采用分等级学习策略的二进制粒子群优化算法

Binary Particle Swarm Optimization Algorithm Using Hierarchical Learning Strategy

针对二进制粒子群优化算法在寻优后期存在多样性丢失、收敛精度低等问题,提出一种分等级学习策略的二进制粒子群优化算法(HLBPSO)。首先,HLBPSO算法借鉴鸡群优化算法中的等级思想,根据适应度值将粒子种群分为优势、中间和劣势三个等级,并依次采用探索、全局和混合学习策略;其次,对于劣势粒子,使其在向优势等级最优粒子与中间等级最优粒子的差分向量进行学习的同时,设计逃逸算子,赋予劣势粒子以一定的概率逃逸;最后,通过计算粒子与全局最优粒子间的距离实现惯性权重更新。实验结果验证了HLBPSO算法比其他算法具有更高的寻优精度和更好的鲁棒性。

Aiming at the problem that the binary particle swarm optimization algorithm has diversity loss and low convergence precision in the late stage of optimization,a binary particle swarm optimization algorithm with hierarchical learning strategy(HLBPSO)is proposed.Firstly,the HLBPSO algorithm draws on the hierarchical idea in the chicken swarm optimization algorithm,divides the particle population into three levels of superiority,intermediate and inferior according to the fitness value,and adopts the exploration,global and mixed learning strategies in turn.Secondly,for the inferior particles,while learning the difference vector between the dominant level optimal particle and the intermediate level optimal particle,the escape operator is designed to give the inferior particle escape with a certain probability.Finally,inertial weight is updated by calculating the distance between particles and global optimum.The experimental results verify that the HLBPSO algorithm has higher precision and better robustness than other algorithms.