Data-Driven Model-Free Adaptive Control for Power Converter Under Multiscenarios in Microgrids

Abstract

Microgrids require efficient control schemes to achieve high penetration of power converters. Model-based control is widely used in converters due to its ease, but it poses challenges when dealing with inaccurate models, critical loads, and multidisturbances. To fix them, this work proposes a data-driven model-free adaptive control (MFAC) for a 3L-NPC power converter system to realize robust, high-quality current and voltage control without any system parameters. Specifically, an MFAC is a potential method for estimating the system dynamics by designing adaptive laws derived intuitively from Lyapunov theory to regulate current and voltage, guaranteeing self-adaptability to unmodeled dynamics, model variations, parameter mismatches, and disturbances. Striving for easier access to adaptive law gains, a particle swarm optimization algorithm is embedded in MFAC to automatically determine them utilizing the fitness function defined by the voltage and current. Experimental data confirm that the proposal outperforms artificial neural networks-based PI and super-twisting algorithm-based model control schemes under multiscenarios in terms of transient-/steady-state, grid current harmonic distortions, and robustness.