多步k最近邻初值寻优的气压模拟系统遗忘迭代学习控制

Forgetting iterative learning control of air pressure simulation system based on multi-step k nearest neighbor initial value optimization

中文关键词  高速列车  气压模拟系统  k最近邻算法  迭代学习控制  PID控制  初值问题  收敛速度

英文关键词  high-speed train  air pressure simulation system  k nearest neighbor algorithm  iterative learning control  PID control  initial value problems  convergence speed

中文摘要

      高速列车车内压力波动过大会对乘客舒适性造成影响, 而气压模拟系统是一套通过对车内模拟气压跟踪 控制, 实现对乘客舒适性进行研究的装置. 为解决系统历史运行数据利用率低以及存在迭代初始误差导致系统收敛 速度慢的问题, 采用k最近邻(kNN)算法, 建立一种基于历史控制信息的最优初次控制信号提取方法, 并根据迭代学 习控制的基本原理, 将最优控制初值输入到带遗忘因子的迭代学习控制器中, 通过不断迭代来实现车内期望气压轨 迹的跟踪控制, 并和基于大数据的迭代学习控制以及传统PID迭代学习控制进行对比分析. 仿真结果表明: 基于多 步kNN的遗忘迭代学习控制收敛速度更快、系统抖动程度更小、控制精度更高以及算法鲁棒性更好.

英文摘要

      The excessive pressure fluctuation in the high-speed train has an impact on the passenger comfort, and the air pressure simulation system is a device to study the passenger comfort by tracking and controlling the simulated air pressure in the train. In order to solve the problems of low utilization rate of system historical operation data and slow convergence speed caused by iterative initial error, the k nearest neighbor (kNN) algorithm is adopted to establish an optimal initial control signal extraction method based on historical control information. According to the basic principle of iterative learning control, the initial value of the optimal control is input into the iterative learning controller with forgetting factor, and the tracking control of the desired air pressure in the train is realized through continuous iteration. And compared with the iterative learning control based on big data and the traditional PID iterative learning control. The simulation results show that the forgetting iterative learning control based on multi-step kNN has faster convergence speed, less system jitter, higher control accuracy and better robustness of the algorithm.