Abstract: Aiming at the slow convergence rate and low optimization accuracy of flower pollination algorithm, a new Beetle Antenna Search Flower Pollination Algorithm（BASFPA） is proposed. Firstly, the algorithm uses the beetle antennae search in global search to improve the convergence rate of algorithm, then it uses perturbation evolution in local search to help the algorithm jump out of local optimal. Six commonly used optimization functions are used in experiment to test BASFPA, the result shows that the BASFPA algorithm has a significant improvement of convergence speed and optimization accuracy in low-dimensional and high-dimensional optimization, and the number of iterations required to achieve the same accuracy is less than other algorithms, it is proved that the improvement of this algorithm is reasonable.

Key words: flower pollination algorithm, beetle antennae search, mutation strategy

摘要： 针对花朵授粉算法后期收敛速度慢，寻优精度低的缺点，提出了一种基于天牛须搜索的花朵授粉算法（BASFPA）。算法首先在全局寻优阶段采用天牛须搜索加快收敛，其次在局部寻优阶段加入变异策略帮助算法跳出局部最优。实验使用6个常用优化函数进行测试，结果表明BASFPA在低维和高维下收敛速度和精度均高于其他算法，达到相同精度所需的迭代次数均小于其他算法，证明天牛须搜索对FPA算法的改进是合理的。

关键词: 花朵授粉算法, 天牛须搜索, 变异策略