Improvement Of BP Algorithm And Research On Its Application In The PID Optimization Control

Proportional-Integral-Derivative control, abbreviated as PID control, is a kind ofmechanism of feedback control that is to reduce the error by proportional function, toeliminate static deviation by the integral calculus function and to predict future by thedifferential calculus function. Because of its clear principle, simple structure and easyimplementation, it has been widely used in the control field. The key to the application of thePID controller is to select appropriate parameters in order to adapt to different objects. But foran object with time-varying and large time delay characteristics, the PID controller’s effect isnot very ideal, even hard to achieve control requirements. The main reason is that controlparameters of the PID controller do not have the capability of justifying the parametersonline.As an intelligent algorithm, the BP neural network algorithm is an application moreextensively used in the field of artificial intelligence. It is a kind of error back propagationalgorithm according to the training of the multilayer network, has the excellent capacity oflearning and adaption.In this paper, a kind of improved BP algorithm is put forward, and this BP algorithm isused in the choice of PID parameters, to realize the parameters setting and online adaption.Because the initial weights of the BP network is selected at random, this will greatly affect thestability of the network, thus, in this paper, the optimization function of the adaptive geneticalgorithm is used by, putting forward an algorithm based on adaptive genetic algorithm tooptimize the weights of the BP neural network, in order to improve the parameter self-tuningand the adaptive ability of the PID controller.This paper is to study the basic principle and the structure of the BP neural network, toanalyze the limitations of the BP algorithm and to make proper improvements to the basic BP algorithm. According to the optimization characteristic of the genetic algorithm, this paper isto train the best weights of the BP network, and to combine the improved algorithm with thePID control, to realize the on-line tuning of the PID controller, so that the PID controller hasbetter adaptability and more excellent dynamic performance, and shows that the effectivenessof the proposed algorithm through the example.