State Estimation Of Pmsm Using Unscented Kalman Filter (Ukf)

Abstract

In this paper a sensorless control is modeled for permanent magnet synchronous motor utilizing UKF in order to estimate the position of rotor and speed in accurate manner. In this thesis we have chosen PMSM over induction motors and DC motors due to its high speed operation, longer life time, high power density, higher efficiency and simple in construction. In industrial applications, we assimilate/integrate different sensors for control of PMSM motors in estimating the rotor position and speed. Generally we use Encoders and Hall sensors to measure the rotor angle increasing the complexity and burden on the system which in succession increases the overall cost of the drive. This control method circumvents the use of sensors thereby reducing its complexity, cost and load on the system.

            The project presents speed or position sensorless control based on Kalman Filter (KF), Extended Kalman Filter (EKF), and Unscented Kalman Filter (UKF) method, implemented using MATLAB code, Kalman Filter(KF) is a recursive Filter applied to linear – dynamic systems. It estimates the state of the system by considering a series of measurements but the output generated will have considerable errors in measurement and process noise. Kalman Filter is preferred for linear systems but PMSM is a non linear model. These drawbacks can be sorted out with use of Extended Kalman Filter (EKF), a modified and non linearized version of Kalman Filter (KF). In EKF, the non linearized relationships can be linearized by obtaining the partial derivatives of noises with the use of Taylor Series. EKF will produce some noise in speed estimation and this process involves calculation of Jacobian matrices which consumes more time. These liabilities are rectified with Unscented Kalman Filter (UKF). It is derivative free filter which adopts a minimum number of sample points called Sigma points for accurate estimation.