Optimal wireless power transfer to hybrid energy storage system for electric vehicles: A comparative analysis of machine learning-based model-free controllers

Abstract

Wireless charging technology for electric vehicles (EVs) is gaining popularity as a result of advancements in battery technology and government incentives. It offers convenience, efficiency, and safety and is much quicker than traditional plug-in charging. However, optimal power transfer in a wireless power transfer (WPT) system remains a challenge. To design an effective control system for hybrid energy storage system (HESS), it is important to have an accurate and reliable model of the system. But still, no model can fully represent the nonlinearity of the system. So model-free controllers are designed to operate without an explicit model, using feedback from sensors to adjust the control inputs. This article presents an LCC series-series network based WPT-HESS model, supported by literature. The HESS system consists of a supercapacitor (SC) and a battery connected to the WPT system through a bidirectional DC-DC buck-boost con verter. Three types of machine learning-based model-free controllers, i.e., an artificial neural network controller, a fuzzy logic controller, and a reinforcement learning-based deep Q-network controller, are designed for the HESS DC-DC converters aimed at track ing optimal power. During charging, the supercapacitor is given priority due to its faster charging time, and current references are generated based on the state-of-charge (SoC) of the supercapacitor and battery. The proposed controllers for WPT-HESS are simulated in MATLAB/Simulink and hardware validation is carried out by doing a hardware-in loop experiment with the C2000 Delfino™ and the MCU F28379D LaunchPad, which has a TMS320F28379D dual-core CPU and runs at 200 MHz.