Identifying AC Control Patterns From a Massive Longitudinal Log Dataset Using Deep Clustering

An air conditioner is a household appliance and one of the most innovative inventions, primarily used to cool indoor temperatures and lower humidity to provide a comfortable environment. The sale of air conditioners increased continuously, reaching approximately 135 million units annually, with near...

Full description

Saved in:
Bibliographic Details
Main Authors: Seunghwan Bae, Donghee Kim, Seokhwan Seong, Jaeyeon Jang, Young-Kyoon Suh, Honghyun Kim, Dongwoo Han, Inuk Jung
Format: Article
Language:English
Published: IEEE 2025-01-01
Series:IEEE Access
Subjects:
Online Access:https://ieeexplore.ieee.org/document/10942373/
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849768410977665024
author Seunghwan Bae
Donghee Kim
Seokhwan Seong
Jaeyeon Jang
Young-Kyoon Suh
Honghyun Kim
Dongwoo Han
Inuk Jung
author_facet Seunghwan Bae
Donghee Kim
Seokhwan Seong
Jaeyeon Jang
Young-Kyoon Suh
Honghyun Kim
Dongwoo Han
Inuk Jung
author_sort Seunghwan Bae
collection DOAJ
description An air conditioner is a household appliance and one of the most innovative inventions, primarily used to cool indoor temperatures and lower humidity to provide a comfortable environment. The sale of air conditioners increased continuously, reaching approximately 135 million units annually, with nearly 1.6 billion devices currently in operation. While it is important to devise and develop technologies for improving their energy efficiency, it is even more important to analytically investigate the performance of its main objective, user experience, or comfortableness. Here, we collected a massive set of logs collected in a longitudinal manner during the summer season of the year 2021 from approximately 1.47 million AC devices manufactured by LG Electronics and located in South Korea. Among the many features in the log, the current temperature, target temperature, humidity, energy consumption, and wind strength features were used to identify clusters embedding meaningful time-course feature patterns. We particularly focused on identifying patterns and user groups who either actively or inactively controlled the target temperature since such behavior likely indicates a discomforting air state. As a result, a total of 10 distinctive patterns and 14 representative user groups were searched. Among the 10 patterns, five showed active target temperature control and the other five inactive control. Regardless of control activity, all patterns are highly correlated with the progressing weather conditions during the summer. The feature patterns of the actively controlled group significantly differed from those of the inactively controlled group. Further investigation of the semantic difference between the groups would provide valuable research directions for improving the user experience of AC usage. It may also be used to devise an AI-driven automatic AC controller to minimize user intervention while providing an optimal air state.
format Article
id doaj-art-75c4a978ab6b41dcb6b902260dbd0a08
institution DOAJ
issn 2169-3536
language English
publishDate 2025-01-01
publisher IEEE
record_format Article
series IEEE Access
spelling doaj-art-75c4a978ab6b41dcb6b902260dbd0a082025-08-20T03:03:49ZengIEEEIEEE Access2169-35362025-01-0113561955620910.1109/ACCESS.2025.355474910942373Identifying AC Control Patterns From a Massive Longitudinal Log Dataset Using Deep ClusteringSeunghwan Bae0https://orcid.org/0009-0005-6233-4983Donghee Kim1https://orcid.org/0009-0005-2941-5220Seokhwan Seong2https://orcid.org/0009-0000-3945-733XJaeyeon Jang3https://orcid.org/0000-0002-2184-9332Young-Kyoon Suh4https://orcid.org/0000-0003-3124-2566Honghyun Kim5https://orcid.org/0009-0003-6920-0932Dongwoo Han6https://orcid.org/0009-0003-4003-4109Inuk Jung7https://orcid.org/0000-0003-0675-4244School of Computer Science and Engineering, Kyungpook National University, Daegu, Republic of KoreaSeoul National University Hospital Biomedical Research Institute, Seoul, Republic of KoreaSchool of Computer Science and Engineering, Kyungpook National University, Daegu, Republic of KoreaSchool of Computer Science and Engineering, Kyungpook National University, Daegu, Republic of KoreaSchool of Computer Science and Engineering, Kyungpook National University, Daegu, Republic of KoreaAir Solution Research and Development Laboratory, LG Electronics Home Appliance and Air Solution Company, Seongsan-gu, Changwon-si, Gyeongsangnam-do, Republic of KoreaAir Solution Research and Development Laboratory, LG Electronics Home Appliance and Air Solution Company, Seongsan-gu, Changwon-si, Gyeongsangnam-do, Republic of KoreaSchool of Computer Science and Engineering, Kyungpook National University, Daegu, Republic of KoreaAn air conditioner is a household appliance and one of the most innovative inventions, primarily used to cool indoor temperatures and lower humidity to provide a comfortable environment. The sale of air conditioners increased continuously, reaching approximately 135 million units annually, with nearly 1.6 billion devices currently in operation. While it is important to devise and develop technologies for improving their energy efficiency, it is even more important to analytically investigate the performance of its main objective, user experience, or comfortableness. Here, we collected a massive set of logs collected in a longitudinal manner during the summer season of the year 2021 from approximately 1.47 million AC devices manufactured by LG Electronics and located in South Korea. Among the many features in the log, the current temperature, target temperature, humidity, energy consumption, and wind strength features were used to identify clusters embedding meaningful time-course feature patterns. We particularly focused on identifying patterns and user groups who either actively or inactively controlled the target temperature since such behavior likely indicates a discomforting air state. As a result, a total of 10 distinctive patterns and 14 representative user groups were searched. Among the 10 patterns, five showed active target temperature control and the other five inactive control. Regardless of control activity, all patterns are highly correlated with the progressing weather conditions during the summer. The feature patterns of the actively controlled group significantly differed from those of the inactively controlled group. Further investigation of the semantic difference between the groups would provide valuable research directions for improving the user experience of AC usage. It may also be used to devise an AI-driven automatic AC controller to minimize user intervention while providing an optimal air state.https://ieeexplore.ieee.org/document/10942373/Air conditionerbig datatime seriesclusteringdeep learning
spellingShingle Seunghwan Bae
Donghee Kim
Seokhwan Seong
Jaeyeon Jang
Young-Kyoon Suh
Honghyun Kim
Dongwoo Han
Inuk Jung
Identifying AC Control Patterns From a Massive Longitudinal Log Dataset Using Deep Clustering
IEEE Access
Air conditioner
big data
time series
clustering
deep learning
title Identifying AC Control Patterns From a Massive Longitudinal Log Dataset Using Deep Clustering
title_full Identifying AC Control Patterns From a Massive Longitudinal Log Dataset Using Deep Clustering
title_fullStr Identifying AC Control Patterns From a Massive Longitudinal Log Dataset Using Deep Clustering
title_full_unstemmed Identifying AC Control Patterns From a Massive Longitudinal Log Dataset Using Deep Clustering
title_short Identifying AC Control Patterns From a Massive Longitudinal Log Dataset Using Deep Clustering
title_sort identifying ac control patterns from a massive longitudinal log dataset using deep clustering
topic Air conditioner
big data
time series
clustering
deep learning
url https://ieeexplore.ieee.org/document/10942373/
work_keys_str_mv AT seunghwanbae identifyingaccontrolpatternsfromamassivelongitudinallogdatasetusingdeepclustering
AT dongheekim identifyingaccontrolpatternsfromamassivelongitudinallogdatasetusingdeepclustering
AT seokhwanseong identifyingaccontrolpatternsfromamassivelongitudinallogdatasetusingdeepclustering
AT jaeyeonjang identifyingaccontrolpatternsfromamassivelongitudinallogdatasetusingdeepclustering
AT youngkyoonsuh identifyingaccontrolpatternsfromamassivelongitudinallogdatasetusingdeepclustering
AT honghyunkim identifyingaccontrolpatternsfromamassivelongitudinallogdatasetusingdeepclustering
AT dongwoohan identifyingaccontrolpatternsfromamassivelongitudinallogdatasetusingdeepclustering
AT inukjung identifyingaccontrolpatternsfromamassivelongitudinallogdatasetusingdeepclustering