人工蜂群算法加速收敛技术研究

Research on Accelerating Convergence Technique of Artificial Bee Colony Algorithm

为提高蜂群算法的收敛速度及精度,提高其工程应用价值,探索了蜂群算法的加速收敛技术。通过分析自然界真实蜜蜂群间的信息共享模式,发现标准蜂群算法在适应度信息共享的处理上存在不足,导致该算法存在易陷入局部最优及收敛速度慢的缺点。文中在标准算法的基础上,修改了适应度共享机制,使得一定邻域内的多个采蜜蜂的搜索信息均可被观察蜂共享,在观察蜂的搜索中引入欧式距离以确定有效邻域,选择邻域内的最优解用以生成新蜜源。通过测试发现改进后的算法收敛速度明显提高,提高幅度高达50%。

Bee colony accelerating convergence technique is studied in order to improve the convergence speed and accuracy, and engi- neering application value of the artificial bee colony algorithm. It is discovered that imperfection is existed in standard ABC algorithm when coping with the share of the information which is assembled by employed bees. This flaw leads to the deterioration of the algorithm. In this paper,the share strategy is modified with the aim to make use of the data collected by multiple employed bees which are in certain neighborhood. Euclidean distance is introduced to identify the valid neighborhood, and the best solution in the region is selected to pro- duce new nectar. Numerical experiment indicates that the convergence speed of the modified algorithm is improved,about 50% better than the original one.