Identification of PMSM based on EKF and elman neural network

Abstract

Permanent magnet synchronous motor (PMSM) is a complex plant to control, due to its high nonlinearity and strong coupling. At the same time, the variations of motor parameters make this problem more serious. So, parameter identification of PMSM seems to be important for the double closed-loop vector control system. To solve this problem, a new method combining Elman neural network(Elman NN) and modified extended kalman filter(EKF) is studied in this paper. The approach of identifying R <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">s</inf> , Ψ <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">d</inf> and Ψ <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">q</inf> is discussed. Simlation results show that it has lots of advantages such as high precision, fast convergence and excellent generalization ability and it is suitable for variable speed and variable load disturbance, even more complex circumstance.