Empowering talkative power technology in wireless power transfer with machine learning

Abstract

In this article, a model is proposed for talkative power (TP) implemented wireless power transfer circuit for constant power load applications that concurrently transmit power and information through a shared channel. The innovative model considers critical factors such as load variations and limited receiver knowledge regarding transmitter component values, which are vital for the seamless operation of TP technology in charging devices with varying loads that are oblivious to the parameters of the transmitter. An efficient transmitter and an optimal receiver are introduced in the framework. The transmitter's design revolves around encoding data into the phase angle between the arm bridges, chosen to optimize energy efficiency. Notably, the model employs a buck-boost converter whose duty cycle is dynamically adjusted by the controller to accommodate changes in information or transmission phases, ensuring smooth system operation. Maximum likelihood detection methods are rendered impractical at the receiver due to the model's assumptions. To address this challenge, a neural network is implemented as a supervised learning classifier to extract information from the output voltage ripple. The simulations utilize the Speedgoat Real-Time Target Machine in conjunction with Simulink RealTime highlighting the effectiveness of the efficient transmitter and optimal receiver demonstrating the robustness of the model.