Convergence Analysis of Extended Kalman Filter for Sensorless Control of Induction Motor

This paper deals with convergence analysis of the Extended Kalman Filters (EKF) for sensorless motion control systems with induction motor. An EKF is tuned according to a six–order discrete–time model of the induction motor, affected by system and measurement noises, obtained by applying a first–order Euler discretization to a six–order continuous–time model. Some properties of the discrete–time model have been explored. Among these properties it is relevant the observability property, which leads to conditions that can be directly linked with the working conditions of the machine. Starting from these properties, the convergence of the stochastic state estimation process, in mean square sense, has been shown. The convergence is also explored with reference to the difference between the samples of the state of the continuous–time model and that estimated by the EKF. The results achieved theoretically have been also validated by means of experimental tests carried out on an IM prototype.