Parameter estimation in induction motors: a comparison between the PE and the TS paradigm

Abstract In this paper, we consider the problem of estimating unknown parameters of the model of an induction motor in a sensorless environment. A common practice consists in performing estimation through a series of bench tests, by letting the motor operate in steady-state. This way of proceeding, however, does not take into account the high sensitivity of some parameters to the motor operating condition. The issue, then, is to set-up an automatic estimator, able to to supply reliable estimates of the parameters from measurements taken from the actual functioning of the motor. Two different estimation paradigms are compared, namely the Prediction Error (PE) paradigm, which has become a standard in the practice of system identification, and the recently introduced Two-Stage (TS) paradigm. Advantages and drawbacks of such methods in the context of induction motors are spotted out by means of simulation experiments. It turns out that the TS method may offer a valid alternative to the PE method.