Robust Control of Quad-Active-Bridge Converter Employing Zeroing Neural Networks and Super-Twisting Sliding Mode Control

Abstract

This article proposes a robust and efficient control framework for quad-active-bridge (QAB) converters based on the integration of super-twisting sliding mode control (STSMC) and zeroing neural networks (ZNNs). An outer-loop STSMC is designed to regulate port voltages and generate corrected power references with finite-time convergence and strong robustness against load disturbances, parameter uncertainties, and measurement noise. To address the inherent multiport coupling of the QAB topology, an inner-loop ZNN-based dynamic solver is developed to compute the inverse power–phase Jacobian in real time, enabling accurate and noise-tolerant power decoupling without iterative optimization or explicit matrix inversion. The proposed STSMC–ZNN architecture achieves fast transient response, precise multiport power sharing, and reduced circulating and rms currents, leading to improved conversion efficiency. Comprehensive simulation and experimental results under various operating conditions, including unequal port power demands and parameter variations, validate the effectiveness and robustness of the proposed control strategy, achieving a 68.3% reduction in steady-state power error, a 16.9% reduction in peak port current, and a 6.47% reduction in rms power ripple compared with a conventional Gauss-Seidel-based decoupling method, while maintaining a comparable settling time of approximately 30 ms.