摘要
针对列车运输能力与客流时空分布不匹配的问题,提出列车灵活编组方案和短编组小间隔自适应控制方法。首先,根据新型列控系统的特点,设计列车自适应闭环控制框架,构建列车最小追踪间隔模型,提出大小交路列车在小交路通信协同运行的编组方案。其次,根据列车编组方案短编组小间隔控制需求,基于深度确定性策略梯度(Deep Deterministic Policy Gradient,DDPG)算法设计列车自适应控制模型,模型中通过设置奖励函数实现列车多目标控制,通过设计神经网络输出列车最优控制策略。然后,利用Matlab/Simulink仿真平台,以上海地铁2号线2A-2A列车通信协同运行为例,设置其追踪间隔为10 s,搭建加入随机阻力扰动的列车自适应控制模型并进行训练。研究结果表明:列车停车误差为0.1 m,准时误差为0.2 s,协同速度差小于2 km/h,区间最小追踪间隔为180.6 m,满足列车精准停车、速度协同和间隔安全性要求。针对DDPG算法估值过高而导致列车运行速度波动的问题,将神经网络中的Critic网络进行改进,改进后训练仿真表明:列车停车误差为0.03 m,准时误差为0.1 s,协同速度差小于2 km/h,区间最小追踪间隔为181.1 m,列车停车精度和间隔安全性均有提高,且速度波动相对平稳。最后,为验证该控制方法的实时性和安全性,假设前车在意外情况下减速停车,经仿真验证:后车停车误差为1.3 s,停车间隔为220.8 m,满足实时性和安全性的要求。研究成果可为城市轨道交通提供一种灵活匹配客流量的行车编组方案,为新型列控系统列车短编组小间隔运行提供一种安全高效的自适应控制方法。
To aim at the problem of mismatch between train transportation capacity and spatial and temporal distribution of passenger flow,a flexible train marshalling scheme and a short-marshalling small-interval adaptive control method were proposed.Firstly,according to the characteristics of the new train control system,the adaptive closed-loop control framework of the train was designed.The minimum tracking interval model of the train is constructed.The marshalling scheme of full-length and short-turn routing trains operating together in short-turn routing communication was proposed.Secondly,according to the requirements of short-marshalling and small-interval control of train marshalling scheme,a train adaptive control model was designed based on Deep Deterministic Policy Gradient(DDPG)algorithm,in which the multi-objective control of train was realized by setting the reward function.The optimal control strategy of train was output by designing a neural network.Then,using the Matlab/Simulink simulation platform,taking the 2A-2A train communication coordinated operation of Shanghai Metro Line 2 as an example,the tracking interval was set to 10 s,and the train adaptive control model with random resistance disturbance was built and trained.The simulation results are as follows.The train stopping error is 0.1 m,the punctuality error is 0.2 s,the coordinated speed difference is less than 2 km/h,and the minimum tracking interval between the interval is 180.6m,which meets the requirements of precise stopping,speed coordination and interval safety of the train.Aiming at the problem of train speed fluctuation caused by excessive valuation of DDPG algorithm,the Critic network in the neural network is improved.The training simulation after improvement shows that the train stopping error is 0.03 m,the punctuality error is 0.1 s,the coordinated speed difference is less than 2 km/h,the minimum tracking interval between the interval is 181.1 m,the precise stopping and interval safety of the train are improved,and the speed fluctuation is relatively stable.In order to verify the real-time and safety of the control method,assuming that the front train decelerates and stops in an unexpected situation,the simulation yields:the rear train stopping error is 1.3 s and the stopping interval is 220.8 m,which meets the requirements of real-time and safety.The research results can provide a flexible marshalling scheme for urban rail transit to match the passenger flow,and provide a safe and efficient adaptive control method for the short-marshalling and small-interval operation of the new train control system.
作者
朱爱红
田晓晴
何明明
ZHU Aihong;TIAN Xiaoqing;HE Mingming(School of Automation and Electrical Engineering,Lanzhou Jiaotong University,Lanzhou 730070,China;CRSC Research&Design Institute Group Co.,Ltd.,Beijing 100070,China)
出处
《铁道科学与工程学报》
EI
CAS
CSCD
北大核心
2024年第3期969-979,共11页
Journal of Railway Science and Engineering
基金
国家自然科学基金资助项目(52162050)
中国国家铁路集团有限公司基金资助项目(N2022G012)。
关键词
新型列控系统
大小交路
协同运行
DDPG算法
自适应控制
new train control system
full-length and short-turn routing
coordinated operation
DDPG algorithm
adaptive control
