Digital Twin-based Automatic Train Regulation for Integration of Dispatching and Control

This paper proposes a framework of digital twin-based integration of train regulation and control for metro systems. We study the integration of train regulation and train control by using deep learning and hybrid search methods. Specifically, a well-trained convolutional neural network is adopted to fit the nonlinear function mapping relationship between the input set (the line conditions and running time of the train in each section) and the output set (the energy consumption of the optimal speed profile and the switching points), which greatly reduces the calculation time of the optimal recommended speed profile in the process of train regulation. An integrated optimization model is proposed with the aim to minimize the total train delay and energy consumption, a hybrid search algorithm based on particle swarm optimization and a genetic algorithm was developed to solve the problem. The effectiveness of the proposed method is verified based on the actual operating data of a Beijing subway line.