Energy-Efficient Rescheduling of Metro Train Timetable under Stochastic Delay Conditions via Deep Reinforcement Learning Method

Metro stations with high ridership are prone to operational delays due to unforeseen factors, disrupting passenger travel and increasing energy consumption. Traditional methods often prioritize restoring the timetable without considering the impact on energy consumption. Considering this deficiency, a timetable rescheduling model based on deep reinforcement learning is proposed for the effective resolution of the aforementioned problem in this paper. This model takes into account both traction energy consumption and regenerative energy consumption. After a train delay occurs, it reallocates the train operation time for subsequent sections, reducing energy consumption while eliminating delay time. A case study of Nanning Rail Transit Line 5 is used to validate the model, showing a 2.1% reduction in energy consumption compared to traditional methods. Moreover, the comparison with other algorithm proves the effectiveness and robustness of the method. The proposed approach demonstrates significant improvements in both delay management and energy efficiency and can be applied to real-time train scheduling in various metro systems to optimize performance and sustainability.