Observer-based neural network dynamic surface control of single-phase LCL-filtered grid-connected inverters under parametric uncertainties and weak grid condition

Abstract

This paper focuses on the current control of single-phase LCL-filtered grid-connected inverters in the presence of parametric uncertainties and weak grid condition. Therefore, a novel neural network dynamic surface control (NNDSC) method is proposed by overcoming the problem of ‘explosion of complexity’. In addition, radial basis function neural networks (RBFNNs) are employed to approximate the system parametric uncertainties. Furthermore, by considering practical considerations, a novel state observer (SO) is designed to estimate the inverter-side current and capacitor voltage. As a result, additional current and voltage sensors are not required, which makes the implementation of the proposed approach straightforward and reliable. The origin neighbourhood convergence of estimated and tracking errors is assured through Lyapunov stability theorem and Young’s inequality. The effectiveness and performance of the proposed NNDSC+SO approach is demonstrated through MATLAB/Simpower simulations in view of the reference current changes, LCL filter parametric uncertainties and weak grid condition.