Paper Title
TOPOLOGY GENERATION METHOD OF SINGLE-SWITCH DC-DC CONVERTERS BASED ON GRAPH VAE

Abstract
Single-switch DC-DC converters (SSCs) are favored in various applications for their simple configuration and ease of control. Traditional methods for deriving SSC topologies almost follow the paradigm where humans learn expertise from existing topologies and then use reasoning to discover new ones. In these traditional methods, human learning ability may limit the diversity of derived results. This paper introduces a novel AI-learning method to address this limitation. In the proposed method, an AI model based on graph variational auto-encoders (GraphVAE) is developed to learn topology generative rules from both existing SSCs topologies and human expertise. Since the proposed method integrates the learning ability of AI and humans, it has the potential to derive more comprehensive topologies. To verify the effectiveness of the proposed method, we collected 104 SSC topologies to train the GraphVAE, deriving 21 new SSC topologies with different performance characteristics. The derived results suggest that combining AI and human insights enhances the diversity and innovation of SSC topologies, improving design efficiency and creativity. Keywords - Topology derivation, Single-switch dc-dc converters, GraphVAE