基于动态多种群的多目标粒子群算法

摘要

研究进化算法在求解多目标优化问题时,极易陷入到伪Pareto前沿(等价于单目标优化问题中的局部最优解),为了提高优化过程,提出一种基于动态多种群的多目标粒子群算法(DMSMOPSO).在DMSMOPSO算法中,为了增加种群的多样性,提升粒子跳出局部最优解的能力,采用多子群进行搜索并且子群是动态地进行构建;采用K-均值聚类算法确定每个子群的搜索行为,提升种群向全局最优位置飞行的概率;根据目标函数的优化难度.通过典型的多目标测试函数和工程上的实际应用对算法进行仿真,仿真结果表明DMSMOPSO比其它算法优越,证明DMSMOPSO可作为求解多目标优化问题的有效算法.%In view of that multi-objective evolution algorithm easily converges to a false Pareto front, which is the equivalent of a local optimum in single objective optimization, a multi-objective particle swarm optimizer based on dynamic multi-swarm (DMSMOPSO for short) is discussed. In the DMSMOPSO algorithm, to increase the diversity of the swarm and improve the ability to escape from the local optima, the multi-swarm strategy is adopted to search for the feasible space. Also, the K-means clustering is used to confirm the center particle of each sub-swarm to guide the swarm flight, which will improve the probability of flying to Pareto front for the whole swarm. At last, in terms of the difficulty of different objective function, each sub-swarm assign different optimization task. Experiments were conducted on a set of classical benchmark functions and engineering application. Simulation results show that the proposed algorithm has better performance compared with other algorithms.