求解动态优化问题的改进多种群引力搜索算法

Improved multi-population gravitational search algorithm for dynamic optimization problems

针对目前多种群算法解决动态优化问题时存在过多冗余计算、寻优精度低等缺陷,提出多种群串行搜索的引力搜索算法。采用多种群串行搜索的策略,便于当前子种群利用其他已收敛种群的进化信息。为解决多峰重复搜索而带来的冗余计算问题,提出具有约束条件的初始化策略,给予初始化的粒子以方向性的指引,避免其初始化在已寻峰区域;采用距离判决的策略发现并终止多峰重复搜索。为全面的监测环境变化及解决多样性丢失问题,提出一种监测环境策略及追踪策略。研究结果表明:所提算法,面对不同的环境变化程度以及不同的峰值数量,其求解精度都优于其他7种对比算法的求解精度,证明该算法在求解动态优化问题上的优越性。

To improve the redundant computing and low accuracy of solving dynamic optimization problems（DOPs） for multi-population algorithm, a novel improved multi-population gravitational search algorithm（IMGSA） was proposed. In IMGSA, the multi-population serial strategy was good for the present subpopulation to use evolutionary information of convergence population. A constraint initialization strategy was proposed to reduce the redundant computing which was generated by multiple populations searching repeatedly. Simultaneously, a distance decision strategy was used to stop multiple populations searching. Eventually, a monitoring and tracking strategy was used to monitor the environmental change and track the local peaks. The results show that IMGSA has a better performance in solving DOPs than those of other seven dynamic algorithms in different degree of environmental change or different peak number. It can prove the validity of proposed algorithm.