A Novel Observer for Induction Motors, with an Application to Soft Starters

This paper proposes a fundamental-model-based observer to estimate the state and load torque of an induction motor. It relies on a novel representation of the motor model, namely a cascade of two linear-time varying systems with known input terms, to which a "true" Kalman filter is applicable. The interest is that the global convergence of the observer can be guaranteed, moreover without any two-time-scale assumption. The good behavior of the observer is illustrated on numerical and experimental data of a motor startup with a soft starter.