HOSMD and neural network based adaptive super-twisting sliding mode control for permanent magnet synchronous generators

Abstract

To solve the maximum power tracking (MPT) control problem of the direct-driven wind power system, a super-twisting integral sliding mode controller with adaptive parameter estimation is designed. A high-order sliding mode differentiator is introduced as the virtual control variate filter, which solves the difficulty of obtaining the derivative of the control variate and the controller saturation in the nonlinear system with disturbances. Since the speed loop of permanent magnet synchronous generator (PMSG) is susceptible to disturbances, a radial basis function neural network (RBFNN) approximator and its adaptive algorithm are proposed to observe the unmodeled part of the system and external disturbances. In addition, the super-twisting algorithm is introduced to improve the robust performance. An improved adaptive parameter estimation algorithm is used to obtain real-time estimated values in the circumstances of uncertain stator parameters and parameters perturbation during operation, which enhances control accuracy and reduces undesirable chatting. The errors of RBFNN approximation and parameter estimation are taken into Lyapunov functions to guarantee the stability of the whole system. The effect of the proposed scheme is verified as compared to the adaptive backstepping terminal and new reaching law sliding mode controllers.