Optimization in railway timetabling for regional and intercity trains in Zealand: A case of study of DSB
Tutor / director / evaluatorLyngby, Kongens
CovenanteeDanmarks tekniske universitet
Document typeMaster thesis
Rights accessOpen Access
The Train Timetabling Problem is one of the main tactical problems in the railway planning process. Depending on the size of the network, the problem can be hard to solve directly and alternative methods should be studied. In this thesis, the Train Timetabling Problem is formulated using a graph formu-lation that takes advantage of the symmetric timetabling strategy and assumed fixed running times between station. The problem is formulated for the morning rush hour period of the Regional and InterCity train network of Zealand. The solution method implemented is based on a Large Neighborhood Search model that iteratively applies a dive-and-cut-and-price procedure. An LP relax version of the problem is solved using Column Generation considering only a subset of columns and constraints. Each column corresponds to the train paths of a line that are found by shortest paths in the graphs. Then, violated constraints are added by separation and an heuristic process is applied to help finding integer solutions. Last, the passengers are routed on the network based on the found timetable and the passenger travel time calculated. The process is repeated taking into account the best transfers from the solution found. A parameter tuning is conducted to find the best algorithm setting. Then, the model is solved for different scenarios where the robustness and quality of the solution is analyzed. The model shows good performance in most of the scenarios being able to find good quality solutions relatively fast. The way the best transfers are considered between timetable solutions does not add significant value in terms of solution quality but could be useful from a planning perspective. In addition, most of the real-life conflicts are taken into account in the model but not all of them. As a result, the model can still be improved in order to provide completely conflict-free timetables. In general, the model appears to be useful for the timetabling planning process of DSB. It allows to test different network requirements and preferences easily. The model not only generates a timetable but also estimates the passenger travel time and the occupancy of the trains quite accurately. Also, any modification in the line plan can easily be included without affecting the core model.