Research On Intelligent Operation Control Of High-speed Maglev Train Based On Stacking Ensemble Learning Method

The operation control of high-speed maglev train is to realize the automatic operation of the train.In view that the existing operation control methods based on state space are insufficient to solve the problem of train operation,thus it is of great significance to conduct the research on a simple and effective operation control method in order to ensure the safe,reliable,high-speed and flexible operation of high-speed maglev train.Through studying high-speed maglev train operation mode and the existing train operation control methods,the dissertation put forward the high-speed maglev train operation control method based on machine learning.In view that the ensemble learning model relative to the single learning model is more stable and suitable for the engineering application research,the dissertation optimized adaptive Stacking ensemble learning method and applied it to high-speed maglev train operation control simulation.The details are as follows.(1)The swarm intelligence fusion algorithm was proposed to optimize the integration selection problem.Aiming at the configuration selection and hyper-parameters optimization in Stacking ensemble learning,an appropriate optimization algorithm was designed to meet the requirements of high optimization accuracy and fast search speed.The algorithm fusion mechanism was designed,and the search method of BAS algorithm was introduced into the optimization process of PSO algorithm to obtain BAPSO algorithm(Beetle Antennae Particle Swarm Optimization).Besides,BAPSO-WCF(Beetle Antennae Particle Swarm With Compression Factor)algorithm was obtained by improving the optimization performance of parameters in BAPSO algorithm.By test function,the algorithm parameters were selected,and the experiments of optimization comparison were performed.The experimental results show that the proposed algorithm fusion strategy and the improved strategy can effectively improve the optimization accuracy and search speed.(2)A self-adaptive Stacking ensemble learning algorithm based on BAPSO-WCF(Self-adaptive Stacking Ensemble Learning Algorithm Based On BAPSO-WCF,BSSEL)was proposed.On the basis of the detailed study of Stacking ensemble learning framework and features,its configuration selection and hyper-parameters optimization both received formalized description and modeling,and the calculation process ofmodel solution based on BAPSO-WCF algorithm was designed.By combing the features of Stacking ensemble learning and the research problem of designing the control method based on traction continuous control variable in the dissertation,both learners the candidate set of hyper-parameters were established.Experimental results show that the BSSEL algorithm can obtain a better prediction effect.(3)A high-speed maglev train operation control method based on BSSEL algorithm was constructed and simulated.The relevant data of train operation were collected,screened and then normalized preprocessing was carried out,and the auto-encoder was designed for features learning.The BSSEL algorithm was used for offline model training,and designing the online application process of the model.Then,the model was applied to the maglev train operation control system,and the operation control simulation was carried out based on the test line.