Adaptive neural network PID controller for nonlinear systems

Abstract

In this paper, we are interested in adaptive neural PID control with a reference model for nonlinear systems. A recurrent neural network architecture is studied, and its parameters are computed to mimic a conventional PID controller. Two neural networks architectures, Feed Forward Neural Network (FFNN) and Recurrent Neural Network (RNN) are used for modeling nonlinear systems. To avoid the errors of the Jacobian system approximation, it is best to model the system with a neural network called an emulator. By adding an emulator in parallel with the dynamic system model, the emulator will be trained to learn the dynamics of the system.The back-propagation method is used as the basis for developing algorithms capable of modeling and controlling our nonlinear systems.