In view of the problem of insufficient accuracy of the traditional model of the full parallel autotransformer traction power supply system in modeling and power flow calculation, this paper proposes a high-fidelity modeling method that combines measured data with model correction. The particle swarm optimization (PSO) is used to identify and correct the model parameters to solve the problem of inaccurate simulation results caused by model parameter deviation and unavoidable simplification. At the same time, in order to overcome the convergence problem that may occur when PSO is directly used for overall parameter correction, a two-layer model correction method is designed: subsystem parameter identification and overall model correction. The simulation results show that the constructed high-fidelity model can effectively reduce the error with the actual system and maintain a high simulation accuracy under the conditions of different locomotive numbers, operating conditions and position changes. This method has great reference value for the operation optimization, scheduling and system design of the traction power supply system.