Fast Moving Horizon Estimator for Induction Motor Sensorless Control

The paper deals with a systematic model based approach to state estimation of an induction machine. A moving horizon estimation algorithm is adopted, which allows to employ fast controllers as model predictive control. The state estimation problem is nonlinear, thus the linearized mathematical model of the machine is considered in order to solve a quadratic optimization problem under linear equality constraints. This problem is solved real-time by computing the direct solution of the equivalent Karush Khun Tucker system with a control sampling rate of 10kHz. Motor speed, which is part of the estimated state, is finally used for a motion sensorless control. The feasibility of the proposed approach is verified by means of experiments on a state-of-the-art embedded control platform. Moving horizon estimator performances are analyzed and discussed in a wide variety of operations, stressing stability and accuracy of the estimated rotor speed.