Comparison of three Kalman filters for speed estimation of induction machines

UKF and SRUKF are two new members of the family of Kalman filters. Speed estimation of induction machines based on them is discussed in depth and compared with the EKF from every side. The effect of the sampling time and the parameters of the filters upon the speed estimation performances are analyzed, and the various aspects of the speed estimation performances of the three Kalman filters, such as stationary error, dynamic response speed, parameter sensitivities and algorithm complexity are evaluated in detail. The simulation results show that neither the UKF nor the SRUKF can replace the EKF with the expected outstanding advantages for the speed estimation of induction machines. There are two main reasons for this, one is that the order of induction machines is relatively high, this makes the algorithm complexity of them increasing greatly, the other is that the time constant of the speed of induction machines is relatively small, thus a small sampling time must be chosen, under this limitation a little more accurate state estimation of the UKF or the SRUKF does not give the speed estimation performances any essential improvement. Simulation and experimental results verify the conclusion that the EKF is still the most efficient and feasible algorithm for speed estimation of induction machines overall.