| 引用本文: | 陈锦铭,蒋玮,陈武,徐俊俊,卢庆宁,袁宇波,陈烨.基于实时量测与滑动时间窗策略的配电网运行方式动态追踪[J].电力自动化设备,2026,46(1):154-162 |
| CHEN Jinming,JIANG Wei,CHEN Wu,XU Junjun,LU Qingning,YUAN Yubo,CHEN Ye.Dynamic tracking of distribution network operation mode based on real-time measurement and sliding time window strategy[J].Electric Power Automation Equipment,2026,46(1):154-162 |
|
| 摘要: |
| 传统配电网的拓扑追踪方法未充分考虑配电网潮流的多样性和低压工况的复杂性，存在工程适用性弱的问题。为此，提出了一种数据驱动方法，利用实时量测实现配电网运行方式的追踪与校核。基于配电网运行时空特性，提出了基于关键路径及边界区段分析的技术路径。构建了基于滑动时间窗的两阶段运行方式调整诊断框架：先对边界区段进行动态监测，利用CatBoost分类器对功率、电压等特征开展诊断，辨识转供行为，再反向遍历馈线拓扑以聚合上游转供区段，明确转供边界并更新动态拓扑。通过横向对比得到滑动时间窗及诊断时延的合理配置，同时证明CatBoost分类算法在性能方面的优异性。实验结果表明所提方法能够准确实现配电网运行方式的动态追踪，并在准确性和时效性方面达到了较好的均衡。 |
| 关键词: 配电网 拓扑辨识 运行方式追踪 实时量测 滑动时间窗 CatBoost算法 |
| DOI：10.16081/j.epae.202507024 |
| 分类号:TM73 |
| 基金项目:国家自然科学基金资助项目(52277172)；国家重点研发计划项目(2022YFB2404200) |
|
| Dynamic tracking of distribution network operation mode based on real-time measurement and sliding time window strategy |
|
CHEN Jinming1,2, JIANG Wei1, CHEN Wu1, XU Junjun3, LU Qingning2, YUAN Yubo2, CHEN Ye2
|
|
1.School of Electrical Engineering, Southeast University, Nanjing 210096, China;2.Electric Power Research Institute of State Grid Jiangsu Electric Power Co.,Ltd.,Nanjing 211103, China;3.College of Automation & College of Artificial Intelligence, Nanjing University of Posts and Telecommunications, Nanjing 210023, China
|
| Abstract: |
| The traditional topology tracking methods of distribution network do not fully consider the diversity of power flow and the complexity of low-voltage working conditions in distribution network, and thus have the problem of weak engineering applicability. To this end, a data-driven method is proposed to track and verify the operation modes of distribution network using real-time measurement. Based on the spatio-temporal operation characteristics of distribution network, a technical path based on critical path and boundary segment analysis is proposed. A diagnosis framework of two-stage operation mode adjustment based on sliding time window is constructed. Firstly, the dynamic monitoring of boundary segments is conducted, and the CatBoost classifier is used to diagnose features such as power and voltage to identify the transfer behaviors. Then, the feeder topology is traversed in reverse to aggregate the upstream transfer segments, so as to clarify the transfer segments and update the dynamic topology. Through horizontal comparison, the reasonable configuration of sliding time window and diagnosis time delay is obtained, and at the same time, the superiority of the CatBoost classification algorithm in terms of performance is proved. The experimental results show that the proposed method can accurately achieve the dynamic tracking of distribution network operation mode and achieve a good balance in terms of accuracy and timeliness. |
| Key words: distribution network topology identification operation mode tracking real-time measurement sliding time window CatBoost algorithm |
