Electricity Consumption Data Cleansing and Imputation Based on Robust Nonnegative Matrix Factorization

被引：0
|
作者：
Liu Q. [1 ,2 ]
Zhong Y. [2 ]
Lin C. [1 ,2 ]
Li T. [2 ]
Yang C. [1 ]
Fu Z. [1 ]
Li X. [1 ]
机构：
[1] School of Electrical Engineering, Chongqing University, Shapingba District, Chongqing
[2] Measurement Center of Yunnan Power Grid Co., Ltd., Yunnan Province, Kunming
来源：
Dianwang Jishu/Power System Technology | 2024年 / 48卷 / 05期
关键词：
cleansing; imputation; low rank; nonnegative matrix; outlier; sparse;
D O I：
10.13335/j.1000-3673.pst.2023.0557
中图分类号：
学科分类号：
摘要：
Given the quality problems of noise, outlier, and loss in the collection and transmission of electricity consumption data under operating conditions, low eigenvalue and sparse outlier of the spatiotemporal distribution of electricity consumption data of a single user are used, and then a unified processing framework for data missing filling, noise reduction, and outlier elimination based on the low-rank matrix completion is proposed. Firstly, because of the huge differences in actual multi-user electricity consumption scenarios and characteristics, a data matrix with low-rank characteristics is constructed only according to the inherent similarity of a single user's electricity consumption behavior. Furthermore, considering the effects of additive background noise such as column and sparse anomalies, a non-negative matrix complete optimization model with low rank lifting regular constraints is constructed. Finally, the iterative least square method is used to solve the optimization problem to fill in the missing data and eliminate the multiple background noise. The effectiveness and accuracy of the proposed algorithm are verified by simulation and experimental results. © 2024 Power System Technology Press. All rights reserved.
引用
收藏
页码：2103 / 2112
页数：9
相关论文
共 36 条
  • [31] HAEFFELE B D, YOUNG E D, VIDAL R．, Structured low-rank matrix factorization：optimality，algorithm and applications to image processing[C], 31st International Conference on Machine Learning, (2014)
  • [32] GUAN Naiyang, Dacheng TAO, Zhigang LUO, NeNMF：an optimal gradient method for nonnegative matrix factorization[J], IEEE Transactions on Signal Processing, 60, 6, pp. 2882-2898, (2012)
  • [33] ZHANG Sheng, WANG Weihong, FORD J, Learning from incomplete ratings using non-negative matrix factorization[C], Proceedings of the 2006 SIAM International Conference on Data Mining, pp. 549-553, (2006)
  • [34] HASTIE T, MAZUMDER R, LEE J D, Matrix completion and low-rank SVD via fast alternating least squares[J], The Journal of Machine Learning Research, 16, 1, pp. 3367-3402, (2015)
  • [35] GIAMPOURAS P V，, RONTOGIANNIS A A, KOUTROUMBAS K D．, Alternating iteratively reweighted least squares minimization for low-rank matrix factorization[J], IEEE Transactions on Signal Processing, 67, 2, pp. 490-503, (2019)
  • [36] Jianfeng CAI, CANDES E J，, SHEN Zuowei, A singular value thresholding algorithm for matrix completion[J], SIAM Journal on Optimization, 20, 4, pp. 1956-1982, (2010)
1234