Abstract: multi-standard heavy-haul railway networks，this paper proposes an intelligent scheduling system based on federated learning. A hierarchical federated learning framework is designed to enable collaborative training and dynamic aggregation of multi-standard operational data，such as CTCS-3/4 and ETCS-2. By integrating 5G-R communication and edge computing technologies，the efficiency of train-ground data transmission and the capability of distributed decision-making are enhanced. The system incorporates a dynamic label generation mechanism and a gradient sparsification strategy，which significantly reduces communication overhead while maintaining the real-time performance of scheduling commands. Simulation results demonstrate that the proposed system can effectively improve the compliance rate of train tracking intervals in multi-standard scenarios，reduce protocol switching delays，and thus provide a viable solution for the intelligent upgrading of complex railway networks.