This work aims to optimize the driver control along a track to ensure minimal energy consumption. The focus is on optimizing a control system, which requires a dynamic model to feed an energy model. This will enable the linking of control and consumption while checking operational constraints such as punctuality and safety that apply to the dynamics of the train. In both models it is necessary to determine parameters that are not directly measurable and potentially variable from one trip to another (such as the mass of the passengers). As we have both expert knowledge and real measurements, this work focuses on Bayesian calibration to deduce an a posterior distribution; from this distribution, we will extract the maximum from this a posteriori distribution in order to perform deterministic optimization. The conclusion of this work is that energy can be reduced. However, the robustness of the model is not sufficient, since a small variation of variable parameters (passenger mass or wind) could cause the operational constraints to be violated.