| Virtual Coupling(VC)can further shorten the operation interval between trains and increase railway capacity,which is an important development direction of train transportation organization in the future.However,there is no complete theoretical system to support the development of its train operation control system technology for VC.Under VC,the VC-oriented train operation control system can be divided into two types: the centralized controller and the distributed controller,according to the communication topology and the distribution of control subjects.The current communication technology can support the realization of these two technologies.Different from that of the distributed controller,it is convenient for the main body of the centralized controller in the virtually coupled train set(VCTS)to coordinate the overall information,enhance the cooperative behavior between trains and implement efficient control of the entire VCTS.The control for VCTS has clear requirements and targets,and also includes complex constraints.For this type of control problem,Model Predictive Control(MPC)is a commonly used control architecture.This paper proposed a centralized control method for virtual marshalling trains based on MPC,and provided solutions to several problems in its design,optimization and application.The content is as follows:(1)A general centralized MPC framework for the VCTS is designed.Based on the traditional MPC control framework for the self-driving vehicle platoon,this paper proposed a controller model suitable for VCTS and a solution algorithm on MPC.The stability of the controller is analyzed.Numerical simulation experiments show that this proposed method can achieve globally stable control of VCTS within a certain range of prediction horizon.(2)The the real-time performance of the proposed general centralized MPC framework for the VCTS is optimized.Aiming at the real-time defect in the proposed controller,based on a deployable model predictive control and the first-order approximation(DMPC-FOA),this paper adjusted the calculation timing in the control cycle process,conducted the time-consuming optimal control calculation process within the reserved time period,and used the first-order approximation to compensate for the defect caused by the distortion of the optimal control output for the too early prediction of state value at the sampling time.Numerical simulation experiments show that this method can greatly restore the optimal control solution under the ideal state,and the real-time performance is guaranteed.(3)The control method facing the VCTS under heterogeneous environment is designed.This paper took the gradient terrains as a heterogeneous environmental factor,proposed a construction method for a nonlinear optimal controller considering the effect of the gradient terrains with the help of the law of conservation of energy and used the continuation/generalized minimum residual method(C/GMRES)to solve it,with the help of which,the complexity of the calculation for the control output is greatly reduced and the real-time performance is improved only with a small loss of accuracy.Numerical simulations show that the proposed controller can still realize the globally stable control of VCTS under the condition of changing gradient.In addition,compared with traditional solutions based on boundary value problems,the results show that this proposed method has great real-time advantages.(4)The control method facing the VCTS under the condition of heterogeneous train performance is designed.This paper took the braking performance of the train as the performance heterogeneous factor and studied the relationship between the braking performance of trains in VCTS and the string stability,and obtain the necessary /sufficient conditions and verification methods of the braking performance of trains to ensure the string stability.Numerical simulation experiments show that the obtained results can provide effective support for the globally stable control for the VCTS with heterogeneous braking performance.There are 35 pictures,10 tables and 83 references. |