Joint State and Parameter Estimation of Squirrel Cage Induction Motor – A System Identification Approach using EM based Extended Kalman Filter

This paper deals with the recursive optimum estimation of rotor resistance, inductance and stator resistance of induction motor. The estimation of parameter and states in the presence of system noise is achieved using EKF, which takes in to account measurement and modelling inaccuracies. A major limitation in the parameter estimation using EKF is that its optimality is dependent on the choice of the right covariance matrices. In this paper the EM(Expectation Maximization) algorithm along with the Kalman smoothing is used to obtain the initial values of process and measurement covariance matrices. The Vas model of the induction motor is used for simulation and the rotor inductance, rotor resistance and stator resistance are considered as parameters as well as states. For the estimation of these parameters, the state vector is augmented with these parameters to be estimated. The parameters used for simulation is that of a three phase, 2.2 kw, 400V, 50 Hz cage induction motor. Matlab is used to simulate the system and from the results it is observed that the parameter estimates converges to true values with a reasonable degree of accuracy. General Terms Parameter estimation, System Identification, Expectation Maximisation