响,进而会影响算法的性能。本文对粒子群优化算法中不同参数区域下的粒子的轨迹特点进行了研究。首先将算法的随机性和进化性进行简化,利用Z变换和离散时不变系统理论对算法的参数区域进行了划分,详细说明了各个区域中的简化模型的粒子轨迹特征。之后在简化模型的基础上,依次加入了算法随机性和进化性的影响,还原了粒子群算法在优化过程中的真实的粒子轨迹形式。In particle swarm optimization algorithm, the adjustable parameters, such as inertia weight and acceleration coefficients, have significant influence on the property of particle trajectory and the algorithm performance. In this paper, the characteristics of particle trajectories with different parameters are studied. First the randomness and evolution property of the algorithm are simplified for an analysis purpose. Z transform and the theory of discrete time system are used to analyze the simplified system and the whole common parameter region is divided into several sub - areas. The characteristics of particle trajectories with parameters from different sub - areas are studied in detail. Then based on the analysis of the simplified model, the randomness and evolution property are called back and the particle trajectories in the actual optimization process are restored.

关 键 词：粒子群优化 粒子轨迹 参数区域 收敛特性 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]