Efficient TR estimation in field coordinates for induction motors

It is well known that detecting of the proper value for the parameter T_R (rotor time constant) is crucial for calculating the correct flux of induction motor drives. Several methods have been proposed to identify or adapt T_R from the voltage, current and speed of the drive (measuring reactive power, correlation methods, observers, Kalman and extended Kalman filters). The latter Kalman filter methods often estimate currents in the stator frame which can easily be compared to the measured currents. This paper presents a suitable control structure and an extended Kalman filter for T_R estimation in field coordinates. The main problem is that, if T_R is not correct, the field coordinate system will not be aligned with the flux. In this case, a correct transformation of the measured currents into the field coordinate system is impossible. Although these currents are not correct, it turns out that the Kalman filter can reliably detect whether the true T_R is lower or greater than the model parameter. The calculation is very efficient due to the low three order filter which is linear in the estimates values. The approach has the potential to be one of the fastest available methods for T_R estimation. The algorithm does not require additional transformations or special signals. It works well under all operating conditions whenever the torque reference is nonzero