文章摘要
梁寿愚,方文崇,王瑾,等.基于图规则化低秩矩阵恢复的用电数据修复与异常检测[J].电力系统自动化,2019,43(21):221-228. DOI: 10.7500/AEPS20181122008.
LIANG Shouyu,FANG Wenchong,WANG Jin, et al.Refinement and Anomaly Detection for Power Consumption Data Based on Recovery of Graph Regularized Low-rank Matrix[J].Automation of Electric Power Systems,2019,43(21):221-228. DOI: 10.7500/AEPS20181122008.
基于图规则化低秩矩阵恢复的用电数据修复与异常检测
Refinement and Anomaly Detection for Power Consumption Data Based on Recovery of Graph Regularized Low-rank Matrix
DOI:10.7500/AEPS20181122008
关键词: 用电异常检测  矩阵补全  稀疏表示  低秩分解
KeyWords: abnormal power consumption detection  matrix completion  sparse representation  low-rank decomposition
上网日期:2019-06-04
基金项目:
作者单位E-mail
梁寿愚 中国南方电网有限责任公司, 广东省广州市 510000  
方文崇 中国南方电网有限责任公司, 广东省广州市 510000  
王瑾 国电南瑞科技股份有限公司, 江苏省南京市 211106  
何超林 中国南方电网有限责任公司, 广东省广州市 510000  
张磊 云南电网有限责任公司德宏供电局, 云南省德宏市 678400  
张骥 国电南瑞科技股份有限公司, 江苏省南京市 211106 zhangji1@sgepri.sgcc.com.cn 
摘要:
      针对电力系统用电数据中的记录误差与异常用电,提出一种基于图规则化低秩矩阵恢复的电力系统用电记录修复与异常检测算法。该方法从用户用电时空矩阵的低秩稀疏分解出发,结合电网拓扑结构与用户相关性的规则化调整,获取修复后的用电数据和异常用户。该方法同时兼顾了用户用电的周期性与异常用户的差异性特点。实验分析表明,与相关方法相比,所提方法在用电数据修复与异常用电模式检测的多项评价标准下均取得了更好的准确性和鲁棒性。
Abstract:
      Aiming at the error and abnormal consumption data in power system, this paper proposes a record data refinement and anomaly detection algorithm for power system based on recovery of graph regularized low-rank matrix. The method starts from the low-rank sparse decomposition of the electricity consumption space-time matrix, and integrates the grid topology and the user correlation to adjust the matrix decomposition results, so as to obtain the repaired power consumption data and abnormal users. The method takes into account both the periodicity of users and the difference of abnormal users. Compared with the related methods, the experimental analysis shows that the proposed method achieves better accuracy and robustness under multiple evaluation criteria of electrical data refinement and abnormal power mode detection.
查看全文(Free!)   查看附录   查看/发表评论  下载PDF阅读器