A Reinforcement Learning-Based Data-Driven Voltage Regulator for Wireless Chargers of Electric Vehicles

As a critical component in the concept of smart cities, autonomous vehicles are under extensive investigation. The need for intelligent fueling without human assistance stimulates the development of wireless charging. There are, however, several issues to be addressed to enhance the efficiency and reliability of charging. The vehicle battery exhibits varying resistance during the charging process. Additionally, the alignment and gap may change with external positioning, which affects the coupling coefficient of the transmitter and receiver coils. Thus, a data-driven control scheme is desired to tackle these environmental uncertainties. This paper adopts the control scheme of embedding a Buck converter at the receiver side. Different from state-of-the-art literature, a reinforcement learning-based data-driven control approach is employed to regulate the charging voltage. Stable charging voltage is attained regardless of the knowledge of the coupling coefficient, load variations, or component values. System simulations in Simulink have proved the effectiveness of the proposed control method.