Deep Learning in Energy Management: Non-Intrusive Load Monitoring for IoT Devices
summary
Claims that deep models can do NILM on low‑volume irregular IoT data using the Chain2 protocol and that training on properly simulated Chain2 samples improves accuracy. Uses a CNN encoder for multi‑label classification on one‑minute inputs (eight‑hour chunks). Evaluations use public clamp datasets (UK, REFIT) with 18 houses—15 houses training 3 testing—and report that event/quarter‑hour framing cuts traffic from 86k vs 2k messages per day, reducing cost while keeping usable disaggregation performance.
AI-written summary · about
chapters (11)
- 0:00Introduction
- 1:10Speaker background
- 3:02Non-Intrusive Load Monitoring overview
- 8:07NILM benefits
- 10:59Smart metering and IoT
- 14:35Chain 2 device data
- 18:36Algorithms for Non-Intrusive Load Monitoring
- 21:51Chain 2 data challenges
- 25:50Experiments and evaluation
- 26:31Summary
- 27:15Q&A
related talks
57mMachine Learning in the Age of IoT
59mML and the IoT: Living on the Edge
06mIntermittent Learning: On-Device Machine Learning on Intermittently Powered System
15mFrom Sensing to Understanding: LoRaWAN + LLMs
52mMachine Learning in IoT Applications
40mDeep Learning and Time Series Forecasting for Smarter Energy
50mApplying Machine Learning to IoT Use Cases with Deeplearning4j
01h 01mML and the IoT: Living on the Edge