The Design And Simulation Of Neural Network PID In Networked Control Systems

The networked control system (NCS) can be defined as a real-time closed-loop in which all portions are connected through network. NCS is a distributed control system which combines the communication network and control system as a whole, and uses the network as the communication to realize the share of the resources between the sensors, controllers and actuators that distributed in different geographic locations. NCS is else a highly intelligent complex control system which involves automatic control theory, computer science, network communication technology, information theory, software engineering, intelligent instruments of reliability theory and optimization methods. As the network control system uses network as communication means, the characteristics of the network such as time delay and packet loss should affect the performance of the control system inevitably. The traditional control algorithm can’t be suitable for the NCS, it is urgent to study new control strategies to deal with the problems mentioned above. More and more attentions are paid on the researches that how to use classical control strategy to achieve better control performances in NCS.The PID controllers are still widely used in project because of the advantages of sample structure, easy installment, strong robustness, and control parameters related to the engineering indicators, how to apply the PID controller in NCS is a problem worthy to considerate. In this paper, a kind of network control system in which the time delay varies in a sample period is studied. The mainly work of this paper can summarized as follows:Firstly, the time delay and the influencing factors of the NCS are studied, the characteristics of the network delay with different driving mode is analyzed.Secondly, for the NCS that sensors are time driven, controllers and actuators are event driven, the space models are built with the delay less or larger than one sample period respectively.Thirdly, due to the issue that PID parameters is difficult to obtained, the single neuron fuzzy PID controller is proposed with the network delay varying in one sampling period randomly which uses single neurons to construct PID controller. As the learning rate of single neuron is too sensitively, a fuzzy algorithm is adapted to adjust the learning rate dynamically online. Simulation results shows that the single neuron fuzzy PID controller has a good control performance when the network delay is less than one sampling period.Fourthly, as it is difficult to model the network control system, the traditional method of setting PID parameters is unsuitable, a novel control algorithm that using extreme learning machine to turn the PID parameters online is presented. The novel algorithm identifies the Jacobin information of the controlled object with online extreme learning machine firstly, and then uses the Jacobin information to adjust the PID parameters. The simulation using Truetime blocks with Simulink/MATLAB is conducted. It is observed from the simulation that the system has a better performance than traditional PID controller.