An RNN-EKF Observer for Time Delay of Large-Scale Motor Control System

For a large-scale motor control system with non-negligible time delay, a Recursive Neural Network- Extended Kalman Filter (RNN-EKF) observer is proposed to estimate the flux and reduce the disturbance caused by the time delay. In this paper, the motor model is designed to simulate the actual operating conditions: the motor resistance varies with temperature and time delay is considered in the current loop. The proposed observer combines the RNN with EKF algorithm, aiming at reducing the error between the observed flux and the actual flux by training weight coefficients of the neural network. Simulation results show that the proposed observer owns good static and transient control performance. In addition, the effect of the unknown time delay is effectively reduced.