Abstract: Moth-flame optimization algorithm is a new swarm intelligence optimization algorithm. It has been applied to many fields such as feature selection and image segmentation. However, the traditional moth-flame optimization algorithm is prone to fall into local optimum, affecting the performance of the algorithm. For solving the deficiency, a moth-flame optimization algorithm based on refraction principle and opposite-based learning（ROBL-MFO） is proposed in this paper. Firstly, it uses the average of the best flame to improve the convergence speed. Then, the opposite-based learning is used to expand the search space of the ROBL-MFO. Finally, for jumping out of the local optimal, it uses refraction principle to improve the diversity of the population. Six test function is used to compare the ROBL-MFO with other algorithms, and the results show that the ROBL-MFO has better convergence speed and can effectively jump out of the local optimal.

Key words: moth-flame algorithm, refraction principle, opposite-based learning, swarm intelligence algorithm, population diversity

摘要： 飞蛾扑火算法是一种新型群智能优化算法，目前已经应用于特征选择和图像分割等诸多领域。然而，传统的飞蛾扑火算法后期收敛速度不足且容易陷入局部最优，从而影响了算法的整体性能。为了提高飞蛾扑火算法的优化性能，提出了一种基于折射原理反向学习的飞蛾扑火算法（ROBL-MFO）。该算法首先在飞蛾的更新公式中引入历史最优火焰平均值，使火焰间的信息能够互相交流，提高算法的收敛能力；其次利用随机反向学习策略对解进行反向学习，扩大算法的搜索空间；最后使用折射原理对解进行折射操作，提高种群的多样性，帮助算法跳出局部最优。在六个标准实验函数上得到的实验结果表明，对比其他算法，ROBL-MFO算法拥有更好的收敛速度，且能够有效跳出局部最优。

关键词: 飞蛾扑火算法, 折射原理, 反向学习, 群智能算法, 种群多样性