Resumo:
Wireless Power Transfer has emerged as a promising solution for contactless battery charging in small electric vehicles, offering enhanced safety, convenience, and system reliability. This work presents the design, modeling, control, and experimental validation of a magnetic resonant wireless power transfer system based on a high-frequency Class-DE resonant inverter for charging lithium-ion batteries. A comprehensive theoretical analysis is carried out considering series, parallel, and series-parallel resonant compensation topologies, enabling efficient energy transfer under varying coupling conditions caused by coil misalignment and distance variation. A novel digital control strategy based on an ADALINE neural network combined with an autoregressive exogenous (ARX) model is proposed to regulate the inverter switching instants and maintain stable output voltage despite load and coupling fluctuations. The control approach is derived from a frequency-domain band-pass filter model of the resonant tank and implemented using a digital signal processing platform. The proposed controller demonstrates fast dynamic response, reduced harmonic distortion, and improved robustness compared to conventional open-loop operation. A laboratory-scale prototype operating in the MHz range is developed to validate the proposed methodology. Experimental results confirm soft-switching operation with zero-voltage and zero-current switching, reduced thermal stress on GaN MOSFETs, and enhanced power transmission efficiency across different coil separations. System performance is evaluated in both time and frequency domains, including harmonic analysis and thermal characterization. The results demonstrate that the proposed Class-DE-based MR-WPT system with ADALINE-based ARX control is a viable and efficient solution for wireless charging applications in light electric mobility systems.