The full length of heavy haul train is several kilometers, and its operation process is a complex dynamic system. The key technology of automatic heavy haul train driving is tracking the running speed profile. Based on the intelligent wagon train scheme of a heavy haul train, this paper establishes a multi-agent model of the heavy haul train by analyzing the dynamic process of heavy haul train operation. Considering the unknown interference of the outside world to the train while the train is running, and ensuring the safe distance between carriages, a composite consistency controller for the heavy haul train is proposed. A consistency algorithm is constructed by using the speed displacement information of adjacent carriage units to make the system speed consistency converge, and then the sliding mode control term is introduced to accelerate the convergence speed. The different disturbances received by different carriages of the train are regarded as unknown disturbances, and the system buffet will be large and the robustness will be weakened with the increase of sliding mode gain. Therefore, an observer is designed to estimate the disturbance and compensate it to the controller to ensure the convergence of the system and improve the anti-interference ability. The artificial potential function is introduced to ensure that the distance between adjacent car units is in a safe range and reduce the longitudinal impulse. The Matlab software is used for simulation experiments to track the given speed curve, and various disturbances are used to simulate the influence of unknown factors on the train, which is compared with the effect of the controller without the observer. The simulation results show that the composite consistency controller can better track the set speed curve, the speed error is within $\pm0.4{km