基于混合系统建模预测控制的列车自动驾驶优化运行  

Automatic train optimization operation based on hybrid system modeling predictive control

作　　者：汤旻安[1,2] 王攀琦 TANG Minan;WANG Panqi(School of Automation and Electrical Engineering,Lanzhou Jiaotong University,Lanzhou 730070,China;School of Mechanical and Electronical Engineering,Lanzhou University of Technology,Lanzhou 730050,China)

机构地区：[1]兰州交通大学自动化与电气工程学院,甘肃兰州730070 [2]兰州理工大学机电工程学院,甘肃兰州730050

出　　处：《铁道科学与工程学报》2019年第6期1527-1534,共8页Journal of Railway Science and Engineering

基　　金：国家自然科学基金资助项目(61663021);甘肃省高等学校科研资助项目(2017A-025)

摘　　要：针对现有列车自动驾驶速度追踪精度不高的问题,提出一种基于混合系统神经网络反馈补偿控制的模型预测控制算法。根据混合系统建模的特点与优势,引入辅助变量,建立混合系统列车运行动力学模型。为了便于求解包含约束的预测控制律,采用二次规划方法求出满足列车各项性能指标的控制作用序列。神经网络反馈控制器用于对系统目标速度与实际速度之间的误差进行在线学习并求出一个补偿控制量,并将补偿后的控制力作用于列车系统模型。研究结果表明:该控制结构包含补偿控制策略,可以较大程度减小系统跟踪误差,保留模型预测控制的优势,同时提高系统的控制精度。Aiming at the problem that the automatic tracking speed of existing trains is not high,a model predictive control algorithm based on neural network feedback compensation control was proposed.According to the characteristics and advantages of the hybrid system,the auxiliary variables were introduced to establish the dynamic model of the hybrid train operation,which was convenient for solving the predictive control law with constraints.The secondary planning method was used to find the control action sequence that satisfies the various performance indicators of the train.The neural network feedback controller was used to learn the error between the system target speed and the actual speed and find a compensation control amount.The compensated control force was applied to the train system model.The simulation and experimental results show that the control structure includes compensation control strategy,which can reduce the system tracking error to a large extent, retain the advantages of model predictive control,and improve the control precision of the system.

关 键 词：混合系统 神经网络 模型预测控制 自动驾驶 优化运行 

分 类 号：U281[交通运输工程—交通信息工程及控制]