Solving multi-depot electric vehicle scheduling problem by column generation and genetic algorithm

Electric vehicles are increasingly applied in bus enterprises due to their advantages of environmental protection and low cost. Although there exist many studies on fuel vehicle scheduling, there still lack studies on electric vehicle scheduling in public transit. In this paper, we study a multi-depot electric vehicle scheduling problem (MD-EVSP) in public transit and propose a genetic algorithm based column generation approach (GA-CG) for MD-EVSP. A column refers to the route of a vehicle. First, a heuristic is devised to create a set of initial columns. Then, starting from the initial columns, the column generation approach with a label correcting algorithm is used to generate a set of candidate columns. Finally, a genetic algorithm is devised to select a subset of columns from the column set to construct the final solution. A solution coding scheme is designed to represent a subset of columns, that is, a candidate solution to MD-EVSP. An elite strategy is integrated into the genetic algorithm to improve the global search capability. GA-CG is applied to a real-world problem with three bus lines in Qingdao, China. Experiments show that compared with the branch and price (BP) algorithm and the manual scheduling scheme, GA-CG can effectively solve the problem and its computational time is about 40 times shorter than the BP algorithm. (C) 2021 Published by Elsevier B.V.

Automated summary

This study focuses on the multi-depot electric vehicle scheduling problem in public transit and proposes a genetic algorithm based column generation approach. Experimental results show that this approach can effectively solve the problem with shorter computational time compared to other algorithms.