Timetable rescheduling of metro network during the last train period: Fid-Bau Portal

Timetable rescheduling of metro network during the last train period

Wang, Yonggang / Chen, Junxian / Qin, Yang / Yang, Xiaofang

Highlights • The period from the first last train departure to network service ending is studied. • An optimization model is proposed to reschedule the train timetable of the network. • Total transfer waiting time and robustness of the optimization scheme is considered. • An improved quantum-behaved particle swarm optimization (QPSO) algorithm is designed. • Shenzhen metro network is used to verify model’s effectiveness and practicality.

Abstract Non-last train passengers in the last train period experience issues with transferability and accessibility in addition to the final service of each line. Therefore, a specified schedule needs to be determined by operators with full consideration of passengers' travel demand for all the trains during this period. Aiming at minimizing the average transfer waiting time, failure transfer rate, and the negative effect of random passenger flow without incurring excessive costs to the operator, a model optimizing the train connection in the last train period is formulated under constraints of operation parameters and transfer passenger flow conditions. To depict the uncertain features of transfer demand, a random passenger flow generation method with stochastic simulation is incorporated into an improved quantum particle swarm algorithm (QPSA). Taking Shenzhen Metro network in 2019 as an example, the rationality of the optimized model, the feasibility of the quantum-behaved particle swarm optimization (QPSO) model, and its superiority over the conventional particle swarm optimization (PSO) model in solving this problem are successfully verified. The results show an 18.54% decrease in the average waiting time for transfer passengers with the optimized last train connections, along with an increase of 8.31% and 6.54% in the number of successful transfer passengers and directions, respectively, after optimization. The proposed QPSO model can reach convergence faster and without falling into local optima. The results also show that the average waiting time of transfer passengers can be significantly decreased after fine-tuning train operation parameters. A reasonable combination of objective weight factors ( $λ = 10$ , $μ = 100$ ) can reflect the game characteristics between passenger’s and operator’s preferences, while extending the service closure time within 20 min can further improve the overall service quality of last train period across the network.