The real-time Energy Consumption Minimization Problem (rtECMP), as introduced by Montrone et al. (2018),has the objective of minimizing both train energy consumption and total delay by deciding speed profiles in a given control area and time horizon. It takes as input the decisions on train routing and precedences coming from a solver for the real-time Rail Traffic Management Problem (rtRTMP) Pellegrini et al. (2014). In addition, to define energy-efficient speed profiles for multiple interacting trains, it takes into accounts infrastructure characteristics, operational constants and train dynamics. The rtECMP ouputs include arrival, departure, passing through and dwelling times along with speed profiles. We extend the research of Montrone et al. (2018) by proposing a graph-based rtECMP model that we solve with three meta-heuristic algorithms. An experimental analysis is conducted on two French infrastructures : the Pierrefitte-Gonesse junction, and a section of the Paris-Le Havre line. We compare the meta-heuristic solutions against those of an exact method for the rtECMP.