Reducing Power Peaks in Railway Traffic Flow Subject to Random Effects

Railway traffic flow in a corridor can be modeled by a string of consecutive trains, each subject to random speed variations that are described by a stochastic process. Despite analogies with car-follower models, railways include specific features and a safety system that forces vehicles to decelerate towards a fixed lower speed if an absolute safety distance with the vehicle ahead is not respected. We simulate such a dynamic system under assumptions that model human drivers and automated train operations (ATO), and compute performance measures focusing on energy consumption and the power peaks arising when multiple trains accelerate simultaneously. We investigate measures to smooth these peaks including the use of regenerative braking energy, potentially coupled with an electric energy storage, and a rule that uses fixed waiting times before re-accelerating. Our findings shed light on when and why these measures can be effective at reducing energy consumption and/or shaving the peaks, and show that employing a well-calibrated ATO controller improves energy performance compared to a model of a human driven. Our results also expose a trade-off between the energy performance and the regularity of the traffic, i.e., strategies to reduce power peaks may slow rail traffic down, leading to a lower capacity utilization.