Validating Train Dwell Time Estimation Model without Considerations of Passengers

摘要

Train dwell time estimation is a critical issue in both scheduling and rescheduling phases. The author proposed a novel dwell time estimation model without passenger data which most of the existing models heavily rely on. It was applied in the Dutch railways and shows promising results. This paper validates the model by applying more datasets from Beijing urban rail transit systems, comparing the estimation accuracy with the Culemborg station in the Netherlands. The validation results show that it fits for 88.6% of the investigated actual dwell time, which is slightly better than the result for the Dutch railway stations (80.1%-88.5%). The mean error is 7.21 s, which is relatively small. The model can be used to estimate the train dwell time in all investigated stations both in peak and off-peak periods. It concludes that the dwell time estimation model is generic within the current datasets.