Despite the increasing improvements of the Wide Band Gap (WBG) semiconductors for power converters, the definition of an accurate losses model still results crucial for better estimating the efficiency, thus optimizing the design of a proper cooling system while minimizing the overall size of the converter itself. With the aim of improving the performance of the current system-level model for SiC-based converters, this paper proposes a losses model based on a data-driven approach able to achieve higher accuracy with a reasonable computational cost. In detail, starting from the creation of an input dataset according to datasheet information or experimental characterization, it is demonstrated that the execution of non-linear fitting operations on the energy losses characteristics can strongly improve the estimation accuracy compared to the most common linear approximations performed by the current literature and the power electronics industries. Referring to the same input dataset, a comparative analysis based on experimental results between the energy losses estimated by a common numerical tool and the proposed modeling approach is presented for highlighting the higher accuracy.
