Computer Engineering and Applications ›› 2019, Vol. 55 ›› Issue (16): 255-258.DOI: 10.3778/j.issn.1002-8331.1805-0046

Previous Articles     Next Articles

Low-Rank Representation for Outliers Detection in Power State Estimation

LI Yongpan, MEN Kun, WU Junyang   

  1. 1.Shenzhen Power Supply Co., Ltd, Shenzhen, Guangdong 518001, China
    2.Sichuan Energy Internet Research Institute, Tsinghua University, Chengdu 610213, China
  • Online:2019-08-15 Published:2019-08-13



  1. 1.深圳供电局有限公司,广东 深圳 518001
    2.清华四川能源互联网研究院,成都 610213

Abstract: Smart grid plays a critical role in national production. The stability and security of smart grid is essentially important. As a result, it is significant to detect the bad and malicious data from daily observations. This paper proposes a novel outlier detection method based on low-rank representation. Specifically, the observation is decomposed into two parts:a low-rank part for clean data and a sparse part for outliers. In addition, this paper deploys ALM(Augmented Lagrange Multiplier) to optimize the objective. Extensive experiments on two popular benchmarks verify the advantages of the proposed method.

Key words: low-rank representation, data mining, cyberspace security, power state estimation

摘要: 智能电网在国民生计中扮演着至关重要的角色,智能电网的稳定性和安全性是网络建设中需要特别考虑的问题。在智能电网的建设和运营中,如何从日常监测数据中及时检测出异常信息和有害信息,比如网络入侵数据和系统隐患数据,对智能电网的稳定性和安全性有着举足轻重的影响。提出一种基于低秩模型的电力数据异常检测算法,将系统观测数据分解为低秩部分和稀疏部分,用低秩部分表达干净的观测,用稀疏部分表达异常数据。然后,通过增广拉格朗日方法来优化目标方程。在公开数据集上的实验结果验证了所提出算法的有效性。

关键词: 低秩表示, 数据挖掘, 网络空间安全, 电力状态估计