What is it about?
NILM infers the consumption of electrical devices in a household based on active power measurements from the mains meter. We found that additional, correlated data can improve these inferences, either by adding whole-home water consumption or reactive power measurements to the models.
Featured Image
Photo by engin akyurt on Unsplash
Why is it important?
NILM still has room for improvement in performance, and the proposed additional data sources are comparatively easy to obtain: reactive power is often measured by the mains meter anyway and could be integrated to the machine learning models in a very simple manner. Similarly, whole-home water consumption can be gathered using just one additional meter and in turn helps a lot with the performance for devices correlated to water use.
Perspectives
We thought about ways to improve the performance of NILM algorithms in a practical manner that does not require a lot of complicated additional sensor setups or redesigning of existing machine learning architectures. This way, our insights can hopefully benefit the community more.
Justus Breyer
Read the Original
This page is a summary of: On the Benefits of Water Consumption and Reactive Power for Eventless NILM, June 2026, ACM (Association for Computing Machinery),
DOI: 10.1145/3765611.3815510.
You can read the full text:
Resources
Contributors
The following have contributed to this page
