Design of Derivativefree State Estimators for a Three Phase Induction Motor � A Comparative Study

Abstract

Particle filters are an alternative to approximate the Kalman filter for nonlinear problems. This paper intends to assess the potential of Particle Filter (PF) and its variants in the context of the state estimation problem of a three phase induction motor. The conventional Particle Filter (SIR-PF), and particle filters that employ importance sampling through proposal distributions such as Particle Filter with Extended Kalman Filter (PF-EKF) and Particle Filter with Unscented Kalman Filter (PF-UKF), which are proposed in the literature within the particle filtering framework that takes into account of the latest observational information to reduce the risk of weight degeneracy is described and the error behaviour is analyzed through Monte Carlo simulations with regard to three scenarios Viz., low speed operation, step changes in load torque and reversal of speed. Simulation results demonstrate the superior tracking performance of PF-EKF at the expense of higher computational effort over the other approaches and can be determined to be a good substitute for the UKF in terms of accuracy of the state vector estimation.