A Robust Model Predictive Control for Dual Active Bridge Converters Based on Super-Twisting Observer and Sensitivity Analysis

Although model predictive control (MPC) offers fast dynamic responses and easier ways to deal with multiple control objectives, its weak robustness is still a crucial issue when there are model or passive component inaccuracies. The previous works either entail significant computational burden or suffer from limited dynamic performance. On the other hand, to pursue low current stress, the existing methods mainly employ Lagrange multiplier methods (LMMs) to calculate optimal solutions. However, the potentially nonconvex feasible region and overlapped power range may compromise its effects. Therefore, this article proposes a robust model predictive control with current stress-optimized (RMPC-CSO) method, which entitles satisfying dynamics and strong robustness using less computational burden while achieving low current stress. Herein, a novel derivation based on piecewise gradient optimization (PGO) is proposed to solve solutions more straightforwardly and provably. Subsequently, a sensitivity analysis is proposed to quantitatively reveal the influence of parameter mismatches. Based on this, a super-twisting observer (STO)-based system is used to observe and compensate the disturbance of DABs. Meanwhile, a guideline for parameter selection is given, while the digital implementation steps are discussed. Finally, the experimental comparisons with other schemes verify its effectiveness and superiority.