Scholarlink

<p>The rapid growth of electric vehicles (EVs) creates challenges for grid stability, peak demand management, and renewable energy integration. Conventional cloud-centric charging coordination systems rely on continuous communication and suffer from latency that limits real-time responsiveness. This study introduces a novel distributed edge-intelligent EV charging coordination framework that integrates behavioral prediction, grid-aware scheduling, renewable-aware optimization, and localized AI inference within a communication-efficient IoT architecture. Lightweight models (CNN+LSTM, XGBoost, and Random Forest) are deployed on Jetson Orin Nano and Raspberry Pi 5 devices to enable low-latency decision-making with reduced reliance on the cloud. The framework is validated using a large-scale U.S. Department of Energy dataset. Experimental results demonstrate 27.6% improvement in station utilization, 24.5% reduction in peak grid load, and 29.8% decrease in user charging costs, with predictive performance reaching R² = 0.92. The findings provide a scalable and communication-efficient reference design for data-driven EV charging coordination in smart grid systems.</p> Graphical abstract <p></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

Edge-intelligent electric vehicle charging coordination for grid load balancing and renewable integration

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年6月29日

查看全文
  • Abdulkadir Gozuoglu,
  • Zafer Dogan

摘要

The rapid growth of electric vehicles (EVs) creates challenges for grid stability, peak demand management, and renewable energy integration. Conventional cloud-centric charging coordination systems rely on continuous communication and suffer from latency that limits real-time responsiveness. This study introduces a novel distributed edge-intelligent EV charging coordination framework that integrates behavioral prediction, grid-aware scheduling, renewable-aware optimization, and localized AI inference within a communication-efficient IoT architecture. Lightweight models (CNN+LSTM, XGBoost, and Random Forest) are deployed on Jetson Orin Nano and Raspberry Pi 5 devices to enable low-latency decision-making with reduced reliance on the cloud. The framework is validated using a large-scale U.S. Department of Energy dataset. Experimental results demonstrate 27.6% improvement in station utilization, 24.5% reduction in peak grid load, and 29.8% decrease in user charging costs, with predictive performance reaching R² = 0.92. The findings provide a scalable and communication-efficient reference design for data-driven EV charging coordination in smart grid systems.

Graphical abstract