Abstract: The moth-flame algorithm is a group intelligence algorithm with a simple structure , few configuration parameters and a wide range of applicability. However, the convergence accuracy, convergence speed needs to be improved, and the algorithm still exists easy to converge to the local optimum, this study proposes a chaotic moth algorithm based on the squirrel-search strategy. A sinusoidal map is employed to initialize the population to obtain high-quality initial populations. The squirrel search path is simulated in the moth searching process, and high-quality and near-flame individuals are introduced to guide the moths to search for high-quality individuals.  The predator probability triggered by the cosine control factor is employed to prompt the moth to break free from the conventional flame’s attraction, so as to improve the global searching ability of the moth algorithm. The adaptive t-distribution and the flame adaptive reduction formula are modified to control random migration via Lévy flights is utilized to increase the local search capability and convergence speed of the proposed algorithm. The proposed algorithm is validated experimentally and compared to fifteen intelligent algorithms using CEC2017 test functions, CEC2022 test sets and two practical engineering problems. The experimental results demonstrate the proposed algorithm exhibits significant advantages in terms of convergence speed, searching capability, and escaping from local optima.

Key words: moth-flame optimization algorithm, squirrel-search algorithm, adaptive control factor, Lévy flight

摘要： 飞蛾算法是一种结构简单、配置参数少且适用范围广的群智能算法，但在收敛精度和收敛速度等方面还有待提高，且存在易收敛到局部最优的问题，为此提出一种融合松鼠搜索策略的混沌飞蛾算法。该策略采用sinusoidal混沌映射获取高质量初始种群；在飞蛾寻优过程中引入松鼠算法中松鼠的寻优途径，设置高质量火焰个体与近距离火焰个体指导飞蛾高质量寻优，通过余弦控制因子触发的捕食者概率促使飞蛾跳出原始火焰对其的吸引，提高飞蛾算法全局搜索能力；改造自适应[t]分布因子与火焰自适应减少公式，控制适应度较差的种群通过列维飞行进行随机迁移，增加算法的局部搜索能力。通过CEC2017测试集、CEC2022测试集与两个工程应用实例分别与其他15种智能算法进行对比验证，结果表明改进算法在收敛速度、搜索能力和跳出局部最优等方面具有一定优势。

关键词: 飞蛾优化算法, 松鼠优化算法, 自适应控制因子, 列维飞行