结合迭代学习和模型预测的重载列车运行控制  

Operation Control of Heavy-Haul Train Based on Combination of Iterative Learning and Model Prediction

作　　者：孙鹏飞[1] 张传鑫 蒋春宏 魏咪 王青元[1] SUN Pengfei;ZHANG Chuanxin;JIANG Chunhong;WEI Mi;WANG Qingyuan(School of Electrical Engineering,Southwest Jiaotong University,Chengdu Sichuan 611756,China)

机构地区：[1]西南交通大学电气工程学院,四川成都611756

出　　处：《中国铁道科学》2023年第2期111-119,共9页China Railway Science

基　　金：国家自然科学基金资助项目(62003283)。

摘　　要：为实现重载列车单次行程的高鲁棒高精度轨迹跟踪,根据列车纵向运动特性,构建重载列车多质点的动力学模型;基于利用批次化的运行过程积累控制经验,结合迭代学习和模型预测控制方法设计1种增强抗扰的重载列车跟踪控制器,将重载列车动力学模型转化为基于模型预测控制框架下的线性二次型最优控制模型,用二次型最优控制的速度和位置状态反馈增益表示迭代学习增益,利用批次化积累的控制经验不断提高跟踪性能,实现单次行程的滚动时域优化,提升轨迹跟踪的鲁棒性和精度;对某货运专线上的2万t重载列车进行跟踪控制仿真,分别从时域稳定性和迭代收敛性验证该控制器的稳定性。结果表明:结合迭代学习和模型预测控制方法能够很好地利用重载列车系统操纵重复性特征并实现全程跟踪控制,较传统控制方法跟踪效果更好并能有效降低列车纵向冲动,同时能够动态响应非重复性扰动,满足重载列车运行控制要求。In order to realize high robustness and high precision trajectory tracking of heavy-haul trains during a single trip,a multi-particle dynamic model of heavy-haul trains is established according to the longitudinal motion characteristics of trains.Based on the control experience accumulated by the batch operation process,combining with the methods of iterative learning and model predictive control,a tracking controller for the heavy-haul trains with enhanced disturbance immunity is designed.The dynamic model of heavy-haul trains is transformed into a linear quadratic optimal control model based on the model predictive control framework,and the iterative learning gain is expressed by the feedback gain of the speed and position state of the quadratic optimal control.Use the control experience accumulated in batch process to continuously improve the tracking performance,to realize the rolling time-domain optimization of a single trip,and to improve the robustness and accuracy of trajectory tracking.The tracking control simulation of a 20,000 t heavy-haul train on a freight special line is carried out to verify the stability of the controller from aspects of time-domain stability and iterative convergence.The results show that the combination of iterative learning and model predictive control method can make good use of the manipulation repeatability characteristics of the heavy-haul train system and realize the whole-process tracking control.Compared with the traditional control method,the tracking effect is better and can effectively reduce the longitudinal impulse of the train.Meanwhile,it can dynamically respond to non-repetitive disturbance and meet the requirements of the heavy-haul train operation control.

关 键 词：迭代学习 模型预测控制 重载列车 多质点动力学模型 线性二次型最优控制 

分 类 号：U284.48[交通运输工程—交通信息工程及控制]