Research On Control Method Of Neural Network Based On PID

The PID controller has been developed for about 70 years. It has become one of the main technologies of industrial control due to its advantages of simple structure, good stability, reliable work and convenient adjustment. With the development of industry, the complexity of object keeps deepening, the conventional PID control is powerless, especially for the non-linear complex system. Therefore, the application of conventional PID control is facing enormous limitations and challenges. The application of neural network in the control system improves the capability of information processing, adaptive capacity and intelligence level of the whole system. In addition, the neural network has the capability of approaching arbitrar y continuous bounded nonlinear function, combined with conventional PID control, being an effective way to solve the problem for the nonlinear system and uncertain system.This thesis mainly applies the neural network subject and control theory to study on the neural network PID controller in depth, and fulfills the following work for the design of neural network PID controllerFirstly, it briefly introduces the basic theory of neural network subject and the learning method of neural network, analyzes the shortcomings of conventional PID controller with the complex, dynamic and uncertain control system, and also presents five improving methods for the problems.Secondly, this thesis mainly studies on the neural network PID controller. The neural network possesses a strong capability of nonlinear mapping, self-learning, associative memory, parallel information processing and fine fault-tolerant performance.By the improvement of the conventional PID controller with the neural network, it has a better control effect for the complex control system in the industrial control, and improves the unsatisfactory control effect caused by the change of structure and parameters effectively. The thesis deeply studies on the neural network neural PID controller, neural network PID controller based on BP algorithm, proposes the conjugate gradient algorithm the and the combination of optimizing neural network PID controller based on the BP network to solve the problem of falling into local minima while using BP algorithm and long learning time, even might not realizing the purpose of learning.Finally, it compares the conventional PID controller and neural network PID controller by simulation. The simulation results show that the application of neural network for improving conventional PID control enhances the robustness and dynamic characteristics of the system, which effectively improves the control effect of the system and achieves the expectant goal.