基于划分和重分布的粒子群算法及优化策略

An improved particle swarm optimization based on division and redistribution

提出了一种新的基于划分和重分布的粒子群优化算法.新算法将粒子划分为普通和优化两类.普通粒子随机产生,速度快,侧重全局搜索;优化粒子紧随群体最优并且速度较慢,侧重局部收敛,以提高收敛精度.当群体最优未发生变更的时间过长时,在保持群体最优的同时将粒子重新分布,以摆脱过早的局部收敛.对典型函数的测试结果表明,新算法没有增加复杂度,在摆脱解的早熟和提高解的收敛精度等方面优于基本粒子群算法.

The authors proposes a novel Particle Swarm Optimization algorithm based on Division and Redistribution（DRPSO）. It divides particles into two classes, i.e. common ＆ optimized particles. Common class is with high velocity and good for global search. Optimized class is with slow velocity and good for local search. When the time of the optimized solution keeps constant for too long time, the best global solution is saved and particles will be redistributed with randomly to get out of local convergence. Experiments show that new method is much better than traditional Particle Swarm Optimization, It improves convergence precise degree of solution and does not increase complexity of algorithm.