Structured Temporal Representation for NILM in Complex Scenarios With Limited Data
IEEE Transactions on Instrumentation and Measurement · 2026 · 75 · 9005315-9005315
DOI: 10.1109/tim.2026.3704179 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Transactions on Instrumentation and Measurement, 2026. Verfasst von Pengfei Hao und 2 weiteren Autor:innen. Die Arbeit nennt 46 Literaturangaben. Gefördert von: Hunan Province Key Research and Development Program.
Abstract (Originalsprache)
In nonintrusive load monitoring (NILM), achieving accurate disaggregation in residential and industrial scenarios is challenging due to highly variable load patterns and the limited availability of instrumented training data. To address these issues, this article proposes a structured temporal representation framework for NILM that explicitly models trend, periodic, and residual characteristics of aggregate power signals within a unified architecture. The proposed framework enhances interpretability and stability in data-scarce environments by jointly learning long-term trends and repetitive temporal behaviors instead of relying solely on black-box mappings. To better reflect uncertainty in complex load behaviors, disaggregation ranges are constructed from data-driven temporal variability, and a refinement module further adjusts the disaggregated load within the feasible range while maintaining physical consistency. Experiments on residential and industrial datasets show that the proposed method achieves competitive disaggregation performance under limited-data conditions. Notably, for periodic appliances such as the fridge, the model achieves performance comparable to or better than models trained on full house datasets despite using only a small fraction of the data. In addition, deployment experiments on a Raspberry Pi 5 further validate the feasibility of the proposed framework under real measurement conditions.
Förderung
- Hunan Province Key Research and Development Program
Zitieren
APA
Hao, P., Yan, Z., & Wen, H. (2026). Structured Temporal Representation for NILM in Complex Scenarios With Limited Data. IEEE Transactions on Instrumentation and Measurement, 75, 9005315-9005315. https://doi.org/10.1109/tim.2026.3704179
Vancouver
Hao P, Yan Z, Wen H. Structured Temporal Representation for NILM in Complex Scenarios With Limited Data. IEEE Transactions on Instrumentation and Measurement. 2026;75:9005315-9005315. doi:10.1109/tim.2026.3704179
BibTeX
@article{Hao2026,
author = {Hao, Pengfei and Yan, Zhongzong and Wen, He},
title = {Structured Temporal Representation for NILM in Complex Scenarios With Limited Data},
journal = {IEEE Transactions on Instrumentation and Measurement},
year = {2026},
volume = {75},
pages = {9005315--9005315},
doi = {10.1109/tim.2026.3704179}
}
EUVIMED