for a metro railway network. The timetable schedules the arrival time and the departure time of each train to and from the platforms it visits such that the total electrical energy consumed is minimized and the utilization of produced regenerative energy is maximized.
1.2. Related work
The general timetabling problem in a metro railway network has been studied extensively over the past three decades ( Harrod, 2012 ). However, very few results exist that can calculate energy-efficient timetables. Now we discuss the related research. We classify the related work as follows. The first two papers are mixed integer programming model, the next three are models based on meta-heuristics and the last one is an analytical study. A Mixed Integer Programming (MIP) model, applicable only to single train-lines, is proposed by Peña-Alcaraz et al. (2012a ) to maximize the total duration of all possible synchronization processes between all possible train pairs. The model is then applied successfully to line three of the Madrid underground system. However, the model can have some drawbacks. First, considering all train pairs in the objective will result in a computationally intractable problem even for a moderate sized railway network. Second, for a train pair in which the associated trains are far apart from each other, most, if not all, of the regenerative energy will be lost due to the transmission loss of the overhead contact line. Finally, the model assumes that the durations of braking and accelerating phases stay the same with varying trip times, which is not the case in reality. The work in Das Gupta et al. (2015) proposes a more tractable MIP model, applicable to any railway network, by consider- ing only train pairs suitable for regenerative energy transfer. The optimization model is applied numerically to the Dockland Light Railway and shows a significant increase in the total duration of the synchronization process. Although such increase, in principle, may increase the total savings in regenerative energy, the actual energy saving is not directly addressed. Simi- lar to Peña-Alcaraz et al. (2012a ), this model too, assumes that even if the trip time changes, the duration of the associated braking and accelerating stay the same.