Robust Signal Offset Identification for Sensorless Control of Induction Machines at Rated Load using One-Active Modulating Pulse Excitation

One way to realize zero-speed sensorless control of ac machines is to establish voltage step excitations, whose current responses deliver information about machine saliencies, finally leading to the flux position required for the control. The use of only one active inverter state for saliency-offset vector calculation results in low current ripple. However, signal offsets appear in the vector, whose magnitude can be significantly bigger than the saliency information, especially in induction machines, where the slotting signal is often used as control signal. Besides, offsets are dependent on saturation level and in addition vary in angle and magnitude, depending on the active inverter state used. Consequently, offset estimation must be performed online and as frequent and accurate as possible to precisely extract saliency information. This paper proposes a novel identification method that estimates offset magnitudes and angles under one active inverter state excitations that modulate the fundamental wave voltage. Measurements performed on a traction induction machine prove the method accuracy.