DNN predictive magnetic flux control for harmonics compensation in magnetically unbalanced induction motor

Space and time harmonics in an induction motor (IM) increases due to motor fault and eccentricity leading to voltage unbalance and harmonic distortion.In this paper, a novel deep neural network (DNN) predicted magnetic flux reference control as shown in Fig. 1(a) has been designed in order to continue operating the faulty motor. The proposed flux predictive control constitutes three major parts: detection block of UMP, on-line harmonics compensation block and finally DNN optimized flux predicted block. The harmonic compensation block operates using Fourier series of magnetic flux density waves as expressed in (1) where Bp, Bμ represent time harmonics and Bλ, Bλe represent space harmonics.