基于比例分配的多目标粒子群改进研究

Research on Improvement of Multi-Objective Particle Swarm Based on Proportional Distribution

针对粒子群算法存在收敛性差、容易陷入局部最优等特点,提出了一种基于比例分配的多目标粒子群改进研究(PDMOPSO)。利用比例分配机制,使个体有针对性地选取全局最优进行有效学习;同时采用网格划分和余弦距离策略动态维护和更新外部存档,有利于非支配解集靠近真实的Pareto前沿;最后,分别使用高引用经典改进的粒子群算法在9个典型测试函数上进行仿真实验。实验结果表明,PDMOPSO算法是一种非常具有竞争力的算法,它在大多数测试函数上优于比较算法的性能。

In order to solve the problemthat particle swarmoptimization is easy to be trapped in local optimum,low convergence precision,a multi-objective particle swarm improvement research based on proportional distribution(PDMOPSO)is proposed.The proportional distribution mechanism was used to enable individuals to select the global optimum for effective learning in a targeted manner.At the same time,the grid division and cosine distance strategy are used to dynamically maintain and update the external archive,which is beneficial to the non-dominated solution set close to the real Pareto front.Finally,simulation experiments were carried out on 9 typical test functions using the classic improved particle swarmalgorithmthat was highly cited.Experimental results showthat the PDMOPSO algorithmis a very competitive algorithm,and it outperforms the comparison algorithmin most test functions.