Scholarlink

With the evolution of electric vehicle (EV) technology, electric vehicles have gained popularity for their various benefits. However, the deployment of EV charging stations faces significant challenges, including high costs, grid voltage deviations, and low charging efficiency. To address these issues, this article proposes an EV wireless charging system that takes advantage of the existing urban public transport network. By integrating on-line electric vehicle systems with microwave power transmission systems, the proposed system offers wireless charging services to electric vehicles. Traditional model-based approaches require prediction models to handle uncertainties in scheduling and optimization. In contrast, we introduce an approach based on multiagent constrained deep reinforcement learning (cMADDPG) to solve this problem. The goal of cMADDPG is to maximize the total remaining energy of all electric vehicles so that they can reach their destinations before the deadline. Extensive simulations demonstrate the effectiveness of cMADDPG. Compared to existing solutions, it increases the average residual energy by 12.66 \(\%\) and reduces the travel time by 10.22 \(\%\) . These results confirm that the proposed method offers a promising solution for EV wireless charging systems.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Constrained Multi-agent Reinforcement Learning Approach on Wireless Charging Scheduling

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Yingjun Liu,
  • Fuchun Liu,
  • Hongzhen Zhu

摘要

With the evolution of electric vehicle (EV) technology, electric vehicles have gained popularity for their various benefits. However, the deployment of EV charging stations faces significant challenges, including high costs, grid voltage deviations, and low charging efficiency. To address these issues, this article proposes an EV wireless charging system that takes advantage of the existing urban public transport network. By integrating on-line electric vehicle systems with microwave power transmission systems, the proposed system offers wireless charging services to electric vehicles. Traditional model-based approaches require prediction models to handle uncertainties in scheduling and optimization. In contrast, we introduce an approach based on multiagent constrained deep reinforcement learning (cMADDPG) to solve this problem. The goal of cMADDPG is to maximize the total remaining energy of all electric vehicles so that they can reach their destinations before the deadline. Extensive simulations demonstrate the effectiveness of cMADDPG. Compared to existing solutions, it increases the average residual energy by 12.66 \(\%\) and reduces the travel time by 10.22 \(\%\) . These results confirm that the proposed method offers a promising solution for EV wireless charging systems.