Research On Modeling And Control Of High Speed Train Based On Support Vector Machine

High-speed trains has been fairly rapid development of times, It is composed of complex equipment including hardware and software system The running environment is complex,with complex characteristics of nonlinear kinematic performance of mobile service system.To a large extent existing high-speed train operation is based on the actual operation condition and the given V-S curve according to the experience of the manual operation mode of operation control, So it is difficult to ensure the high efficiency of high speed trains the performance of the complete operational standards.To improve the complicated high speed trains systems, The process model of high speed train can be got on the basis of the actual data collection of existing lines running process information and traction/braking characteristics. In order to further realize the speed tracking control of the train and the future study of the preparation for the design of driving system.1.The study takes CRH380 AL as the research objective.Through the actual data information and the drawing of the reference braking characteristic curve, Using the data-driven modeling method,and the high-speed train LSSVM forecasting model,Real-time computation and other adverse characteristics of long time reference put forward improved least squares support vector model of online learning,to improve the modeling speed and accuracy. To a great extent to meet the online real-time requirements and improve the modeling speed and accuracy, CRH380 AL as the result of the experiment show that the reliability and validity of the established model.2. According to the complexity of the high-speed train operation process study, Due to the influence of external random uncertainty we put forward the idea of multiple model modeling because of data usually contains a large amount of information, The movement process of the high-speed train traction and braking condition on the data using fuzzy clustering to classification, and then phase model is established according to the classification of data, at the same time, consider a kernel function of support vector machine(SVM) model have a greater influence upon the accuracy and effectiveness of the model, so the phase model is established on the basis of seeking the most suitable kernel function optimal LSSVM model,experiments show that under the test of classification modeling, model preserves more information, to establish the optimal model provides more accurate, the basis of the experimental model prediction accuracy is higher.3. Optimal LSSVM model are built on the basis of multiple support vector machine(SVM)is put forward with the combination of generalized predictive control model of high-speed train speed tracking control method, the high-speed train traction and braking force and speed prediction tracking control objectives provide a reference for further research on autopilot system design, as well as CRH380 AL as experimental object using multi-model data-driven modeling combined with predictive control method for the simulation experiment, the results show that the single model and mechanism model predictive control method has more advantages.