Research On Key Technologies Of Network Measurement And Control In Modern Distributed Monitor And Control System And Their Applications

With the development of industrial network in modern distributed monitor and control system, the control network has been connected to wide area network and enterprise network tightly. The architecture of monitor and control network became more open and complex that many characteristics of the network have been changed. In order to make industrial network work well for all monitor and control sub-systems, it is necessary to do some researches on network measurement, prediction and control of industrial network systems.In this paper, some key technologies of network measurement and control in modern distributed monitor and control system are researched, which include survey and analysis of network delay, static prediction model of network delay in industrial system, dynamic prediction model of network delay in industrial system, reliability of topology of industrial network, fault-tolerant of gateway in control network, network behavior of control systems, traffic control of streaming media based surveillance system, and scheduling of network resource for control systems. Furthermore, the feasibility for practical implementation is also mainly concerned in this thesis.The main contents and innovations of this thesis are as follows:(1) In chapter 1, the background and significance were reviewed. Then, the development of modern industrial network, construction of network-based remote surveillance system, and status of network measurement were discussed. At last, the challenges of industrial network in modern distributed monitor and control system were analyzed.(2) In chapter 2, the basic theories of network measurement, the research status of network prediction, and the classification of packet identification were introduced. The performance of industrial network and the methods of traffic control are discussed. Later, quality of service of network and scheduling of traffic were summarized.(3) In chapter 3, the differences of characteristics of some algorithms in network measurement and prediction were compared. And prediction of network delay in monitor and control system was discussed. One nonlinearε-support vector regression-based static prediction model of industrial network delay was proposed. The periodicity of this model is constructed by working day. And the data of model's prediction samples are calculated by round trip time. At last, this model was implemented in real system. The results of experiments show that the static prediction model based on e-support vector regression can work well in real systems, and it will give more precise results than other similar models.(4) In chapter 4, the short-period characteristic of network delay in dynamic monitor and control system was analyzed. The network reliabilities in industrial network with different topologies were discussed, and the influences of dynamic network were studied. Then, one nonlinearε-support vector regression-based dynamic prediction model of industrial network delay was proposed. The periodicity of this model is constructed by network delay of every control circulation. Some nonlinear kernel functions were researched respectively. The results of experiments show that the dynamic prediction model based on e-support vector regression can work well in real-time and provide accuracy comparable to or even better than that of BP-ANN methods.(5) In chapter 5, the characteristics of streaming media based surveillance system with many sub-control systems were analyzed, and an abstract model of video compression and network transportation was provided. Subsequently, the utility function of sub-systems was designed, and the existence and uniqueness of Nash Equilibrium in non-cooperative network scheduling game was proved. At last, a method of optimization based on genetic algorithm was presented. The results of experiment and simulation show that the non-cooperative model can directly influence on the competition of sub-systems and the method can optimize resource allocation of network bandwidth.(6) In chapter 6, the network behaviors of different control systems in industrial network were presented. Then, different network protocols in mixed network control system were analyzed. One priority queuing model based on network behaviors for traffic modeling of data packet queue of gateway was provided. In order to control P2P applications, one model of receiving and forwarding P2P packets is constructed. The algorithm of traffic control in P2P includes two important factors: rejection probability of P2P request and disconnection probability of P2P channel. The results of experiment show that the traffic control algorithms can successfully identify different protocols of sub-systems and control bandwidth fairly.(7) In chapter 7, the advantages of authority software and standard software of network measurement were analyzed, such as TCPDUMP and Wireshark. Based on contents and innovations in previous chapters, one suite of network measurement, prediction and control software system was developed, which can be deployed in distributed monitor and control network. This software system supports multi-linkage unit including some clusters of measurement unit and control unit. And the convenient graphic user interface was designed. In real systems, this software system can identify data of different control system precisely and control all traffic in industrial network well.Finally, the conclusions of this thesis were summarized and discussed.