Deltaflow: Submetering by Synthesizing Uncalibrated Pulse Sensor Streams

Abstract

Current submetering systems suffer from prohibitive device costs, invasive installations, and burdensome maintenance. In this paper we present Deltaflow, a submetering system that can estimate the power draw of individual loads by augmenting aggregate measurements with very simple sensors. The key insight is that we can drastically reduce sensor complexity by encoding information in the mere existence of a radio transmission, rather than the contents of that transmission. A sensor consisting simply of a radio and an energy-harvesting power supply tuned to harvest a side-channel emission of energy consumption (e.g. light, heat, magnetic field, vibration) will exhibit an activation frequency that is correlated with the power draw of the load to which it is affixed. These sensors report their activations to the data-processing backend, which can determine the actual power draw by incorporating ground truth aggregate measurements such as those provided by utility meters. The server maps sensor activations to energy consumption by observing when the aggregate measurement and the sensor activation frequency change simultaneously. The server iteratively partitions the system history into discrete states which are used to construct and solve instances of a linear optimization problem. Solutions to the problem reveal the mapping from pulse frequencies to individual load power draw. This systems approach to submetering results in deployments that are easy to install and maintain, while contributing zero additional load, enabling building owners and occupants to simply affix tags to energy consumers and automatically begin receiving real-time power draw readings.

Publication
Proceedings of the 5th International Conference on Future Energy Systems

This was sweet.

Dr. Meghan Clark
Dr. Meghan Clark
IoT Research Scientist

My research interests include sensor networks, radio communications, and network monitoring.