IMPLEMENTATION OF NARMA-L2 NEUROCONTROLLER FOR SPEED REGULATION OF SERIES CONNECTED DC MOTOR

Abstract

Series connected DC Motor Drive (SCDM) has been the most suitable drive for electric rapid transit railway system and electric vehicles, characterized by speed and torque control using PID based choppers. However there are some serious problems underlying such as nonlinear switching, variable nature of load and difficulty in exact mathematical model representation. These may cause poor dynamic response of the system. To improve the performance of the system these nonlinearities are to be properly incorporated in system representation and appropriate model of SCDM to be considered. Based on the nonlinear mathematical model of series connected DC motor system dynamics, the analysis and design of two speed controller presented, which are non linear PID and nonlinear autoregressivemoving average L-2 (NARMA L-2) controllers. The entire system has been modeled using MATLAB 8.0 Simulink and power system block sets. The speed response of SCDM is observed by giving reference inputs as speed and load torque in terms of step variation. The response of the system using NARMA L-2 Neuro Controller and classical PID Controller are compared. It was observed that the NARMA L-2 Neuro Controller performance was better in terms of rising time, overshoot and steady state error over classical PID Controller.