In the past few decades, multi-objective particle swarm optimization (PSO) has increasingly attracted attention from scientists. To obtain a set of more accurate and well-distributed solutions, many variations of multi-objective PSO algorithms have been proposed. However, for complicated multi-objective problems, existing multi-objective PSO algorithms are prone to falling into local optima because of their weak global search capability. In this study, a modified multi-objective particle swarm optimization algorithm based on levy flight and double-archive mechanism (MOPSO-LFDA) is proposed to alleviate this problem. On one hand, in the evolution process of the particles, levy flight is combined with PSO to avoid the algorithm falling into local optima. By expanding the search scope of the particles, levy flight can improve the global search ability of the particles and make them jump out of local optima with a high probability. On the other hand, when maintaining external archives, in addition to the primary external archive, a secondary external archive is created to avoid unnecessary removal of the particles that may be generated by traditional maintenance approaches. With the proposed double-archive mechanism, more useful particles can be kept, and thus the diversity of the solutions is increased. Moreover, in terms of accelerating the convergence rate, a novel leader selection strategy is presented, which selects particles closer to the boundary of the attainable objective set and with larger crowding distance as leaders in optimization. The proposed algorithm outperforms existing state-of-the-art multi-objective algorithms on benchmark test functions for its fast convergence and excellent accuracy.