In the context of the preservation of natural resources, many research and development efforts have been made to reduce the energy consumption of buildings for professional and domestic use. The use of Information and Communication Technologies (ICT) allows to bring new tools to the understanding of these consumption, by adapting the production of electricity to the consumption, but also by informing the user of the most important consumption items so that he can adapt his behavior. Indeed, “the cheapest energy is the one we do not consume”. It has been shown that the simple fact of communicating this information to users could reduce the consumption of a building by up to 15%.


The Non-Intrusive Load Monitoring (NILM) was invented by George W. Hart in 1992. It consists in analyzing the power variations observed on the electrical circuit of a building in order to deduce, by signal processing algorithms, the operating status of the different devices of the building, as well as their individual consumption in real time.
This system is lightweight and non-intrusive because it not need direct measurement on any specific device (such a power meter in every plugged device studied), but only a global power measurement.

NILM systems are a very interesting as research or technological topic: the issues surrounding NILM are not all resolved, and research is still very active on this subject. Development of this technology will have heavy economic impacts with promising capabilities in energy saving and them strong environmental preserving potential. NILM address a true need in Smart Building marked.


Indeed, NILM techniques currently show very good results for domestic installations: these generally have a limited number of devices, rare simultaneous activations, and very few identical devices installed in multiple copies. These three elements make it easier to disaggregate electrical usage in these installations.


However, Office buildings as the actual host of SmartSense platform introduce more issues in NILM techniques:

  • Lot of device to disaggregate
  • Important proportion of continues state changing devices (computers)
  • Many copies of a same device type
  • Large users increasing probabilities of simultaneous device activations


For these reasons, additional information is needed to perform the disaggregation of electrical uses with acceptable accuracy.

Our solution

The SmartSense platform allows us to get access to a large spectrum of information about the context of the experiment with a good sampling rate. After raw data processing locally in sensor nodes, this information will be available :

  • room occupancy probability
  • Estimated number of people in room
  • Luminosity in room
  • sound envelop in room

We propose to use these data of SmartSense platform to help disaggregation algorithm. The aim is to deduct individuals state probabilities for every device by coupling with this data to classic NILM technologies.

La solution que nous proposons est d’utiliser les données ci-dessus de la plateforme SmartSense pour aider les algorithmes de désagrégation. L’objectif est donc de déduire de ces informations des probabilités individuelles d’état pour chaque appareil, et en coupler celles-ci aux techniques classiques de NILM. More information can be found in the “publication” section of our website.

Comments are closed.