Predictive Control For Automatic Operation Of High-speed Trains Based On Multi-point Model

ABSTRACT:Over the last decade, with the rapid development of high-speed rail and urban mass transit, the scale of high-speed railway network in China has reached the top level in the world. In this circumstance, Automatic Train Operation (ATO), which can improve the operation efficiency and performance of railway system, has become a research focus in the field of rail transportation.In this dissertation, several multi-point dynamic models of train are constructed to deal with the characteristics existed in the train dynamic model, such as the nonlinear running resistance, the uncertain disturbance caused by rail and ambient condition, and the uncertainty of model parameters. Then, considering the operation constraints of the train, namely the actuator saturation, the speed limitation for safe driving and the maxi-mum in-train coupling force, several controllers of train operation are designed based on model predictive control (MPC) correspondingly. Finally, an active fault-tolerant MPC (AF-MPC) controller is developed to tackle a variety of failure cases when the train actu-ator breaks down.The main contents and contributions of this dissertation are listed in the following.Firstly, a hybrid multi-point dynamics model of train is built by piecewise lineariz-ing the train’s running resistance and introducing an integer variable of train’s running status. An optimization problem is formulated based on the operation constraints of the train, as well as a cost function that penalizes the punctuality of trains, operation energy consumption and comfort of passengers. Subsequently, a MPC controller for train oper-ation is designed by solving the formulated optimization problem. Moreover, in order to reduce the computation complexity, two improved controllers are designed based on the input move-blocking technique and an explicit MPC respectively.Secondly, a multi-point dynamics model of train with bounded disturbances is estab-lished based on the model uncertainty. In order to avoid the unnecessary emergent stops caused by the train automatic protection, a robust MPC controller for train operation is designed by transforming the bounded disturbance from the model to constraints by using a constraints tightening algorithm, introducing a pseudo reference as a decision variable for the optimization problem, and adopting a modified cost function, an extended termi-nal constraint and a nominal train model. Finally, the theoretical analysis of feasibility and stability for the closed-loop system is presented.Thirdly, a multi-point dynamics model of train with unknown parameters and bounded disturbance is built to deal with the unknown and time-varying model param-eters. An adaptive parameter estimation law is designed to update the unknown model parameters, which enables the prediction of a monotonically decreasing worst-case es-timation error bound over the prediction horizon of the MPC. Then, an adaptive robust MPC controller for train operation is designed based on the nominal train model and the transforming constraints, as well as an additional comparison model for worst-case analysis based on a robust control Lyapunov function. Finally, the theoretical analysis of feasibility and stability for the closed-loop system is presented.Finally, the fault models of train are developed with considering several kinds of train actuator faults. Based on the known faults information, an AF-MPC controller for train operation is designed by reconstructing the predictive model or constraints of the fault-free MPC controller. Furthermore, to tackle the cases of complete and partial loss of actuators, a fault identification algorithm is designed based on the Moving Horizon Estimation method. The fault identification algorithm is integrated to the AF-MPC con-troller, which realizes an integral design of both fault identification and fault-tolerant control.