Machine Learning-Based Prediction of Magnetic Field Patterns in Wireless Power Transfer

Wireless power transfer (WPT) has gained growing significance for supplying energy to contemporary electronic gadgets and electric vehicles, making accurate prediction of magnetic field distribution essential during the design phase. Conventional techniques for estimating the magnetic field in WPT systems often suffer from slow processing times, limited precision, and inefficiencies in the modeling process. With the rapid progress in Artificial Intelligence (AI), its capability to conduct swift, precise computations and optimize complex variables has emerged as a promising solution to enhance WPT design. This paper introduces a novel method for forecasting the magnetic field distribution in WPT systems by leveraging machine learning (ML). A customized U-Net model was used to learn from coil input parameters and produce spatial magnetic field distribution maps as outputs. These maps are crucial for determining safe design parameters that meet magnetic field safety standards. Experimental findings show that the proposed approach delivers highly efficient predictions, requiring on average only 4.1 milliseconds for computation and achieving a normalized prediction error of 0.0028. This marks a notable advancement compared to traditional simulation software such as COMSOL, which takes about 21.5 seconds for equivalent calculations.