Title :
Design Method and Application of Wavelet Neural Network for Direct Torque Control System
Author :
Guangbin, Ding ; Peilin, Pang
Author_Institution :
Hebei Univ. of Eng., Handan
Abstract :
A torque ripple reduction technique based on wavelet network of direct torque control running in low-speed is presented. The wavelet network can accurately localizes the characteristics of a signal both in the time and frequency domains, and the occurring instants of the signal change can be identified by the multi-scale representation of the signal.Taking advantage of complex wavelet transform, combined information can be obtained from both the magnitudes and arguments of complex wavelet transform coefficients to extract the desired feature of the transient signal. The input nodes of the WN are the stator current error and the change in the stator current error and the output node of the WN is the stator resistance error. The synthesized method of recursive orthogonal least squares algorithm and improved Givens rotation is used to fulfill the wavelet network structure initialization and parameter identification, and then the accurate stator flux vector and electromagnetic torque are acquired by means of state estimator, optimizing the inverter control strategy. The simulation results show effectiveness of the proposed control algorithm.
Keywords :
feature extraction; induction motors; least squares approximations; machine control; neural nets; signal representation; stators; torque control; wavelet transforms; complex wavelet transform; direct torque control system; dynamic system identification; electromagnetic torque; feature extraction; frequency domain; improved Givens rotation; induction motor; inverter control strategy; multiscale signal representation; parameter identification; recursive orthogonal least squares algorithm; signal change; state estimator; stator current error; stator flux vector; stator resistance error; time domain; torque ripple reduction; transient signal; wavelet network structure initialization; wavelet neural network; Data mining; Design methodology; Feature extraction; Frequency domain analysis; Neural networks; Signal processing; Stators; Torque control; Wavelet domain; Wavelet transforms; Induction motor; combined information; complex wavelet transform; direct torque control; dynamic system identification; wavelet network;
Conference_Titel :
Electronic Measurement and Instruments, 2007. ICEMI '07. 8th International Conference on
Conference_Location :
Xi´an
Print_ISBN :
978-1-4244-1136-8
Electronic_ISBN :
978-1-4244-1136-8
DOI :
10.1109/ICEMI.2007.4351044
