The growing deployment of grid-connected voltage source inverters (GC-VSIs) in the smart grids poses some significant issues that include harmonic distortion, grid impedance fluctuations, nonlinear loading, and external disturbances. Traditional model predictive control (MPC) is known to be accurate in tracking but computationally expensive, and neural network-based methods are known to be fast but not substantiated by formal control. This paper suggests a strictly built hybrid control with the introduction of an offline MPC trajectory optimization, real-time Recurrent Neural Network (RNN) implementation, and an Adaptive Barrier-Condition Super-Twisting Sliding Mode Controller (ABC-STSMC) to strengthen the robustness. Training to the RNN utilizes MPC generated optimal control data and allows a corresponding reduction in computational complexity when implemented online. The ABC-STSMC layer achieves convergent behavior to finite time, provides a solution to chattering and guarantees stability in the presence of nonlinear and uncertain grid conditions. A Lyapunov analysis provides a bounded error criteria of the tracking error and convergence of the sliding surface. The controller parameters are optimally adjusted with an Improved Grey Wolf Optimization (IGWO) algorithm. Under weak-grid conditions, unbalanced loads, and harmonically distorted grid voltages extensive simulations and Hardware-in-the-Loop experiments indicate the high harmonic mitigation, high dynamic response, and minimized Total Harmonic Distortion (THD) of the ABC-STSMC compared to standalone MPC and RNN controllers. The suggested hybrid system offers a computationally efficient and powerful solution to sophisticated inverter control in current smart grid systems.
Zeb et al. (Mon,) studied this question.