Enhancing Speed Loop PI Controllers with Adaptive Feed-forward Neural Networks: Application to Induction Motor Drives

Abstract

This paper proposes the idea to improve the performance of the speed loop PI controller by using feed-forward and adaptive control actions. Indeed, when the system to be controlled is required to track a rapidly changing reference frame, higher bandwidth is usually required, making the system more sensitive to noise and consequently less robust. In such cases, to achieve a better performance in reference tracking while keeping noise rejection capacity, one idea is to use a feed-forward controller, employed to enhance the required tracking, leaving the feedback action to stabilize the system and suppress higher frequency disturbance. As such, this paper analysis the classical PI based field oriented control of induction machine with the proposed adaptive feedforward technique. Simulation and experimental results are presented under nominal conditions, speed varying conditions and loading conditions. Upon analyzing, it is seen that with the proposed controller integrated into the speed loop, the overall response of the system has significantly improved.