基于拓展卡尔曼滤波的货运列车载重估计算法

Load Estimation Algorithm Based on Extended Kalman Filter for Freight Train

货运列车的作业场景多样、线路条件恶劣,其牵引、制动系统具有大时延、强约束等特点。目前在货运列车的控制系统中,通常通过建立复杂的列车多质点运动模型来设计控制算法以实现精确控制。但在实际运营中,货运列车的编组情况、作业时间和牵引质量都不固定,车载设备获取精准模型参数的代价高,这也是货运列车自动驾驶(automatic train operation,ATO)系统难以大规模应用的主要原因。针对货运列车牵引质量不固定的问题,文章提出了一种基于拓展卡尔曼滤波(EKF)的货运列车载重估计算法,其在ATO系统控制列车起动过程中,以ATO系统的控制输出和列车运动状态作为输入,实时估算列车的载重,校正控制系统模型和参数,减少了司机在人机交互界面频繁输入/确认列车数据的操作,同时也提高了ATO系统的控制精度和效率。仿真结果显示,在司机输入相对准确载重值的基础上,所提算法可以估计得到更准确的载重值,在2 000 t载重仿真实验中,估计误差控制在3.5%~4.0%之间。这表明该算法能有效降低ATO系统对多质点运动模型的要求,提高了ATO系统控车的精度。

Freight trains operating in various scenarios with severe track conditions require traction and braking systems characterized by large delays and strong constraints.Researchers usually build complex train multi-particle motion models and design control algorithms,seeking to establish accurate control in the control system of freight trains.However,in actual operation,due to flexible formation,operation schedule,and traction weight of freight trains,the cost for on-board equipment to obtain accurate model parameters is very high,which is also the main reason hindering the large-scale application of the automatic train operation(ATO)system on freight trains.This paper presents a load estimation algorithm based on extended Kalman filter(EKF)for freight trains,in response to their flexibility in traction weight.This algorithm is designed for real-time load estimation upon train starting under the control of the ATO system,based on inputs composed of ATO system control outputs and train movement status.The following corrections to the control system model and parameters can reduce the frequent operation of drivers by entering/confirming train data in the human-computer interface.The algorithm also improves the control accuracy and efficiency of the ATO system.Simulation results showed that the proposed algorithm generated more accurate load values,compared with relatively accurate load values entered by drivers.In the simulation experiment of 2,000-ton load,the estimation error ranged from 3.5%to 4%,demonstrating the proposed algorithm could effectively improve the accuracy of the ATO system,with lower requirements on multi-particle motion models.