Initial Rotor Position Estimation of PMSM Based on DC-Link Voltage Dip Analysis Using Neural Network

This paper proposes an initial rotor position estimation scheme of a Permanent Magnet Synchronous Motor (PMSM). It is based on an analysis of the DC-link voltage dips that result from voltage pulse injection, where a neural network (NN) is applied. The effect of magnetic saturation due to sequential voltage injections to different phases and magnetic flux is utilized to incur a variation in inductance. The injected voltage pulse causes a dip in the DC-link voltage. The change in inductance and consequently the voltage dip depend on the rotor position. The NN is used to analyze the dips in the measured DC-link voltage, where the additional process needed to determine the rotor polarity is eliminated since the measured voltage differs in accordance with the polarity. The effectiveness of the proposed method has been verified through the simulation, with an absolute average error of 0.157° in the estimated initial position. The proposed method is used to estimate the initial rotor position, 151°, which is then utilized to perform the starting ramp speed operation; 0 - 1200 rpm.