Exploring Sequential and Association Rule Mining for Pattern-based Energy Demand Characterization

  • Authors:
  • Leneve Ong;Mario Bergés;Hae Young Noh

  • Affiliations:
  • Department of Civil and Environmental Engineering, Carnegie Mellon University;Department of Civil and Environmental Engineering, Carnegie Mellon University;Department of Civil and Environmental Engineering, Carnegie Mellon University

  • Venue:
  • Proceedings of the 5th ACM Workshop on Embedded Systems For Energy-Efficient Buildings
  • Year:
  • 2013

Quantified Score

Hi-index 0.00

Visualization

Abstract

The relationship between occupant activity and electricity consumption is inextricably linked. It has been difficult to both gather detailed energy data and information about occupants' daily lives as well as understand their relationship quantitatively. There is significant past work on activity recognition in homes and load prediction, but there is limited understanding of how activities can inform consumption or vice versa. Our work begins by characterizing power data as provided by plug-level meters from one household. Association and sequential rule mining techniques are applied to extract explicit rules that may be useful for forming the basis of demand patterns. Initial findings include the identification of device groups but highlight the challenges of modeling complex patterns and event rarity.