Comparison of Different Signal Processing Methods for Saliency-Based Sensorless Control of Induction Machine Using Optimized Excitation Sequence

Sensorless machine control at zero or low frequencies is typically achieved through external voltage step excitation which, by means of the current responses, produces quantifiable transient inductance values, containing the machine saliencies used to determine rotor position. However, additional excitation implies increasing current ripple, losses and acoustic noise, unacceptable for certain applications. The integration of these pulses into standard SVPWM has become, consequently, a key objective especially in the last decade. Nevertheless, the quality of the computed transient inductance, traditionally done by two sample points, is mainly constrained by the need of a minimum voltage excitation amount used for the current derivative. To deal with this limitations, this paper calculates, in online operation, the current derivative through a linear regression for different excitation strategies.