Abstract: In order to achieve real-time prediction of train longitudinal force that balances calculation accuracy and speed, the paper establishes a real-time and accurate calculation method for heavy haul train longitudinal force based on the vehicle-carried data and test data. The locomotive force of traction/electric-braking is updated through vehicle-carried monitoring data, and the basic running resistance coefficient of the train is adaptively adjusted accordingly. Furthermore, the empirical model and fluid model of the air braking system are compared, and a high-precision balance iteration method is established to solve the longitudinal dynamic equation of the train; To further improve computational efficiency, a simplified calculation method for longitudinal force of trains using Duhamel integral to solve vibration response is established; Taking the example of heavy haul train passing through the typical long heavy down grade from -9 ‰ to -12 ‰, the precise method has a maximum error of 3.72 km/h in calculating the train's operating speed, and has a good predictive effect on the relative displacement trend, magnitude level, and key impact of the buffer. The simplified method can achieve a probability distribution of coupler force error similar to the precise method in some sections where the train's operating conditions are relatively stable, that is, about 68% of the probability error is less than 64.6 kN, 95% of the probability is less than 129.2 kN, and 99.7% of the probability is less than 193.8 kN; However, the simplification method is too simplistic in terms of buffer characteristics, resulting in poor simulation of short-term severe longitudinal impacts during the transition of train operation conditions.