DNN-Based Predictive Magnetic Flux Reference for Harmonic Compensation Control in Magnetically Unbalanced Induction Motor

A stator inter-turn fault occurring in one of the phases of a three-phase induction motor (IM) gives rise to high harmonics distortion in air-gap flux density, increased torque ripple, temperature rise in the stator windings, and mechanical vibrations due to varying magnetic forces and magnetic noise. The fault leads to a change in the electromagnetic field generated when compared to that during the normal motor operation. An incipient stator fault leads to variation of machine's parameters, causing malfunction of the motor drive. Hence, it is of significant importance to detect the incipient fault before complete motor breakdown occurs. In this paper, a novel magnetic flux reference predictive method for control has been presented by using a harmonic compensation block in coordination with deep neural network (DNN) as a feedforward method to continue safe operation of motor after occurrence of incipient stator fault. This method takes into account both time and space harmonics discrepancies produced due to the fault. The proposed method has been implemented on a 7.5 hp IM using online observer of unhealthy conditions and compensated using DNN predictive methodology.