Abstract: Compared with stationary train wind tunnel test, moving train model test can realize more accurate identification of train aerodynamic characteristics. However, some train model parameters (such as model length) cannot fully reflect the real operation scenario due to some limitations of test equipment, site and other conditions. In order to study the effect of train model length on the aerodynamic characteristics of trains passing through bridges-tunnel junctions under cross wind for selection of train length in wind tunnel test, the LES model and dynamic mesh are adopted in this paper to establish the CFD simulation model of moving trains passing through the bridge-tunnel junction under cross wind. The simulation results are verified by the moving train model test. The aerodynamic performances of the trains with different lengths are analyzed, and the flow field around the vehicle is also examined. In some restricted test conditions, the one-car plus half-car length compensation section model can replace the three-car train model. The results show that the similarity between the aerodynamic coefficient of the head car of single-car train model and that of the three-car train model can be effectively improved by adding a half-car length compensation section. In addition to the large error in drag force coefficient, the error of the average aerodynamic force coefficients between the one-car model and the three-car model running on the bridge section is 1.7 to 6.3 times that of the results after adding the compensating section, and the error of the results when the train running in the tunnel can be reduced by more than 47% by adding a compensating section. The same trend is observed for the train pressure coefficient. The difference is that for the three-car formation model and the model with additional compensation section, the pressure coefficient during the train's entry into the tunnel can be maintained near 0 for a certain distance in the tunnel, and its distance increases with the train length. However, the decreasing amplitude of the pressure coefficient of the three models during entering tunnel portal is similar; after subtracting the stable section, the decreasing rates are basically the same, and the change patterns at the exit are also similar. Therefore, under limited test conditions, the addition of half compensation section in the single-car train model can replace the simulated three-car formation.