Induction motor, state estimation, extended Kalman filtering, recurrent neural networks

摘要

This study presents a recurrent neural network (RNN)-basednonlinear state estimator that uses an Elman neural network structure(ENN) for state estimation of a squirrel-cage induction motor. Theproposed algorithm only uses the measurements of the stator currentsand the rotor angular speed, and it learns the dynamic behavior of thestate observer from these measurements through prediction errorminimization.A squirrel-cage induction motor was fed from sinusoidal, 6-step, andpulse-width modulation (PWM) supply sources at different times inorder to observe the performance of the proposed estimator fordifferent operation conditions. Estimation results showed that theproposed algorithm is capable of estimating the states of an inductionmotor and performs better than extended Kalman filtering (EKF) interms of accuracy and convergence speed.