Non-intrusive load monitoring for water (WaterNILM)

Abstract

Better water consumption decisions benefit from detailed use information. Easily installed non-intrusive vibration sensors provide a "no-fuss" retrofit solution for detecting the operation of water consuming appliances. The sensors measure pipe vibration, which are revealed to be a rich source of information for identifying loads. Vibration is processed to extract power spectral density based features which are classified with a clustering algorithm trained during install. The results can be used to track load operating schedule from the vibration data collected from as little as one pipe in a home. Mechanics governing the observed signals, and signal processing to extract operating information, are discussed in this paper. Field data from three different homes demonstrates the accuracy of this approach. © 2014 WIT Press.