Sensorless position estimation from PWM-induced transient excitation in induction machines

This paper investigates the feasibility of sensorless position estimation from PWM-induced transient excitation sequences, present during inverter-driven operation of induction machines (IM) with standard space vector PWM (SVPWM). The applied method exploits existing machine saliencies and their correlation to the current slopes measured in response to a voltage pulse at the machine terminals. Determination of this quite linear current response and its position dependent slope modulation allows for extraction of the flux/rotor position information. However, an initial transient oscillation process delays the first possible measurement instant, what imposes minimum pulse duration and limits the possible operational range of SVPWM-integrated transient excitation for sensorless methods. Previous publications showed, that by application of current signal high frequency sampling it is possible to estimate the current derivative before these oscillations decay, thus further reducing the minimum pulse duration requirement. The paper at hand contributes an analysis of different signal processing algorithms and their capabilities when used for this purpose. The investigations include: mean value evaluation (MVE), Theil-Sen regression (TSR), least-squares regression (LSR) and maximum-likelihood estimation (MLE). In a first step an offline comparison is carried out with previously sampled measurement data. The second part features an online evaluation to verify the possibilities for SVPWM-integration.