Neural Network Pid Controller And Decoupling Applications

This paper discusses the standard PID algorithm and the incomplete differential PID algorithm firstly, and analyzes some advantages of the incomplete differential PID algorithm. In terms of dealing with the system's time variation and nonlinear, neural networks have obvious preponderances. Neural network controllers are constituted by combining the standard PID algorithm and the incomplete differential PID algorithm with neural networks, especially the RBF neural network controller based on the incomplete PID algorithm and the PID neural network controller based on the incomplete PID algorithm. These neural network controllers are applied to the linear and nonlinear plant's single-variable control, and examples with simulation and comparison have been carried on, especially focusing on RBF neural network control based on the incomplete PID algorithm. Simulation results show that the neural network control based on the incomplete differential PID algorithm has a quicker response, a smaller error, a better stability, and a better adaptive ability than the neural network control based on the standard PID algorithm has.The research of the neural network control technology is used for industrial practice, considering multivariable decoupling control in this paper which is one of its applications. The neuron control based on the incomplete PID algorithm is applied to decoupling control, the RBF neural network control based on the incomplete differential PID algorithm and the RBF neural network control based on the standard PID algorithm using for decoupling control are proposed, and comparing the decoupling effects of two algorithms. According to the flaw of the traditional RBF neural network, a new structure of RBF neural network which has dynamic performance is used, and the new RBF neural network controller by combining this network with standard PID algorithm is applied to multivariable systems decoupling. Then, examples of RBF neural network decoupling control based on the standard PID algorithm is compared with new RBF neural network controller. The results show that although the effect of RBF neural network decoupling based on the incomplete differential PID algorithm is good, the quality of new RBF neural network used is more satisfactory, that is to say, a higher precision, a stronger ability of adapting to the environment and a little computation.Finally, the paper summarizes the researching work, lists some conclusions achieved and proposes further research.