Optimization Algorithms for Energy-Efficient Train Operations in Real-Time Rail Traffic Management

This thesis addresses the problem of achieving energy-efficient train operations in real-time rail traffic management. Due to traffic perturbations, dispatchers have to re-schedule train operations repeatedly to maintain circulation as smooth as possible, therefore minimizing the total delay. Along with re-scheduling decisions, train speed profiles have to be determined in such a way that a feasible train schedule is preserved and the total energy consumption is minimized. First, we address a single-train real-time Energy-Efficient Train Control (EETC) problem, envisioned as a sub-problem of a hypothetical multiple-train Energy-Efficient Train Timetabling (EETT) problem for real-time traffic management applications. Precisely, we focus on determining an energy-optimal speed profile for a single train with a given schedule. We propose three algorithms: a constructive heuristic; a multi-start randomized constructive heuristic; and a Genetic Algorithm. We run experiments on real-life case studies provided by our industrial partner ALSTOM, which is a world leader in rail transport. Second, we address a real-time multiple-train EETT problem known as the real-time Energy Consumption Minimization Problem (rtECMP), which is a subproblem of the real-time Rail Traffic Management Problem (rtRTMP). The rtECMP asks for deciding the speed profiles of multiple trains circulating in a given network during a given time window. The objective is to minimize the weighted sum of total delay and energy consumption. Rail infrastructures are represented with a microscopic level of detail including the interlocking system and signals. We propose a graph-based rtECMP model and three meta-heuristic algorithms: Ant Colony Optimization (ACO); Iterated Local Search (ILS); and Greedy Randomized Adaptive Search Procedure (GRASP). We run experiments on two real-life case studies, considering mixed dense traffic subject to perturbations.