Robust Nonlinear Control of a 3×2 Stacked Multicellular Inverter Using Neural Network-Based Regulation

Abstract

The increasing demand for high-performance power converters has led to the development of advanced control strategies that enhance system robustness against load variations. This paper presents a novel control approach for a 3×2 stacked multicellular inverter, integrating nonlinear feedback decoupling control with neural network - based regulators. Unlike conventional Proportional-Integral (PI) controllers, which struggle with parameter variations and dynamic uncertainties, the proposed neural network (NN) controller adapts in real-time to maintain system stability and optimal performance. The NN is trained to approximate the nonlinear control law while compensating for parametric variations and external disturbances. A comparative analysis between the NN-based control and the traditional PI approach is conducted using MATLAB / Simulink simulations, evaluating response time, harmonic distortion (THD), and robustness against load changes. The results demonstrate that the NN-based controller significantly improves system adaptability, robustness and voltage regulation, making it a promising solution for modern power conversion systems.