Publication Type
Conference Proceeding Article
Version
publishedVersion
Publication Date
11-2023
Abstract
In the energy domain, the classification of power meters has become an increasingly significant area of interest, such as appliance identification and characteristics prediction, enabling targeted and efficient energy management. However, the limited availability of labeled data for power meters and the inconsistencies in labeling and naming conventions have constrained the potential of metadata for further application. This study aims to bridge the gap by employing semi-supervised Generative Adversarial Networks (SGAN) to classify 1805 power meters distributed globally. This approach explores and assesses the advantages of incorporating unlabeled power meter data into the learning process. A comparative analysis is performed between supervised and semi-supervised baseline models using different proportions of labeled data. The results reveal that SGAN can achieve an accuracy rate exceeding 0.8 with just 100 labeled samples, whereas a Two-dimensional Convolution Neural Network (2D-CNN) requires a minimum of 300 samples to attain the same performance level. The innovative contribution of this study lies in formulating a refined classification model and a label propagation method that optimizes the use of unlabeled energy meter data. Additionally, the classification framework established in this study has the potential for expanded application in categorizing other metadata.
Keywords
Building energy, Smart meter, Semi-supervised learning, Classifica-tion, Generative models
Discipline
Energy Policy | Urban Studies
Research Areas
Integrative Research Areas
Publication
BuildSys '23: Proceedings of the 10th ACM International Conference on Systems for Energy-Efficient Buildings, Cities, and Transportation, Istanbul, Turkey, November 15-16
First Page
450
Last Page
453
Identifier
10.1145/3600100.3626633
Publisher
ACM
City or Country
New York
Citation
FU, Chun; KAZMI, Hussain; QUINTANA, Matias; and MILLER, Clayton.
Enhancing classification of energy meters with limited labels using a semi-supervised generative model. (2023). BuildSys '23: Proceedings of the 10th ACM International Conference on Systems for Energy-Efficient Buildings, Cities, and Transportation, Istanbul, Turkey, November 15-16. 450-453.
Available at: https://ink.library.smu.edu.sg/cis_research/595
Creative Commons License
This work is licensed under a Creative Commons Attribution-NonCommercial-No Derivative Works 4.0 International License.
Additional URL
https://doi.org/10.1145/3600100.3626633