{"title":"Lightweight Federated Learning for On-Device Non-Intrusive Load Monitoring","authors":"Yehui Li;Ruiyang Yao;Dalin Qin;Yi Wang","doi":"10.1109/TSG.2024.3482363","DOIUrl":null,"url":null,"abstract":"Non-intrusive load monitoring (NILM) is a critical technology for disaggregating appliance-specific energy usage by only observing household-level power consumption. If NILM can be performed on end devices (such as smart meters), it can facilitate electricity demand identification and electricity behavior perception for real-time demand-side energy management. However, implementing high-performance NILM models on end devices presents an unresolved issue, encompassing two primary challenges: hardware resource constraints and data resource paucity on end devices. To this end, this paper proposes a lightweight federated learning approach for on-device NILM by combining neural architecture search (NAS) and federated learning. Firstly, a memory-efficient NAS approach is investigated to determine a personalized model within the resource constraints of end devices. Secondly, a federated mutual learning approach is designed to orchestrate the cooperation of distributed end devices with heterogeneous personalized models in a privacy-preserving manner. Case studies on two real-world datasets verify that the proposed method for appliance-level power disaggregation outperforms conventional methods in accuracy and efficiency.","PeriodicalId":13331,"journal":{"name":"IEEE Transactions on Smart Grid","volume":"16 2","pages":"1950-1961"},"PeriodicalIF":8.6000,"publicationDate":"2024-10-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"IEEE Transactions on Smart Grid","FirstCategoryId":"5","ListUrlMain":"https://ieeexplore.ieee.org/document/10720908/","RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"ENGINEERING, ELECTRICAL & ELECTRONIC","Score":null,"Total":0}
引用次数: 0
Abstract
Non-intrusive load monitoring (NILM) is a critical technology for disaggregating appliance-specific energy usage by only observing household-level power consumption. If NILM can be performed on end devices (such as smart meters), it can facilitate electricity demand identification and electricity behavior perception for real-time demand-side energy management. However, implementing high-performance NILM models on end devices presents an unresolved issue, encompassing two primary challenges: hardware resource constraints and data resource paucity on end devices. To this end, this paper proposes a lightweight federated learning approach for on-device NILM by combining neural architecture search (NAS) and federated learning. Firstly, a memory-efficient NAS approach is investigated to determine a personalized model within the resource constraints of end devices. Secondly, a federated mutual learning approach is designed to orchestrate the cooperation of distributed end devices with heterogeneous personalized models in a privacy-preserving manner. Case studies on two real-world datasets verify that the proposed method for appliance-level power disaggregation outperforms conventional methods in accuracy and efficiency.
期刊介绍:
The IEEE Transactions on Smart Grid is a multidisciplinary journal that focuses on research and development in the field of smart grid technology. It covers various aspects of the smart grid, including energy networks, prosumers (consumers who also produce energy), electric transportation, distributed energy resources, and communications. The journal also addresses the integration of microgrids and active distribution networks with transmission systems. It publishes original research on smart grid theories and principles, including technologies and systems for demand response, Advance Metering Infrastructure, cyber-physical systems, multi-energy systems, transactive energy, data analytics, and electric vehicle integration. Additionally, the journal considers surveys of existing work on the smart grid that propose new perspectives on the history and future of intelligent and active grids.