基于搜索策略的烟花算法研究

Research on fireworks algorithm based on search strategy

为了克服烟花算法容易早熟,提高其寻优精度,提出一种基于搜索策略的烟花算法。首先,通过最小爆炸半径检测,得到种群适应度值。其次,在烟花种群多次迭代过程中对当前最佳烟花个体进行动态随机搜索,增强对当前阶段最佳个体邻域范围内的搜索。最后,根据当前最佳个体之间的拥挤程度,存留10%的最佳个体,对剩余烟花个体采用佳点集策略进行初始化操作,辅助种群个体逃离局部最优。实验结果表明,所提算法相比同类烟花算法有效提高了求解精度,且收敛速度较快。

A fireworks algorithm based on search strategy is proposed to overcome the premature explosion of the fireworks algorithm and improve its optimization accuracy. The fitness value of the population is obtained by detecting the minimum explosion radius. In the course of the multiple iterations of the fireworks population,the dynamic random search for the current best fireworks individuals is carried out to enhance the search within the neighborhood scope of the optimal individuals at the current stage. The 10% best fireworks individuals are retained according to the current crowdedness degree among them,and the rest is initialized by means of the good point set strategy to assist the individuals to escape from the local optimum. The experimental results show that in comparison with similar fireworks algorithms,the proposed algorithm can more effectively improve the solution accuracy,and has faster convergence speed.