A novel LLM-AI based automation framework for high frequency wireless power transfer design

Abstract

Designing high frequency wireless power transfer (WPT) systems typically involve complex modeling, simulation, and prototyping steps that demand significant time and expertise. In this study, we adopted and extended a previously proposed large language model (LLM) based design framework to accelerate design process. Seven customized generative pretrain transformers (GPT) based agents were developed using specific design guidance in WPT design to improve the accuracy of the generated outputs. A total of seven generated designs proposed by each agent were evaluated using the same design prompt. The generated designs were compared against four mathematical models based on design accuracy, completeness, and design time. The most accurate design was validated using PSIM and 3D Ansys simulation. The results demonstrate the potential of LLM-driven workflows to significantly reduce design effort and time while maintaining high reliability in WPT system development.