ML-Based Fully-Numerical Design Method for Load-Independent Class-EF WPT Systems
Naoki Fukuda, Yutaro Komiyama, Wenqi Zhu, Yinchen Xie, Ayano Komanaka, Akihiro Konishi, Kien Nguyen, and Hiroo Sekiya
IEEE Transactions on Circuits and Systems I: Regular Papers, June, 2025. [pdf document]

<Abstract>

This paper proposes a machine learning (ML)-based design for a load-independent (LI) wireless power transfer (WPT) system. The proposed design strategy formulates the WPT design as an optimization problem and solves it by fully-numerical computation to achieve high-frequency LI operation. A multi-objective evaluation function is given to achieve the LI operation, which evaluates output voltage, power-delivery efficiency, and total harmonic distortion. The class-EF WPT system designed using the proposed method achieved narrower output voltage variations and higher power-delivery efficiency than the analysis-based design system. Additionally, the proposed design algorithm identifies parameter sets that were never found through the analysis-based design. Namely, only a small amount of current flows through the previous harmonic resonant filter of the class-EF inverter. These results demonstrate the potential of ML-based design in the field of power electronics.