کنفرانس بین المللی ماشین و محرکه های الکتریکی

صفحه اصلی / The 5th International Conference on Electrical Machines and Drives

Active Equalization for an Integrated Bidirectional Electric Vehicle Charger Using a Hybrid Model Predictive and Reinforcement Learning Framework

نویسندگان :

Peyman Bayat¹ Pezhman Bayat²

1- Department of Electrical Engineering, Hamedan University of Technology, Hamedan, Iran 2- Department of Electrical Engineering, Hamedan University of Technology, Hamedan, Iran

کلمات کلیدی :

Battery charging systems،Charge equalization،Electric vehicles،Model predictive control،Reinforcement learning

چکیده :

The escalating global adoption of electric vehicles (EVs) necessitates the development of advanced battery charging systems that deliver not only high efficiency but also proactive battery health management through effective charge equalization. This paper introduces a novel bidirectional battery charger architecture that seamlessly integrates a modular charge equalization circuit into its core power conversion stage. The system's performance is optimized by an innovative adaptive predictive equalization control (APEC), which synergistically combines the predictive capabilities of model predictive control (MPC) with the adaptive level of reinforcement learning (RL). A key differentiator of this design is its unified control framework, which consolidates the traditionally separate functions of battery charging and cell equalization. This integration facilitates rapid active balancing of cell voltages, minimizes input current ripple, and provides precise reactive power support to the grid. A comprehensive simulation analysis validates the proposed APEC methodology, demonstrating outstanding performance metrics: a peak system efficiency of 94.2%, a cell equalization time of less than 15 minutes, and notable improvements in additional system parameters during both grid-to-vehicle and vehicle-to-grid operational modes. The results clearly demonstrate that the proposed charger, integrating MPC with RL, significantly outperforms existing state-of-the-art solutions, setting a new benchmark in speed, efficiency, and operational reliability for next-generation EV power electronics.