融合多策略的增强海鸥优化算法

Enhancing seagull optimization algorithm by applying multiple strategies

针对海鸥优化算法(SOA)求解精度低、种群多样性差、易陷入早熟收敛的缺点,提出了一种融合多策略的海鸥优化算法(ESOA)。首先,在每次迭代的过程中,引入改进的自适应差分变异策略,对单个海鸥个体进行差分变异操作并通过自适应机制扩大海鸥的全局搜索范围及提高种群的多样性;其次,设置了基于粒子群算法的机制来处理最差的海鸥个体位置;最后,针对海鸥的最优位置,采用了动态透镜映射的策略增加算法跳出局部最优的能力。采用CEC2017测试函数中的14个函数作为基准测试函数,将ESOA与麻雀算法(SSA)、飞蛾扑火算法(MFO)、灰狼算法(GWO),以及改进的GSCSOA、CCSOA进行性能对比。实验结果表明ESOA在统计学意义上具有显著的性能优势。

Aiming at the shortcomings of the seagull optimization algorithm(SOA), which has low solution accuracy, and rela-tively poor population diversity, and is easy to trap in local optimal, this paper proposed an enhanced seagull optimization algorithm(ESOA) that integrated multiple improvement strategies. Firstly, this paper used an improved self-adaptive differential mutation strategy to perform differential mutation operation on a single seagull individual in each iteration process which could expand the global search range of seagulls and improve the diversity of the population.Secondly, it established a processing mechanism for the worst position of individual seagulls based on particle swarm algorithm.Finally, for the optimal position of the seagull, it adopted a dynamic lens mapping strategy to jump out of the local optimum for the optimal position of the seagull. This paper used 14 functions in the CEC2017 test suite as the benchmark function to compare the performance of ESOA, algorithm with sparrow search algorithm(SSA), moth-flame optimization algorithm(MFO), grey wolf optimizer(GWO), and the improved GSCSOA and CCSOA. The experimental results show that ESOA has a statistically significant performance advantage.