分组信息共享的量子粒子群优化算法的改进

摘要

Quantum-behaved particle swarm optimization(QPSO) suffers the drawback of premature-convergence.To overcome the problem,a novel improved particle swarm optimization based on quantum(PSO-Q) is presented.In PSO-Q,a new group strategy is proposed to supervise all the particles updating the positions according to different search mechanism.Thus PSO-Q can enhance the local and global search abilities.All the particles in the swarm share the useful information between the two groups and the achieved information to balance exploration and exploitation.Without decreasing the accuracy of search,group strategy enlarges search space in the evolutionary process.The particles in one group can keep the basic search ability to develop the achieved space.The other group particles utilize the efficient message in whole swarm to explore new regions and avoid reducing the diversity of the swarm.In the experiment on benchmark functions,comparison results prove the powerful and potential search ability of the proposed algorithm with higher optimization accuracy.%标准量子行为的粒子群优化(Quantum-behaved particle swarm optimization,QPSO)算法依然存在早熟收敛的缺点,针对此问题,提出了一种改进的量子粒子群算法(Particle swarm optimization based on quantum,PSO-Q).在PSO-Q算法中,采用分组策略基于不同的更新公式同时提高局部搜索和全局搜索能力,并且共享组间有用的信息,达到探索与开发能力的平衡.在不降低搜索精度的情况下,分组策略扩大了种群搜索过程中的搜索范围,其中一组保持QPSO搜索方法的基本搜索能力,主要开发已有搜索空间.另外一组共享整个群里的有效信息,增加新领域探索能力,可以避免种群多样性的不断下降.在标准测试函数的对比实验中,仿真结果表明该算法具有较强的搜索能力并且达到了较高的优化精度.