Herconfigureerbare resonante DC-DC-omzetter voor een breed spannings- en vermogensbereik in DC-snellaadsystemen

This PhD research investigates reconfigurable resonant DC-DC converter architectures and design methodologies for electric vehicle DC fast-charging systems. The transition from 400 V vehicle platforms to 800 V and future kilovolt-class powertrains creates stringent requirements for converter voltage compatibility, power scalability, semiconductor voltage stress, efficiency, and design robustness. Conventional LLC resonant converters offer soft-switching capability and high efficiency near the resonant operating point, but their use in ultra-wide voltage-range applications typically requires a large switching-frequency variation, which complicates magnetic design, increases circulating current, and degrades efficiency. To address these limitations, this work first proposes a reconfigurable single-module LLC-based resonant DC-DC converter for wide-voltage EV charging. By combining a stacked half-bridge inverter, reconfigurable rectification, and three operating modes, the converter extends the achievable output-voltage range while maintaining soft-switching and a constrained switching-frequency span. The stacked structure reduces the semiconductor voltage stress to approximately half of the DC-bus voltage and provides inherent voltage self-balancing without auxiliary balancing circuitry. Experimental validation on a 500 W prototype demonstrates continuous wide-range operation, smooth mode transitions, and high efficiency over the tested voltage range. To enable power scalability, the research further develops a reconfigurable input-parallel output-series resonant converter architecture. This modular configuration decouples voltage-range extension from power scaling, reducing the burden on each individual module. A coupled-inductor-based balancing mechanism is introduced to achieve inherent current sharing and voltage self-balancing under modular operation. A quantitative design methodology is established to link resonant voltage-gain characteristics with current-sharing and voltage-balancing requirements. Experimental results from a 1 kW prototype validate wide output-voltage operation and high conversion efficiency. Finally, this thesis studies resonant-parameter design methods for wide-range LLC converters. The limitations of fundamental harmonic approximation and time-domain analysis are analyzed in terms of accuracy, computational effort, and practical design applicability. A knowledge-aware neural network framework is proposed to improve voltage-gain prediction by embedding analytical trends into a data-driven model. This enables more accurate resonant-parameter selection and reduces the need for conservative design margins that compromise efficiency. Overall, this research demonstrates that converter-level and system-level reconfiguration, combined with knowledge-aware data-driven design, provides an effective approach for achieving ultra-wide voltage compatibility, scalable power processing, voltage-stress reduction, and efficiency-oriented design in next-generation EV DC fast-charging systems.