Moving horizon estimation for induction motors

In this paper we present a systematic model based approach to state and parameter estimation for the induction machine. We use moving horizon estimation (MHE), an optimization based scheme that yields excellent performance and can be used with aggressive controllers such as model predictive controllers. The past measurements within a given horizon are combined with an a priori estimate based on the induction machine model. Under mild assumptions, this yields a maximum-likelihood estimate of the states and parameters over the horizon. The resulting optimization problem is solved using the Generalized Gauss-Newton method. A real-time iteration approach can be used to significantly reduce execution and response times. Simulation results indicate superior performance of MHE over established methods such as model reference adaptive system (MRAS) or Extended Kalman Filter (EKF). Real-time feasibility of the proposed approach up to 3.5 kHz sampling rate is demonstrated by experiments on a state-of-the-art embedded control platform.