Multi-objective optimization of train speed profiles on the Ayat-Megenagna Line
In this paper, a new approach has been developed for train speed profile optimization, discrete space based modeling followed by the determination of an optimal set of riding modes using multi-objective optimization techniques. The optimization problem is formulated by making energy and time as the components of the two element objective vector function. A point mass model of the operation of trains is developed by considering all the important force components acting on the train. The distance to travel between stations is discretized into 20 equal length elements where a two stage solution procedure has been applied to get to the final results. The first stage of the solution procedure is the application of a Non dominated Sorting Genetic Algorithm II (NSGA II) based optimization technique taking vector of riding modes as the decision variable. Using the developed algorithms for the calculation of cost functions, a Pareto-optimal set of riding modes are determined. The second stage of the solution process smoothes out the results found in the previous stage without bringing about considerable change in the values of the cost functions. Various speed profiles are generated as optimal for the case of Ayat to Megenagna line of Addis Ababa Light Rail Transit (AALRT). The speed profiles that are generated as the fastest can bring about up to 30% reduction in headway over the plan. Furthermore, by choosing the slowest trajectories over the fastest ones, it is possible to save up to 38.18% of energy, while 23.98% of reduction in riding time can be achieved by preferring the fastest profiles over the slowest ones.
Keywords: Speed profile, energy consumption, running time, multi-objective, optimization, AALRT