基于神经网络延时预测的自适应网络控制系统

针对网络控制系统存在着随机、时变、不确定的信息传输延时,采用带有时间戳的线性神经网络(TSLNN)进行在线延时预测,实时地获得当前采样周期的网络传输延时预测值.该方法选取3个先验的网络实测延时值作为神经网络的输入样本,选用widrow-hoff学习规则作为神经网络的训练算法;应用网络传输延时预测值,并采用一阶Pade方法,对数学模型中的延时环节进行线性化处理,从而获得无刷直流电机调速网络控制系统的线性数学模型;最后,利用模型参考自适应控制方法(MRAC)设计闭环控制器.仿真结果表明,将基于TSLNN在线延时预测的MRAC方法应用于无刷直流电机调速网络控制系统中,可以获得令人满意的系统动、静态性能.

Adaptive networked control system based on delay prediction using neural network

Aiming at existent randomness,time varying and uncertainty of the information transmission time delay in the networked control systems,the time-stamped linear neural network(TSLNN) is adopted to predict the time delay in real-time,on line.Using measurement time delay values of the previous sampled period in actual network as the input data set for the neural network,the widrow-hoff learning rule is chosen as the training algorithm of neural network;With predicted network delay value,and using one-order Pade method to linearize the delay element,the linear mathematical model of the brushless direct-current motor drive networked control systems is established;The model reference adaptive control(MRAC) strategy is introduced to design the close-loop controller for the control system;The results of digital simulation prove that based on TSLNN time delay prediction in real-time,the MRAC for the brushless direct-current motor drive networked control system is feasible,and the dynamic and static response performances of the system are satisfied.