Adaptive Neural Network PID Controller

Abstract

Complex and coupled nonlinear systems cannot be efficiently controlled using conventional Proportional Integral Derivative (PID) controller. To obtain good control performance, the parameters of the PID controller should be suitably tuned and adapted to the system parameters variation. In this paper, the PID controller parameters are obtained from a Multi-Layer Perceptron (MLP) neural network and their values are online tuned using the back propagation method. The Adaptive Neural Network PID (ANNPID) controller is analyzed and compared with the conventional PID controller, the type 1 Fuzzy Logic Controller (type 1 FLC) and the type 2 Fuzzy Logic Controller (type 2 FLC) through computer simulation and experimental study. The obtained results show that the neural network PID has better performance than the other controllers and its parameters can be easily tuned to compensate the system parameters variation and the disturbances effects.