Adaptive stator and rotor resistance identification for induction machine without rotational transducer

This paper introduces a new method of stator resistance identification based on wavelet network in order to improve the low-speed dynamic performance of induction motor in direct torque control. Because of the advantage of wavelet transform, the desired feature of the transient signal can be extracted conveniently from both the magnitudes and arguments of wavelet coefficients and the control precision for direct torque control can be increased. The input node of wavelet network is the stator current error and the change in the stator current error. The output node of the wavelet network is the stator resistance error. The wavelet network structure and parameter identification are fulfilled by the evolutionary algorithm. Then in order to optimize the inverter control strategy, the accurate stator flux vector and electromagnetic torque are acquired by means of state estimator. The experiment results show that this new method can efficiently reduce the torque ripple and current ripple and is better than that of the back-propagation neural network.