Phase-locked loop (PLL) type sensorless control of PMSM using neural network filter

Abstract

This paper describes a Phase-Locked Loop (PLL) type sensorless control strategy of permanent-magnet synchronous machines with pulsating high frequency (HF) voltage signal injection. The HF model of the PMSM and the demodulation scheme are analyzed in detail. A neural network (NN) based adaptive filter replaces the common bandpass filter (BPF) to extract the desired frequency signal which includes the rotor position information. The proposed NN filter has an adjustable bandwidth which helps the filter to capture the desired signal even in the presence of high speed estimation errors. The recursive least square (RLS) algorithm is used to update the neuron weights. The simulation results verify the performance of the proposed NN adaptive filter in the sensorless control strategy.