Data Regression Compensation Algorithm for Improving the Current Dynamic Performance of Grid-Tied Inverters

Abstract

Grid-tied inverter performance is degraded by its inherent nonlinear characteristics, which can be suppressed by internal model principle-based controllers. However, designing these controllers requires a tradeoff between time consumption and model uncertainties. This article proposes a data-driven approach that can eliminate the influences of nonlinearities and improve grid-tied inverter performance by compensating for low-frequency harmonics, enhancing current control accuracy and improving system stability. The proposed data-driven compensation method trains data based on the internal model principle, establishes a data regression model through lightweight offline regression, and constructs a compensation loop based on the regression model. The loop is then combined with traditional low-order control methods to improve the overall performance of the inverters through real-time compensation. A loss function is used to validate the accuracy of the data regression model, and experimental results show the effectiveness and exceptional performance of the proposed method.