AI-Based Inductive Wireless Charging for Electric Vehicles Under Different Circumstances

Abstract

In recent years, wireless power transfer (WPT) has gained substantial interest due to its capability to enhance convenience in different applications. Specifically, WPT for Electric Vehicles (EVs) has emerged as a crucial element in accelerating the growth of the EV industry. However, coil design optimization remains a challenging task due to the complexity of electromagnetic analysis. To address this, the Ansys Maxwell simulator is used to build various coil structures. Although Ansys Maxwell provides a practical analysis with high-accuracy results, its computing time is prohibitively long, averaging 45 minutes per design. To overcome this limitation, a machine learning (ML) model is proposed to estimate the parameters required for calculating coil efficiency under different circumstances. A Random Forest (RF) model was trained using a dataset of 1,200 samples collected from the Ansys Maxwell simulator. This machine learning model shows an effective solution, achieving an accuracy of 98% and significant reduction in computation time. Moreover, the model evaluates the effects of the vertical air gap, horizontal misalignment, spacing between coil turns, and the number of coil turns on the WPT efficiency. Additionally, a search algorithm is used to identify the top 5 receiver designs based on the transmitter coil parameters and its positioning that yield the highest efficiency. The results show that adjusting the spacing between turns can reduce the total wire length by 43.4%, leading to a significant reduction in system cost, size, weight, and manufacturing complexity, all while maintaining high efficiency.