Scholarlink

<p>A Vehicle-to-Grid (V2G) enabled smart grid (SG)-based Electric Vehicle Charging System (EVCS) facilitates intelligent energy exchange between electric vehicles (EVs) and the power grid through effective energy management and grid performance optimization. However, scalability and efficiency are often hindered by inaccurate load estimation, system instability, and poor coordination of charging sessions. This paper proposes an innovative hybrid framework that integrates the Kookaburra Optimization Algorithm (KOA) with a Heterogeneous Context-Aware Graph Convolutional Network (HCAGCN) to enhance classification and prediction in V2G-enabled SGs. The proposed model is validated using an Electric Vehicle Charging Dataset (EVCD) comprising more than 130,000 charging session records collected over an 18-month period (2019–2021) from geographically distributed charging stations across the United States. The dataset captures diverse public and workplace charging behaviors, including energy consumption, charging duration, temporal attributes, and station-level information. Fast Resampled Iterative Filtering (FRIF) is employed to preprocess and temporally align the EVCD, followed by feature selection using the Snow Ablation Optimizer (SAO), which identifies the most informative attributes such as sessionId, kWhTotal, startTime, endTime, chargeTimeHrs, and stationId. These features are then used by HCAGCN to classify charging sessions into four load categories: Low Load, Normal Load, High Load, and Overloaded, while KOA optimizes model hyperparameters to improve predictive accuracy. Implemented in MATLAB, the proposed KOA-HCAGCN framework achieves a classification accuracy of 99% with a low RMSE of 0.258, demonstrating superior robustness, precision, and efficiency for intelligent V2G-enabled SG applications.</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

Smart Grid-Based Electric Vehicle Charging System: Intelligent Decision Making through Neural Networks and Optimization for V2G Integration

  • 期刊论文

  • 发表日期：

    2026年2月25日

查看全文
  • A. Arunya Revathi,
  • Pandikumar Maniraj,
  • Papana Venkata Prasad,
  • Manam Ravindra

摘要

A Vehicle-to-Grid (V2G) enabled smart grid (SG)-based Electric Vehicle Charging System (EVCS) facilitates intelligent energy exchange between electric vehicles (EVs) and the power grid through effective energy management and grid performance optimization. However, scalability and efficiency are often hindered by inaccurate load estimation, system instability, and poor coordination of charging sessions. This paper proposes an innovative hybrid framework that integrates the Kookaburra Optimization Algorithm (KOA) with a Heterogeneous Context-Aware Graph Convolutional Network (HCAGCN) to enhance classification and prediction in V2G-enabled SGs. The proposed model is validated using an Electric Vehicle Charging Dataset (EVCD) comprising more than 130,000 charging session records collected over an 18-month period (2019–2021) from geographically distributed charging stations across the United States. The dataset captures diverse public and workplace charging behaviors, including energy consumption, charging duration, temporal attributes, and station-level information. Fast Resampled Iterative Filtering (FRIF) is employed to preprocess and temporally align the EVCD, followed by feature selection using the Snow Ablation Optimizer (SAO), which identifies the most informative attributes such as sessionId, kWhTotal, startTime, endTime, chargeTimeHrs, and stationId. These features are then used by HCAGCN to classify charging sessions into four load categories: Low Load, Normal Load, High Load, and Overloaded, while KOA optimizes model hyperparameters to improve predictive accuracy. Implemented in MATLAB, the proposed KOA-HCAGCN framework achieves a classification accuracy of 99% with a low RMSE of 0.258, demonstrating superior robustness, precision, and efficiency for intelligent V2G-enabled SG applications.