A data-driven modulation strategy for a triple-active-bridge (TAB) DC-DC converter is developed to directly minimize the total semiconductor losses, including both conduction and switching components, in wide-bandgap (WBG) SiC MOSFET-based systems. An offline exhaustive parameter sweep in PLECS, integrating detailed device-level SiC loss models and enforcing zero-voltage switching (ZVS) constraints, generates a high-fidelity dataset. The resulting dataset is further augmented and used to train a compact, artificial neural network (ANN) that predicts loss-optimal duty-cycles in real time, eliminating the need for large multidimensional lookup tables and computationally intensive online searches. Simulation results demonstrate consistent loss reductions of up to 23% compared to conventional phase-shift modulation across a wide operating envelope spanning 200-1000V and 2-10kW. The trained ANN requires only 870 bytes of memory and achieves mean absolute prediction error below 0.8%, enabling practical, high-efficiency control of multiport converters with minimal computational overhead and facilitating scalable implementation in embedded real-time power electronic controllers.

This paper discusses the power converter interface between the 400 VDC DC microgrid and the European threephase grid (400 VLL). Therefore, power bidirectional step-down topologies are considered in this paper. Among the various topologies, three single-stage non-isolated topologies are compared in terms of semiconductor losses, magnetic component size and heat sink size. These topologies are the seven-switch buck converter, the SWISS converter and the Y-converter. The comparative evaluation is conducted for a 10-kW converter targeting residential or small commercial applications.

Isolated multi-port converters show the merits of hosting several sources and loads with different voltage and power ratings, allowing power routing among multiple ports with high power density. However, many degrees of freedom are available for modulation, and exploiting them for optimal converter operation is challenging. This paper proposes an artificial neural network (ANN) approach that minimizes the rms ports currents of a triple active bridge (TAB) converter for the entire range of operation. The ANN is trained to determine the optimum duty-cycles for total true rms current minimization. The effectiveness of the ANN implementation is shown by considering an experimental TAB converter prototype rated 5kW.

In this paper, two-level and three-level modules for a three-phase multi-port converter (MPC) are evaluated and compared regarding the number of components and semiconductor losses. The MPC connects two medium-voltage ac (MVAC) microgrids to a common storage unit. The two-level MPC unit comprises a full-bridge converter for the ac-side and an isolated triple active bridge (TAB) dc-dc converter connecting the three ports. The full-bridge is replaced by a three-level neutral point clamped converter in the three-level solution. The comparative study was performed using PLECS simulation.