Because of the high-dense population and complex structure, the large public building faces a unique challenge in developing effective emergency evacuation plans. And due to the large scale and numbers of evacuees in real evacuation, real tests are impractical. Therefore, the simulation of evacuation becomes a wonderful choice in program planning. Particle Swarm is as one of the multi-agent based simulation method that can simulate complex behaviors of individuals. NSGA-II (Non-dominated Sorting Genetic Algorithm II) is a kind of optimization method for multi-objective optimization problem. In this paper, we propose a novel multi-objective evolutionary algorithm (named as PNMO, Particle swarm & NSGA-II based Multi-objective Optimization) which simulates evacuation process as well as optimizing the generated evacuation plans. The experiment shows that this method possesses superior performance in evacuation planning.