一种具有自我更新机制的量子粒子群优化算法

Quantum-behaved Particle Swarm Optimization algorithm with self-renewal mechanism

自然界中生命体都存在着有限的生命周期,随着时间的推移生命体会出现老化并死亡的现象,这种老化机制对于生命群体进化并保持多样性有重要影响。针对量子行为粒子群(QPSO)算法中粒子存在老化并使得算法存在早熟收敛的现象,将生命体的自我更新机制引入了QPSO算法,在粒子群体进化中提出领导者粒子和挑战者粒子,随着群体粒子的老化,当领导者粒子领导力耗尽不能引导群体进化时,挑战者粒子通过竞争更新机制成为新的领导者粒子引导群体进化并保持群体多样性,并证明了算法的全局收敛性。将提出的算法与多种典型改进QPSO算法通过12个CEC2005 benchmark测试函数进行比较,对结果进行了分析。仿真结果显示,该算法具有较强的全局搜索能力,尤其在7个多峰测试函数中,综合性能最优。

Life body has limited life in nature;it will be aging and die with time. The aging mechanism is very important to keep swarm diversity during evolutionary process. For the phenomenon that Quantum-behaved Particle Swarm Optimization（QPSO）is often premature convergence, self-renewal mechanism is proposed into QPSO, and a leading particle and challengers are introduced. When the leading power of leading particle is exhausted, one challenger will select to be the new leading particle and continues keeping the diversity of swarm with a certain renewal mechanism. Furthermore,global convergence of the proposed algorithm is proved. Finally, the comparison and analysis of results with the proposed method and classical improved QPSO algorithm based on twelve CEC2005 benchmark function is given, the simulation results show stronger global searching ability of the modified algorithm. Especially in the seven multi-model test functions, the comprehensive performance is optimal.