网络控制系统前向通道随机时延的在线多步预测

Online Multi-step Prediction for the Random Delay of the Forward Channel in Networked Control System

针对网络控制系统中前向通道随机时延带来的不确定性问题,在时序分析的基础上,提出一种基于自回归(AR)模型的时延在线预测模型.采用True Time构建闭环仿真系统,由反馈通道采集执行器端的时延信息,经预处理后用于建立AR模型,并由递推最小二乘法在线更新模型参数;最后,通过滚动预测实现前向通道时延的多步预测.实验结果表明:与基于广义回归神经网络的在线预测模型相比,本文提出的前向通道随机时延的在线预测模型具有更好的预测性能.

To address uncertainty caused by random delay of the forward channel in a networked control system,we propose an online predictive model based on an autoregressive(AR) model on account of timing analysis.We use the True Time toolbox to establish the simulation system in which the delay series is collected through the feedback channel.After pretreatment,we establish the AR model with the delay series,and the parameters are updated online by using a recursive least-squares algorithm.A rolling forecast is considered to realize multi-step prediction for the delay of the forward channel.Experimental results suggest that this online predictive model has better predictive performance than the GRNN-based model.