A data-driven optimization framework for high-frequency three-port DC–DC converters is presented. The approach is enabled through Monte-Carlo design exploration, where electrical, magnetic, and thermal constraints are evaluated to generate valid design samples across a wide operating range. These samples are then used to train regression-based AI models, which are subsequently employed within evolutionary optimization algorithms to identify operating points and component selections that maximize efficiency, power density, and overall converter performance. The method is structured to support MHz-range operation, incorporate practical magnetic-component modeling, and integrate real hardware feedback for model refinement. The resulting AI-assisted workflow enables systematic co-optimization of topology parameters, switching conditions, and magnetic design for advanced multi-port power-conversion applications.
