Current signal denoising of PMSM by matching pursuits algorithm with undercomplete dictionary

Abstract

Motor stator current signal with noise in current loop would reduce the performance of permanent magnet ac servo system. This paper introduces a new method-matching pursuit which is applied in the denoising of current signal. The proposed method designs the optimal atom based on the similarity between original signal and atoms by using matching pursuit (MP) algorithm. The conventional MP algorithm needs large amount of calculation due to the overcomplete dictionary. To overcome this problem, this parper demonstrates the effectiveness of MP algorithm with undercomplete dictionary. The undercomplete dictionary is consist of sinusoidal atoms after analyzing the characteristic of actual current signal. The signal set partitioning method is proposed to optimize the configuration of dictionary. Simulation and experimental results verify the feasibility and efficiency of the proposed method.