Back propagation neural network‐based torque ripple reduction strategy for high frequency square‐wave voltage injection‐based interior permanent magnet synchronous motor sensorless control

Abstract

Abstract In interior permanent magnet synchronous motor (IPMSM) position‐sensorless drives, the high‐frequency (HF) square‐wave voltage injection method is often used to estimate the rotor position and speed in low‐speed range by tracking the salient polarity of the motor. In order to reduce the torque ripple caused by HF signal injection, a strategy to update the magnitude of the injected signal online by back propagation neural network is proposed in this paper. With the proposed method, the neural network can update the magnitude of the injected signal online according to the d‐axis current and the position error information. It can not only ensure the accuracy of position extraction but also effectively reduce the current harmonics caused by the injected signal, and then the torque ripple can be reduced. In addition, the proposed method is easy to implement, resulting in low computation burden. Finally, the experiments are implemented on a 1‐kW IPMSM drive. The experimental results show that compared with the conventional fixed magnitude injection, the peak‐to‐peak value of the torque ripple is reduced by nearly half along with the decrease of the injected magnitude.

Keywords

• back propagation neural network
• interior permanent magnet synchronous motor
• sensorless control
• square‐wave voltage injection method