Scholarlink

<p>The rapid penetration of electric vehicles in recent years has imposed substantial charging demands on power grids. The concentration and volatility of such demands can exacerbate peak–valley differences, increase operational stress, and compromise supply stability. Effective charging scheduling can flatten the load curve, mitigate peak–valley differences, and thereby maintain grid stability. However, current studies often fail to account for user anxiety, which affects users’ dynamic charging behaviors and subsequently reduces the effectiveness of scheduling decisions. This leads to inefficient or premature charging. In addition, current studies struggle to address sparse load data in new stations, limiting the practical applicability and effectiveness of scheduling strategies. Therefore, an intelligent charging scheduling method that considers users’ travel anxiety is proposed in this study. First, a Cognition-Perception Markov decision process model was developed to quantify the user’s cognitive characteristics such as anxiety, experience, and risk preference. Next, we propose a Deep Reinforcement Learning(DRL)-based model for charging scheduling. To enable the DRL to adapt fast to unseen environments, we further extend it to Meta-DRL, a meta-learning-based scheduling framework. Simulation results demonstrate that the proposed approach significantly reduces user charging cost and anxiety levels while enhancing training efficiency and policy robustness. This study offers a promising direction for intelligent charging scheduling in complex and dynamic environments.</p>

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

Electric vehicle charging optimization scheduling strategy considering users’ travel anxiety: a meta-deep reinforcement learning method

  • 期刊论文

  • 发表日期：

    2026年7月11日

查看全文
  • Ruixia Hu,
  • Ruijuan Zheng,
  • Junlong Zhu,
  • Yongsheng Dong,
  • Mingchuan Zhang,
  • Xuhui Zhao

摘要

The rapid penetration of electric vehicles in recent years has imposed substantial charging demands on power grids. The concentration and volatility of such demands can exacerbate peak–valley differences, increase operational stress, and compromise supply stability. Effective charging scheduling can flatten the load curve, mitigate peak–valley differences, and thereby maintain grid stability. However, current studies often fail to account for user anxiety, which affects users’ dynamic charging behaviors and subsequently reduces the effectiveness of scheduling decisions. This leads to inefficient or premature charging. In addition, current studies struggle to address sparse load data in new stations, limiting the practical applicability and effectiveness of scheduling strategies. Therefore, an intelligent charging scheduling method that considers users’ travel anxiety is proposed in this study. First, a Cognition-Perception Markov decision process model was developed to quantify the user’s cognitive characteristics such as anxiety, experience, and risk preference. Next, we propose a Deep Reinforcement Learning(DRL)-based model for charging scheduling. To enable the DRL to adapt fast to unseen environments, we further extend it to Meta-DRL, a meta-learning-based scheduling framework. Simulation results demonstrate that the proposed approach significantly reduces user charging cost and anxiety levels while enhancing training efficiency and policy robustness. This study offers a promising direction for intelligent charging scheduling in complex and dynamic environments.