A Hybrid Data-Driven Approach in Magnetic Core Loss Modeling for Power Electronics Applications

The accurate magnetic core loss estimation in power electronics applications, which is crucial for designing efficient converters, is a highly challenging task. Traditional methods like the Steinmetz equation and its evolutions fail to describe various operating conditions accurately. This paper presents a hybrid approach to address these challenges, proposing a workflow that integrates machine learning concepts with the physical understanding of magnetic phenomena for accurate and fast magnetic core loss predictions under various operating conditions, offering a promising path for effective loss models in power electronics applications.