Hybrid model for proactive dispatching of railway operation under the consideration of random disturbances in dynamic circumstances

Abstract

With the increasing traffic demand and limited infrastructure expansion, railway networks are often operated close to the full capacity, especially in heavily used areas. As a result, the basic timetable is quite susceptible to the operational disturbances, and thereby the propagation and accumulation of delays significantly degrade the service level for customers. To solve this problem, extensive researches have been conducted by focusing on the predefined robust timetables and the real time dispatching algorithm development. However, it has been widely recognized that excessive robust timetables may deteriorate the operating capacity of the railway network and the addition of recovery time and buffer time can be hardly implemented in the congested area. Moreover, most of the conventional dispatching algorithms ignore the further potential random disturbances during the dispatching process, which yield non-implementable dispatching solutions and, as consequences, inferior punctuality and repetitive dispatching actions. To this end, this project aims to develop a new algorithm for real-world dispatching process with the consideration of risk-oriented random disturbances in dynamic circumstances. In the procedure of this project, an operational risk map will be firstly produced: by simulating considerable amount of disturbed timetables with random disturbances generated in a Monte-Carlo scheme and calculating the corresponding expected negative impacts (average total weighted waiting time among all the disturbances scenarios), different levels of operational risk will be assigned to each block section in the studied railway network. Within a rolling time horizon framework, conflicts are detected with the inclusion of risk-oriented random disturbances in each block section, and the near-optimal dispatching solutions are calculated by using Tabu search algorithm. Finally, three indicators including total weighted waiting time, the number of relative reordering and average absolute retiming, are chosen to testify the effectiveness and advantages of the proposed dispatching algorithm. The sensitivity analysis of dispatching-related parameters is conducted afterwards to investigate the most appropriate relevant parameters for the specific studied area. The proposed algorithms are expected to be capable of automatically producing near-optimal and robust dispatching solutions with sufficient punctuality achieved.