Research On Generation And Optimization Of Assistant Driving Strategy Of Heavy Haul Train Based On Ensemble Learning

Heavy haul railways play an important role in bulk goods transportation owing to the advantages of low energy consumption,low cost and high efficiency.At present,the heavy haul train in China is mainly operated by the drivers according to the guidance driving curve formed by experience.Due to the heavy load and long length of heavy haul trains,and the complex line conditions of "West to East Coal Transportation" heavy haul railways,it is very difficult and laborious of the manual driving for the train drivers.Moreover,with the increase of traffic density brought by the demand for improving the transport capacity of heavy haul railways,higher requirements are put forward for the driver’s operating quality.Hence,to ensure the safety and smoothness of train operation,it is of great significance to study the assistant driving strategy of heavy haul train.This thesis focuses on the research of the manipulation of the timing of exerting and releasing air braking and the value of the traction or electric braking force in the driving strategy of heavy haul train.At first,the data set is obtained through data preprocessing based on the historical train operation data of an actual heavy haul railway in China.Then,according to the discrete characteristics of air pressure reduction and the continuous characteristics of traction or electric braking force,an ensemble learning method is applied to construct the air braking control model and traction / electric braking control model respectively.Then,on the basis of the practical operational specifications of heavy haul train,the outputs of two trained models are constrained and adjusted to realize the intelligent control of the two control quantities.In view of the issue that the timing of driver’s manipulation and output value of control quantities affect the smoothness of train operation under the scenario of complex variable slopes,taking the minimization of intrain force as the optimization objective,considering the line conditions and operational regulations of heavy haul railway as constrains,and regarding the train working condition as the decision quantity,the decision-making optimization model of driving strategy for heavy haul train is formulated.The proposed improved genetic algorithm and the established dynamics model of heavy haul train are introduced to solve the optimization model,thereby obtaining the optimal driving curve of heavy haul train on the complex variable slopes.Based on the line data and train formation data of an actual heavy haul railway,the driving strategies of heavy haul train under different scenarios are simulated and verified.The simulation results show that under the scenario of long steep down-slopes,the final controller can reasonably output the air pressure reduction and traction or electric braking force,and the maximum in-train force during train operation meets the safety evaluation criteria of train longitudinal force.Under the scenario of complex variable slopes,the smooth driving curve generated by the optimization algorithm proposed in this thesis meets the requirements of safe driving indicators,and compared with the actual manual driving curve,the maximum in-train force of the smooth driving curve during train operation is reduced by 18.7%,which further illustrates the effectiveness of the proposed method.There are 26 figures,13 tables and 67 references.