Research On Assistant Driving Algorithm For Freight Trains Based On Neural Network

Due to the complicated gradient profiles of freight line in China,the basic resistance characteristics of the train changes under the influence of weather,time and other factors,and the driving process is also affected easily by signal system and other factors,which increases the difficulty of driver manipulation.Therefore,designing a train assistant operation system to guide the drivers can effectively reduce the driver's workload,save operation energy consumption,and improve the line transportation capacity.The core of the train assistant operation system is the real-time speed profile planning algorithm.Firstly,a real-time speed profile planning algorithm based on neural network model is proposed to meet the requirements of safety,stationarity,energy saving and punctuality when train running.The core computing unit of the algorithm is composed of an input building module,a neural network planning module and a traction calculation module.The input and output of the neural network model in the neural network planner are determined by extracting the data features from the results of the quadratic programming offline speed profile planning algorithm.The structure and weight of the neural network model are determined by training and validating abundant speed profile results generated by the quadratic programming offline data generator under different line conditions.In order to verify the correctness,real-time and effectiveness of the planning algorithm,the virtual complex line and the actual line are tested respectively.The simulation results show that the proposed algorithm can be applied to different operating scenarios under different line conditions,and the speed profile calculated by the algorithm proposed in this paper is identical with the speed profile calculated by the offline global planning algorithm based on quadratic programming under the guarantee of real-time computation.The energy-saving rate is 5.98% compared with the experienced driver's operation results.Secondly,aiming at the problem that the output of actual train's basic resistance is time-varying and uncertain caused by weather and other factors during the driving process,the basic resistance coefficient identification module is introduced to improve the speed profile planning algorithm.The recursive least square method with variable forgetting factor to identify the basic resistance parameters.The simulation results show that the parameter identification algorithm converges quickly and the error is very small.The speed profile planning algorithm can re-plan a new speed profile in real time according to the identified basic resistance coefficient.Finally,a driver assistant system framework is proposed,which consists of two parts: off-line training stage and on-line guidance stage.The speed profile planning algorithm proposed in this paper is transplanted to the embedded platform,and the semi-physical simulation with train operation monitor and record device verifies the validity and correctness of the proposed algorithm.