Modified model reference adaptive observer for rotor speed and position estimation in wound rotor induction machine

Abstract

Sensorless estimation of rotor speed and position has numerous advantages in terms of maintenance, cost, robustness and cabling requirements. The existing Model Reference Adaptive System (MRAS) observer considers the cross-product of estimated and actual rotor current vectors as the error input to the PI controller. This results in a non linear plant model, necessitating the use of small signal analysis for controller design. The modified MRAS observer presented here results in a linear plant model, enabling straight-forward design of PI controller and also resulting in improved performance. Simulation and experimental results, presented on a 10-HP Wound Rotor Induction Machine (WRIM), driven by a Squirrel Cage Induction Motor (SCIM) clearly bring out the superior dynamic performance of the modified estimation scheme over the existing scheme. The problem of integrator drift is also addressed in the hardware implementation.