Initial Rotor Position Estimation for an Ultra-low Inductive Saliency Machine Based on Eddy Current Loss

Initial rotor position detection is essential to ensure the accuracy of sensorless control for surface permanent magnet synchronous motor (SPMSM) drives. In order to extract position information for SPMSM with an ultra-low inductive saliency, this paper presents the estimation method of rotor position based on eddy-current-reflected resistive saliency. This method separated high-frequency (HF) eddy-current-reflected resistance from HF impedance and extracted the HF voltage vector corresponding to the resistance. To recognize accurate subtle position signatures in the extracted HF voltage, the method harnessed the convolutional neural networks (CNN) to establish the relationship between the HF voltage image and the rotor position. Finally, the proposed method is experimentally evaluated and proven to have a promising performance on initial position estimation for an ultra-low saliency machine.