Adaptive PID control algorithm for urban rail train with a partially updatable weights neural network

Abstract

To meet the requirements of precise speed profile tracking of urban rail transit system in complex environments and to reduce energy consumption, a back propagation (BP) neural network control strategy with partially updatable weights is proposed, it not only achieves better control performance but is also easy to deploy in large-scale industrial applications. Firstly, a train dynamic model is established, and the additional resistance encountered during train operation is considered to simulate complex operating conditions. Then, the BP neural network architecture is determined, the rules of each structural layer and the definition of weight parameters are included, and the algorithm’s adaptability to changes in model parameters and external disturbances is ensured. Finally, a simulation model is established for comparative studies, and the motor speed control experiment is conducted as an initial exploration of real train operation control. Results indicate that the BP neural network controller with partially adjustable weights exhibits high speed tracking accuracy and robustness.