Neural network based regulator for linear and nonlinear dynamic systems

The aim of this paper is presenting a new structure based on a neural network for online regulating unknown systems which can be linear and nonlinear. Regulator design in this paper means that building a structure in order to have steady state zero response for any inputs. The proposed structure contains a modelling block based on neural network for unknown system and this block is trained with backpropagation algorithm. In addition, a predictive block is embedded before the unknown plant and a low-pass filter is also applied to reduce the high frequency components of the response at the beginning of the process.

Keywords:

Artificial neural network , Backpropagation Algorithm , Regulator , Plant Identification , Feedforward Neural Network.