Scholarlink

The widespread adoption of electric vehicles (EVs) impacts the power grid, with uncoordinated charging potentially worsening load peaks and reducing renewable energy efficiency. Vehicle-to-Grid (V2G) technology enables EVs to support grid regulation through bidirectional charging/discharging. This paper proposes a multi-agent reinforcement learning strategy for efficient EV charging/discharging management. It models the problem as a spatio-temporal sequential Markov game, using a Stackelberg game mechanism to guide charging stations toward sequential decisions for near-optimal global solutions. A Transformer-based autoregressive structure enhances agents’ ability to consider historical data and other stations’ strategies, improving coordination and decision efficiency. Experiments demonstrate superior performance in revenue optimization, grid load balancing, and computational efficiency, offering a scalable framework for EV scheduling.

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

Charging and Discharging Scheduling Optimization Based on Multi-agent Reinforcement Learning

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Peng Wu,
  • Peng Liu

摘要

The widespread adoption of electric vehicles (EVs) impacts the power grid, with uncoordinated charging potentially worsening load peaks and reducing renewable energy efficiency. Vehicle-to-Grid (V2G) technology enables EVs to support grid regulation through bidirectional charging/discharging. This paper proposes a multi-agent reinforcement learning strategy for efficient EV charging/discharging management. It models the problem as a spatio-temporal sequential Markov game, using a Stackelberg game mechanism to guide charging stations toward sequential decisions for near-optimal global solutions. A Transformer-based autoregressive structure enhances agents’ ability to consider historical data and other stations’ strategies, improving coordination and decision efficiency. Experiments demonstrate superior performance in revenue optimization, grid load balancing, and computational efficiency, offering a scalable framework for EV scheduling.