1. Introduction
Non-intrusive load monitoring allows appliance-level energy insights from a single point of measurement at the meter, avoiding the cost of per-appliance sub-metering, and NB-IoT extends this capability to rural and semi-urban residential areas with poor WiFi or cellular data coverage.
2. Methodology
A smart meter prototype sampled aggregate current and voltage at 4kHz and applied a lightweight edge-based event detection algorithm to identify appliance on/off transitions and classify appliance type from transient current signatures, reporting 15-minute aggregated appliance-level summaries over an NB-IoT connection rather than raw high-frequency waveforms, deployed across 25 households for an eight-week field trial.
3. Results
The on-device event-based classifier achieved 88.4 percent identification accuracy averaged across six common appliance types (refrigerator, water heater, washing machine, air conditioner, television, and lighting circuits), with average system power draw of 38mW, projecting 14 to 18 months of operation from a single 10,000mAh battery pack depending on household appliance-switching frequency.
4. Conclusion
Combining edge-based non-intrusive load monitoring with NB-IoT reporting enables appliance-level residential energy insight in coverage-limited areas without the cost of per-appliance sensors. Future work will extend the appliance signature library to cover a broader range of household equipment.
References
[1] Hart G. W., Nonintrusive appliance load monitoring, Proceedings of the IEEE, 1992. [2] Ratasuk R. et al., NB-IoT system for M2M communication, IEEE WCNC, 2016.