Adaptive predictive control system with disturbance compensation based on self-recurrent wavelet neural network

In permanent magnet synchronous motor (PMSM) speed drive system, the uncertainties including model uncertainty and non-model uncertainty influence the system performance. When serious uncertainty exists, system performance may be worse or even unstable. An adaptive Smith predictive controller based on wavelet neural network (WNNSP) is proposed in this paper to relieve the affection brought by model parameter uncertainty and disturbance. Two wavelet networks are introduced, one is used for feed forward compensating disturbance such as load changing, and the other is utilized to eliminate the influence of model uncertainties by compensating the model error and obtain a accurate predictive output. These solutions efficiently restrain the effect of disturbances and provide good static and dynamic performance so that the system stability and robustness are guaranteed. Performance comparisons among PI controller, PI controller with Smith Predictor and proposed controller are simulated and the results prove the validity of the proposed control scheme.