Realization of WPT System with Incorporated Metamaterial: Employment of Machine Learning into the Design Process

Abstract

This project presents the design, implementation, and experimental validation of a compact wireless power transfer (WPT) system operating at 433 MHz, incorporating metamaterial (MTM) enhancement and a machine learning (ML)-assisted design pipeline. The system employs Defected Ground Structure (DGS) resonators fabricated on Rogers RO4350B substrate within a 20×20 mm footprint, addressing the challenge of achieving high power transfer efficiency (PTE) in miniaturized IoT sensor applications. An ANN-Ensemble machine learning model was trained on a parametric sweep dataset generated in CST Studio Suite to predict resonator geometries, achieving a test R² of 0.916. A high-side baseline frequency strategy was adopted, targeting approximately 1.5 GHz to ensure fabrication robustness, with resonance subsequently tuned to 433.2 MHz via a 3.5 pF external capacitor. Six reflective metamaterial placement configurations were systematically evaluated, with the Bottom + Top arrangement (Case 5) yielding the optimal result of 64% PTE at 20 mm transmission distance, representing a 23.3 percentage point improvement over the unassisted baseline of 40.7%. All six project requirements were satisfied, including size, efficiency, regulatory compliance, ML accuracy, component availability, and material safety. The findings demonstrate that compact, high-efficiency near-field WPT systems can be realized through the combined application of ML-driven inverse design and metamaterial field concentration, with potential applications in battery-free IoT sensors and implantable medical devices.