This paper presents a MIMO-bidirectional converter designed for aerospace applications and an optimized model using Neural-Network(NN) techniques. The converter comprises a five-level T-inverter, a two-cell cascaded-inverter, and a high-frequency transformer. It addresses the newer High DC voltages proposed in the literature and offers ground isolation between the 800[V] High Voltage DC grid, the 28 [V] low-voltage grid, and loads. The converter also supports extra modules for bidirectional operation. The paper's second contribution is training a neural network to optimize the system's behavior, including switching angles, output power, bi-directionality, and noise emission. The study includes simulations and a case-study-experimental result using a low-voltage/low-power prototype. Additional experimental results will be included in the final manuscript to strengthen the draft further.
