Propagation of train delays in stations

Abstract

Train delays occur on a daily basis, and their propagation is a main source for delays experienced by the Dutch Railways. A statistical analysis shows that the mean train delay increases 15-45 seconds for most train lines at The Hague HS station, the Netherlands. Late arrival delays and corresponding dwell delays can generally be fitted to an exponential distribution and a normal distribution, respectively. In this paper, a stochastic model is presented that forecasts the distribution of late arrival and dwell delays. The model incorporates some main constraints of train operations, and is solved by the Monte Carlo sampling approach. Validation shows that the model predicts the propagation of train delays reasonably well in railway stations.