A Novel Neuro-Wavelet Based Self-Tuned Wavelet Controller for IPM Motor Drives

Abstract

This paper presents a hybrid neuro-wavelet scheme for on-line tuning of a wavelet-based multiresolution PID (MRPID) controller in real-time for precise speed control of an interior permanent magnet synchronous motor (IPMSM) drive system under system uncertainties. In the wavelet-based MRPID controller, the discrete wavelet transform (DWT) is used to decompose the error between actual and command speeds into different frequency components at various scales. The MRPID controller parameters are tuned by the wavelet neural network (WNN) to ensure optimum performance of the drive system. The proposed neuro-wavelet based MRPID controller is trained online with adaptive learning rates in the closed-loop vector control of the IPMSM drive system. The adaptive learning rates are derived using discrete Lyapunov stability theorem so that the convergence of the tracking error is guaranteed in the closed-loop system. The performances of the proposed hybrid controller are investigated in both simulation and experiments at different dynamic operating conditions. The complete vector control scheme incorporating the proposed self-tuning MRPID controller is successfully implemented in real-time using the ds1102 digital signal processor board for the laboratory 1-hp IPM motor. The superior performances of the proposed WNN-based self-tuning MRPID controller are also validated over fixed-gain controllers.