With an aim to reduce the energy consumption in rail transit system, an optimized bi-level two-phase method has thus been put forward to achieve an energy saving operation for metro trains based on Particle Swarm Optimization. A new model for train operation adjustment has been established based on an analysis of the characteristics of train operation organization. In the first phase, Particle Swarm Optimization (PSO) has been used to comprehensively search the energy-saving operation curves of metro trains, thus obtaining the optimum operation curve under the premise of accordant running time constraints. In the second phase, the energy consumption-time curve for each interval is to be generated based on the optimization results produced in the first phase so as to further optimize the train schedule. The model has been simulated by using the data of a Metro Line 2. Experimental results show that the improved algorithm outperforms the traditional algorithm in that it presents a better effect on energy-saving operation and the formulation and optimization of time table in actual train lines.