Abstract: To keep the diversity of particles when multi-objective particle swarm optimization is running, a multi-objective optimization algorithm was proposed based on particle swarms initialization and dynamic multiple particle swarms cooperation. The quantity of swarms was increased or decreased dynamically according to the distribution of particle swarms in the decision space. To avoid converging too quickly, the factors, which affected the flying speed of a particle, were improved to depend on the current velocity inertia of the particle, the best value of the particle, the best value of the swarm which the particle belonged to, and the optimal value of all swarms. This algorithm was tested by five benchmark functions and compared with the multi-objective particle swarm optimization. The experimental results indicate that the proposed algorithm is superior to the multi-objective particle swarm optimization.

Key words: Multi-Objective Optimization (MOO), Particle Swarm Optimization (PSO), local search, global optimal solution, local optimal solution

摘要： 为了让多目标粒子群优化算法在运行过程中保持粒子的多样性，提出了一种初始化方法和动态多粒子群协作的多目标优化算法。根据粒子群在决策空间中的分布情况动态增加或者减少粒子群数量；为避免粒子收敛速度过快，改进了决定粒子飞行速度的因素，速度值依赖于粒子当前速度惯性、粒子最优值，群最优值和所有群最优值。用五个测试函数对算法进行了测试并与多目标粒子群优化进行了比较，测试结果表明提出的算法优于多目标粒子群优化算法。

关键词: 多目标优化, 粒子群优化, 局部搜索, 全局最优解, 局部最优解