中文

少量标签足矣:针对智能表计系列中的电器定位的弱监督框架

机器学习 2025-06-09 v1

摘要

提高智能电网系统管理至关重要,这有助于应对气候变化,而让消费者在这一过程中发挥积极作用则是一个重要挑战。近年来,数以百万计的智能表计已在全球部署,记录着单个住户消耗的主要电力。然而,这些数据聚合了房屋中不同电器同时运行的耗电,使得难以把握。非侵入式负荷监测(NILM)指的是使用主表计信号来估计单个电器的功率消耗、模式或开/关状态激活的挑战。最近的一些方法基于完全监督的深度学习方法来解决此任务,需要同时具备聚合信号和单个电器功率的真实标签。然而,这些标签在实践中昂贵且极其稀缺,因为它们需要对住户进行侵入性调查以监控每个电器。本文提出CamAL(A weakly supervised approach for appliance pattern localization that only requires information on the presence of an appliance in a household to be trained. CamAL merges an ensemble of deep-learning classifiers combined with an explainable classification method to be able to localize appliance patterns. Our experimental evaluation, conducted on 4 real-world datasets, demonstrates that CamAL significantly outperforms existing weakly supervised baselines and that current SotA fully supervised NILM approaches require significantly more labels to reach CamAL performances. The source of our experiments is available at: https://github.com/adrienpetralia/CamAL. This paper appeared in ICDE 2025.)的方法,这种方法只需要电器存在于住户中的信息即可进行训练。CamAL融合了深度学习分类器的集成方法,结合可解释的分类方法,以便对电器模式进行定位。我们的实验评估在4个真实数据集上进行,表明CamAL显著优于现有的弱监督基线方法,而当前最先进的完全监督NILM方法需要显著更多的标签才能达到CamAL的性能。我们的实验代码可在https://github.com/adrienpetralia/CamAL获取。本文已发表于ICDE 2025。

关键词

引用

@article{arxiv.2506.05895,
  title  = {Few Labels are all you need: A Weakly Supervised Framework for Appliance Localization in Smart-Meter Series},
  author = {Adrien Petralia and Paul Boniol and Philippe Charpentier and Themis Palpanas},
  journal= {arXiv preprint arXiv:2506.05895},
  year   = {2025}
}

备注

12 pages, 10 figures. This paper appeared in IEEE ICDE 2025