Model-free predictive control strategy for three-level dual active bridge converter based on novel sliding mode observer

For three-level dual-active-bridge (3L-DAB) converters, conventional control strategies are susceptible to performance degradation due to modeling inaccuracies, parameter mismatches, and external disturbances. This paper proposes a model-free predictive control strategy based on a novel sliding-mode observer. By analyzing the converter’s switching conduction modes and operational characteristics, a first-order ultra-local model is established, containing only the control input, system output, and an aggregated disturbance term. A novel sliding-mode observer is designed to estimate the aggregated disturbance in real time, while a nonlinear saturation function is introduced to effectively suppress system chattering. Finally, the optimal phase-shift expression for the three-level DAB converter under model-free predictive control is obtained by discretizing and solving the cost function. Experimental results demonstrate that the proposed strategy reduces the system’s dependence on model parameters, enables rapid and stable output voltage tracking under parameter mismatches, and enhances both dynamic response and disturbance rejection performance.