延时预测内模网络控制系统

Internal model control with delay prediction for networked control systems

网络控制系统中存在的随机延时会导致系统的性能变差甚至不稳定.针对该问题提出了时间戳BP神经网络延时预测方法,利用历史延时数据预测下一个时刻的延时值,根据得到的延时预测值,结合内模控制原理进行网络控制系统的设计.延时环节采用Pade一阶展开近似表达法,分析了延时预测对系统性能的影响.仿真结果表明,延时预测精度有保证,系统性能仅受内模控制器滤波系数α和延时预测精度的影响,网络控制系统性能得到了很大的提高.

A delay prediction method based on time-stamped BP neural network was proposed to solve the problem that the performance of networked control systems goes bad and unstable due to the stochastic delay. This method used the measured historical delay data to predict the delay value during the next period, according to which the networked control system with internal model control was designed. The delay loop was approximately described by one-order Pade expended method. The effect of delay prediction on the system performance was also analyzed. Simulation illustrates that this method guarantees the precision of delay prediction, and the system performance is only affected by the filter coefficient a and the delay prediction precision, thus the performance of networked control system is greatly improved.