Research On Speed Tracking Control And Parameter Intelligent Optimization Method Of High-speed Train

With the gradual coverage of China’s high-speed railroad network and the continuous acceleration of urbanization,rail transit plays an important role as a key infrastructure,an important means of transportation and a major artery of the national economy.High-speed trains,with their high operating speed,high passenger capacity,comfort and punctuality,are now becoming an increasingly popular choice.High-speed train operation has to adapt to a variety of operating environments,there are various unknown uncertainties,coupled with the train travel speed and density requirements gradually increased,there are still unmodeled parts inside the train,which also puts new demands on the train speed tracking control accuracy.Therefore,research to improve the anti-interference and speed tracking accuracy of trains under the premise of ensuring safety is of great significance to promote the intelligent development of automatic train driving.The main elements of the research in this paper are as follows:(1)By analyzing the principle of train operation and the action of each part of forces on the train during operation,comparing the different points of single-point model and multi-point model,and then considering that this paper focuses on speed tracking,we choose to establish a single-point model for high-speed trains and derive the train dynamics equations;the active disturbance rejection control is introduced and its composition structure and the working mechanism of each part are analyzed.(2)In order to simplify the active disturbance rejection control structure and improve the controller nonlinear performance while preserving nonlinearity,an improved Nonlinear Active Disturbance Rejection Controller(NLADRC)is proposed.Firstly,considering the complex structure of the original transition process and the noise interference in the actual signal input,the simulation experiment is designed to compare the effect of two tracking differentiators,and the Levant differential tracker with simple structure and good de-noising effect is selected.Secondly,for the problem that the nonlinear function is not smooth at the inflection point and the train may have large disturbances leading to large errors during operation,an improved nonlinear smooth function impfal(e,α,δ)is proposed and improved Nonlinear Extended State Observer(NLESO)and improved Nonlinear State Error Feedback(NLSEF)are designed.Finally,the controller and train model are designed in Simulink to verify that the improved algorithm can overcome the chattering phenomenon,give small gain when the error is large,and improve the anti-interference ability and tracking accuracy.(3)To address the problem of the high cost of trial and difficulty of parameter adjustment due to the many and coupled train speed controller parameters and long train running times,the independent updateability of Continuous Action Reinforcement Learning Automata(CARLA)is used for adjustment.Considering that inappropriate parameter sets can lead to system divergence and learning method failure,combining the characteristics of Levant differential tracker and impfal(e,α,δ)function to avoid unnecessary parameter coupling to aggravate the learning burden,the Long Time Continuous Action Reinforcement Learning Automata(LT-CARLA)optimization algorithm is proposed.The speed tracking accuracy is used as the index and verified under different line conditions.The results show that the parameters derived from the parameter learning method in this paper improve the controller response speed,tracking accuracy,and have better anti-interference performance.