An Unscented Kalman Filter Design on Rotating Reference Frame For Induction Machines

Abstract

The need for high performance and high control accuracy in electric motors has made modern motor control techniques popular. Modern control techniques require position sensors integrated in the electric motor. An alternative method to sensor feedback systems is the cost-effective prediction methods that offer solutions within acceptable limits. In this study, a new observer model is proposed for induction machines. In the design of the observer, Unscented Kalman Filter is used. In the Unscented Kalman Filter design, the rotor position information is defined as a state variable within the mathematical model. Therefor the observer model has been developed without the need for additional sensor costs. As an observer model output, machine internal parameters such as rotor fluxes which cannot be measured directly and rotor speed can be estimated without requiring a sensor by using this method.