Development of optimal real-time metro operation strategy minimizing total passenger travel time and train energy consumption

Abstract

The optimization of the total passenger travel time and total train energy consumption are critical factors in metro operation optimization. However, deriving an optimal train operation plan that incorporates both passenger travel time and total train energy consumption is a complex task because it should consider numerous variables representing the operational status of the urban railway, such as the number of boarding and alighting passengers, number of on-board passengers in each train, and entire train operation status along the line. Moreover, owing to the fluctuating nature of passenger demand, which can change rapidly over time, its optimization becomes challenging. To address this challenge, this study develops a recurrent neural network-based real-time metro operation optimization model trained using data representing the moments when the trains departed from the stations. These data are derived and reconstructed from various simulated operation plans while searching for optimal daily metro timetable. Consequently, the proposed model derives the real-time optimal operation strategies for trains departing from the next station within an average of 0.18 s. The result of metro operation simulations using proposed optimal operation strategies reveals a 7–14% improvement in efficiency compared to the current train operation strategies.