湖南电力 ›› 2026, Vol. 46 ›› Issue (1): 19-28.doi: 10.3969/j.issn.1008-0198.2026.01.003

• 电网运行与控制 • 上一篇    下一篇

基于改进双向门控循环单元的电动汽车可调节能力辨识方法

李永军1, 潘明明1, 郑博文1, 王小明2, 赵文广2, 汪宇航2   

  1. 1.中国电力科学研究院有限公司,北京 100192;
    2.国网安徽省电力有限公司电力科学研究院,安徽 合肥 230601
  • 收稿日期:2025-08-11 修回日期:2025-10-16 出版日期:2026-02-25 发布日期:2026-03-10
  • 通信作者: 李永军(1992),男,工程师,从事用电安全、负荷管理工作。
  • 基金资助:
    国家电网有限公司科技项目(5400-202320581A-3-2-ZN)

Improved BiGRU-Based Method for Identifying the Adjustable Capabilities of Electric Vehicles

LI Yongjun1, PAN Mingming1, ZHENG Bowen1, WANG Xiaoming2, ZHAO Wenguang2, WANG Yuhang2   

  1. 1. China Electric Power Research Institute, Beijing 100192, China;
    2. State Grid Anhui Electric Power Company Limited Research Institute, Hefei 230601, China
  • Received:2025-08-11 Revised:2025-10-16 Online:2026-02-25 Published:2026-03-10

摘要: 为提升电动汽车可调节能力的识别精度与鲁棒性,提出一种融合自适应提升算法(adaptive boosting,AdaBoost)机制的双向门控循环单元(bidirectional gated recurrent unit,BiGRU)深度学习模型。首先,基于用户、区域、时间3个维度的充电行为刻画,构建涵盖充电时段、电量及外部环境等13项关键变量的输入体系,对电动汽车群体进行集群划分,为电网中充电负荷的调节奠定基础。其次,利用BiGRU模型挖掘时序特征,并通过AdaBoost集成学习机制增强模型泛化能力,提高对电动汽车可调节能力辨识的准确性和鲁棒性,进一步提升电动汽车调度效率。最后,在真实历史数据集上进行实验验证,评估模型在不同场景下的辨识效果。结果表明,本文所提方法能够有效提升电动汽车调节能力辨识的准确率,提升电动汽车可调度能力,同时对不同日期、不同场景的电动汽车调节能力辨识作用较好。

关键词: 电动汽车, 可调节能力, 行为特征辨识, 双向门控循环单元

Abstract: To improve the recognition accuracy and robustness of the adjustable capability of electric vehicles, a Bidirectional Gated Recurrent Unit(BiGRU) deep learning model that integrates Adaptive Boosting(AdaBoost) mechanism is proposed. Firstly, based on the characterization of charging behavior from three dimensions: user, region, and time, an input system covering 13 key variables such as charging period, electricity consumption, and external environment is constructed to cluster the electric vehicle population and lay the foundation for regulating charging load in the power grid. Secondly, the BiGR model is utilized to mine temporal features, and the AdaBoos ensemble learning mechanism is used to enhance the model's generalization ability, improve the accuracy and robustness of identifying the adjustable capabilities of electric vehicles, and further improve the efficiency of electric vehicle scheduling. Finally, experimental validation is conducted on a real historical dataset to evaluate the identification effect of the model in different scenarios. The results show that the proposed method can effectively improve the accuracy of identifying the adjustable capability of electric vehicles, enhance their schedulability, and has a good effect on identifying the adjustable capability of electric vehicles in different dates and scenarios.

Key words: electric vehicles, adjustable capabilities, behavior feature identification, bidirectional gated recurrent unit(BiGRU)

中图分类号: