Application of genetic algorithms in EKF for speed estimation of an induction motor

Abstract

Genetic algorithm (GA) is applied in this paper to optimize parameters of the extended Kalman filter (EKF) in a speed-senserless field-oriented controller (FOC) system. The main parameters of EKF are the covariance matrics Q and R, which are bound respectively to the state and measurement noises. As for speed-sensorless FOC system, the convergence and precision of both rotor speed and flux estimation depend on the accuracy of the models of system noise and measurement noise, i.e. Q and R. A GA training simulation system of optimum parameters of EKF is given and the simulation results show the efficiency and rationality of the algorithm.