Machine Learning-based Power Control Technique for Wireless Power Transfer

Abstract

This paper proposes a power control technique to overcome the efficiency drop caused by impedance variation for wireless power transfer (WPT). The technique is designed to control power without an impedance converter in the power receiver. Systems using BLE have very slow data feedback of 250 ms. In this case, when the existing control is used, oscillation occurs at the control stage or the control speed becomes slow. Based on the data provided intermittently by the receiver, we propose a technique for controlling power based on machine learning to implement fast control at the transmitter. The data for the proposed model are measured based on the DC voltage and current in the transceiver. The proposed technique is verified using a 6.78MHz magnetic resonance (MR) WPT transceiver. Bluetooth low energy (BLE) technology is used for communication. The proposed model is trained based on experimental data and controls power with 93.3% accuracy.