系统仿真学报 ›› 2023, Vol. 35 ›› Issue (10): 2193-2201.doi: 10.16182/j.issn1004731x.joss.23-FZ0800

• 论文 • 上一篇    下一篇

基于日志信息的电网离散事件动态模型学习和分析

朱丹龙1(), 闫云琦2, 陈颖2(), 张家琦2, 晋龙兴1, 付威1   

  1. 1.深圳供电局有限公司,广东 深圳 518048
    2.清华大学 电机系,北京 100084
  • 收稿日期:2023-07-02 修回日期:2023-09-12 出版日期:2023-10-30 发布日期:2023-10-26
  • 通讯作者: 陈颖 E-mail:zhudanlong@sz.csg.cn;chen_ying@tsinghua.edu.cn
  • 第一作者简介:朱丹龙(1988-),男,工程师,本科,研究方向为电力系统自动化。E-mail:zhudanlong@sz.csg.cn
  • 基金资助:
    国家自然科学基金企业创新发展联合基金重点资助项目(U22B2096)

Learning and Analysis of Dynamic Models for Grid Discrete Events Based on Log Information

Zhu Danlong1(), Yan Yunqi2, Chen Ying2(), Zhang Jiaqi2, Jin Longxing1, Fu Wei1   

  1. 1.Shenzhen Power Grid Utility Co. LTD, Shenzhen 518048, China
    2.Department of Electrical Engineering, Tsinghua University, Beijing 100084, China
  • Received:2023-07-02 Revised:2023-09-12 Online:2023-10-30 Published:2023-10-26
  • Contact: Chen Ying E-mail:zhudanlong@sz.csg.cn;chen_ying@tsinghua.edu.cn

摘要:

随着电网规模的增加,电网设备产生的海量日志信息为人工分析电网异常状态带来了挑战。电网运行中的日志信息具有典型的离散序列特征,通过分析电网告警信息日志,提出了厂站事件转移概率模型和事件序列风险计算 方法 ,可以有效建模和分析厂站一二次系统运行异常水平。该方法不仅可以成功地识别出已被调度员人工标记的电网故障对应事件序列,还能识别出未曾被调度员记录的异常告警信息序列,有助于分析研判电网设备运行态势,发现潜在风险,提升变电站运维效率。

关键词: 电力系统, 告警信息, 动态特征学习, 异常识别, 风险评估

Abstract:

With the increasing scale of power grid, the massive amount of log information generated by devices in the power grid poses a challenge to the manual analysis of abnormal grid conditions. The log information generated during the operation of the power grid has the typical discrete sequential characteristics. By analyzing the log information of grid alarm messages, a station event transition probability model and an event sequence risk calculation method are proposed to effectively model and analyze the abnormal operation level of primary and secondary systems in substations. The proposed method not only successfully identifies the event sequences corresponding to grid failures that have been manually recarded by dispatchers but also identifies the abnormal alarm information sequences that have not been recorded by dispatchers. This helps analyze and assess the operational situation of power grid equipment, discover potential risks, and improve the efficiency of substation maintenance.

Key words: power system, alarm information, dynamic feature learning, anomaly detection, risk assessment

中图分类号: