Robust Quasi-integral sliding mode control combined with an advanced neural network algorithm for PMSG-based wind energy conversion systems in noisy environment

Abstract

With the global increase of wind energy integration into modern electrical power systems worldwide, Wind Energy Conversion Systems (WECS) are nowadays increasingly expected to contribute actively to grid operations through advanced control strategies powered by novel algorithms of artificial intelligence. To address this need and ensure high-quality power injection from a variable speed wind turbine into the grid under a randomly noisy environment, this paper introduces a novel control scheme that combines robust Quasi-Integral Sliding Mode Control (QISMC) with an improved Radial Basis Function (RBF) technique for more accurate modeling and regulation of a Permanent Magnet Synchronous Generator (PMSG)-based WECS. In this configuration, the PMSG is interfaced with the grid via a back-to-back converter, which is connected to the distribution network via a DC-link capacitor, ensuring efficient energy transfer. Additionally, our proposed method integrates an RBF neural network (RBFNN) to estimate the uncertain dynamics of the wind turbine and external disturbances of the system, enabling faster convergence, reduced switching gains, and zero steady-state errors. Within this setup, the Machine-Side Converter (MSC) is designed to control the wind turbine rotor speed and the PMSG current vector, ensuring efficient Maximum Power Point Tracking (MPPT) and efficient active power transfer. Meanwhile, the Grid-Side Converter (GSC) stabilizes the DC-link voltage and manages the injected active and reactive power into the grid network. A systematic comparative study is carried out under system uncertainties and stochastic noise conditions, wherein the proposed QISMC-RBF is systematically benchmarked against conventional control methods published in the literature. Finally, simulation results obtained using the MATLAB/Simulink environment demonstrate the superior performance of the proposed control approach, with a rotor-speed convergence time of about 3 s, a DC-link settling time of about 5 s, a 78% reduction in tracking error compared to QISMC, and more than 96% chattering reduction.