Research On Random Time Delay Prediction Of Networked Control System

The rapid development of communication, control and computer technology affect the structure of control system greatly, with the control objects become more and more complex and distribution region expanding constantly, the disadvantage of conventional point-to-point control system such as complex wires deployed, hard maintenance, poor extensibility and high cost is increasingly prominent. In this context, the networked control system which exchanges the information between controller, actuators and sensor through the shared network emerged. Compared with conventional point-to-point control systems, the advantages of NCS are less wiring, lower install cost as well as greater agility in diagnosis and maintenance so that NCS has gotten wider application. Now NCS has became a hot research issue in the international control field.However, network is not a reliable data transmission medium, many new problems are introduced in analyze and design NCS by using band limited network to transmit a huge number of real time date. Where the time delay and the data packet dropout are the most two common problems, these two problems will induce the performance of network control system and even lead the system instability. Because the data packet dropout can be regarded as infinite time delay, so the most important issue is how to decrease the effect of network induced time delay on NCS. This paper mainly researches the way of predicting network time delay, a combined forecasting model to predict the network time delay is present, in order to obtain higher accuracy prediction, improve the compensate effect and alleviate the influence of network induced time delay on networked control system. The details are as follows:First, compiling a program of C++language in Visual Studio2010environment to measure the one way network transmission time delay in LAN by using time stamp, nearly300-thousand data have been obtained for their distribution characteristic analysis.Secondly, ARMA model and RBF neural network are adopted to predict network time delay and then compare their predicting effect.Finally, a combined forecasting model is expressed, ARMA model and RBF neural network are chosen as the submodel to make up the combined model and how to construct and operate the combined model was finally given. According to the simulation experiment, the predicted accuracy of combined model is better than the accuracy of single model.