A cascaded deep learning framework for simultaneous non-intrusive load and occupancy monitoring using multi-channel aggregated smart meter data
<p>Non-intrusive load monitoring (NILM) and non-intrusive occupancy monitoring (NIOM) are critical for smart home management, enabling device-level energy optimization, fault detection, and improved energy efficiency, comfort, and security. However, most existing methods treat NILM and NIOM se...
محفوظ في:
| المؤلف الرئيسي: | |
|---|---|
| مؤلفون آخرون: | , , |
| منشور في: |
2025
|
| الموضوعات: | |
| الوسوم: |
إضافة وسم
لا توجد وسوم, كن أول من يضع وسما على هذه التسجيلة!
|
| الملخص: | <p>Non-intrusive load monitoring (NILM) and non-intrusive occupancy monitoring (NIOM) are critical for smart home management, enabling device-level energy optimization, fault detection, and improved energy efficiency, comfort, and security. However, most existing methods treat NILM and NIOM separately, limiting their combined potential. Traditional NILM struggles with overlapping device loads, while NIOM often relies on intrusive sensors, raising concerns about privacy, scalability, and integration. The proliferation of smart meters presents an opportunity to unify NILM and NIOM in a single, non-intrusive framework, though challenges remain in accurate device-level disaggregation and complex occupancy pattern recognition. To address this, we propose non-intrusive load and occupancy monitoring (NILOM), a dual-phase deep learning framework for real-time NILM and NIOM using multi-channel smart meter data. The framework cascades a load disaggregation network (LD-Net), a 1D sequence-to-sequence model for appliance-level load disaggregation, and an occupancy detection network (OD-Net), a 1D sequence-to-point classifier for occupancy detection from reaggregated household-level load event patterns. We also introduce the dice of energies of interest (DEOI) metric for robust load disaggregation assessment. Evaluated on the Electricity Consumption & Occupancy (ECO) dataset, NILOM outperforms domain benchmarks for both load disaggregation and occupancy detection, achieving approximately 90 % accuracy and F1-score in occupancy classification. Further improvements are realized through targeted device selection for generating household-level load event pulses after load disaggregation. Overall, NILOM provides a scalable, accurate, and privacy-preserving solution for integrated energy and occupancy management in smart homes.</p><h2>Other Information</h2> <p> Published in: Journal of Building Engineering<br> License: <a href="http://creativecommons.org/licenses/by/4.0/" target="_blank">http://creativecommons.org/licenses/by/4.0/</a><br>See article on publisher's website: <a href="https://dx.doi.org/10.1016/j.jobe.2025.113731" target="_blank">https://dx.doi.org/10.1016/j.jobe.2025.113731</a></p> |
|---|