International Review of Electrical Engineering, cilt.6, sa.2, ss.706-718, 2011 (Scopus)
This study presents a nonlinear state estimator based on recurrent neural network (RNN) which uses a PI Elman neural network (PI-ENN) structure for state estimation of a squirrel-cage induction motor. Proposed algorithm uses the measurements of the stator currents and rotor angular speed. It learns the dynamic behavior of the state observer from these measurements, through the prediction error minimization. Since the stator currents are available for measurement it may appear that the stator current estimates are redundant but these estimates are actually filtered version of the measured stator currents. We also include these variables to the state vector for completeness of the algorithm and to check the results. In order to observe the performance of the proposed estimation algorithm for different operation conditions the squirrel- cage induction motor was fed with various supply voltages, such as sinusoidal, six-steps, and pulse width modulation (PWM) waveforms. Estimation results show that the proposed algorithm performs better than for the extended Kalman filter (EKF) in terms of accuracy and convergence speed. © 2011 Praise Worthy Prize S.r.l. - All rights reserved.