Position estimation at zero speed for PMSM using probabilistic neural network

Abstract

The paper presents a method for estimating the shaft position of a synchronous motor with permanent magnets (PMSM) for the zero and very low speed range. The method is based on the analysis of the high frequency currents, which are induced by the additional test voltage in a stationary coordinate system associated with the stator. Although this method involves the identification of currents hodograph, the method does not need to calculate the current ellipse position. Presented method involves a comparison of obtained shape to the reference pattern using probabilistic neural network (PNN). The method can achieve satisfactory accuracy in a case the high asymmetry of the inductance, as well as in the case of small values of the inductance asymmetry ratio, also in the case of a high level of noise.