Pre-trained non-intrusive load monitoring model for recognizing activity of daily living

Gyubaek Kim, Sanghyun Park

Research output: Contribution to journalArticlepeer-review


Non-intrusive load monitoring (NILM) is a technology that analyzes total electricity consumption data to determine whether a specific type of residential appliance is operated. However, despite many previous studies conducted in this field, NILM is still limited to developing models in the form of regression and presenting relative performance results. However, to be used in commercial applications, it must be finally developed as a classification model to present the absolute performance results from the service viewpoint. In this paper, a new methodology is proposed to build a pre-trained NILM model that ensures reliable classification performance, whereby differentiated steps are adopted to improve NILM accuracy. In particular, the methodology includes an intelligent result processing scheme that accurately identifies appliance activation based on past profiles and translates the NILM results into proper information for service utilization, which has not been addressed in previous studies. The developed pre-trained NILM model can be used to detect appliance activation to indirectly monitor the activity of daily living (ADL) of single-person households in social safety net services as a vital sign. The detailed explanation and experimental results are presented as reference to build a pre-trained NILM model in many domains.

Original languageEnglish
JournalApplied Intelligence
Publication statusAccepted/In press - 2022

Bibliographical note

Funding Information:
This work was supported by Institute of Information & communications Technology Planning & Evaluation (IITP) grant funded by the Korea government(MSIT) (IITP-2017-0-00477, (SW starlab) Research and development of the high performance in-memory distributed DBMS based on flash memory storage in IoT environment).

Publisher Copyright:
© 2022, The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature.

All Science Journal Classification (ASJC) codes

  • Artificial Intelligence


Dive into the research topics of 'Pre-trained non-intrusive load monitoring model for recognizing activity of daily living'. Together they form a unique fingerprint.

Cite this