A novel control method for dynamic performance improvement of induction motor using neural network

In the conventional direct torque control system, both of the torque and stator flux errors between reference and estimated values are directly compared, and the appropriate voltage vector is produced by a switching table. On basis of quick torque response and robustness against parameter variation, the direct torque control system has been widely used in the industrial production. To improve its dynamic performance, a novel approach using wavelet network is introduced. The wavelet function is adopted as the basic function of neural network taking advantage of wavelet transform behaving good localization property in both time and frequency space and multi-scale property. The wavelet transform can accurately detect and localize signal characteristic in time-frequency domains. By means of improved training algorithm, the parameters of wavelet network are adjusted offline to minimize the approximation error. The simulation results prove that the proposed method can efficiently reduce the torque ripple and current ripple, optimizing system control strategy.