Duration: 59:03

41 views

0

Residential and commercial buildings in the US account for roughly one-third of the national energy usage and one-eight of its water use. Modest improvements in this sector can therefore have a significant impact. Technology for fine-grained real-time monitoring of in-building resource consumption can help achieve this goal by yielding information that is useful for motivating resource efficient behaviors, helping assess efficacy of resource efficiency measures and new technologies under realistic conditions, and identifying inefficiency-causing degradations in building infrastructure. Indeed, the US Green Building Council (USGBC) recommends existing buildings to have resource monitoring technology as part of its Leadership in Energy and Environmental Design for Existing Building (LEED-EB) program. Typically, however, only coarse grained information is available, making it difficult to improve the consumption efficiency. Monitoring end-point level resource consumption is difficult with available technologies because the use of expensive sensors or professional installation of in-line sensors is necessary. In addition, device aesthetic and the inherent intrusiveness of direct in-line sensors to measure the resource usage at every end-point complicate installations. This talk introduces an alternative approach for very fine grained monitoring of resource consumption in buildings. Since resource consuming end-points emit measurable signals when they are consuming resources, indirect sensors can be used for inferring real-time resource consumption. However, indirect sensors cannot be calibrated during manufacturing because of varying ambient conditions and sensor placement. Thus, the main challenge is to provide a method that autonomously calibrates the indirect sensors. This concept will be exemplified through two pilot studies: ViridiScope, a whole-house appliance-level power monitoring system using the combination of magnetic, acoustic and light sensors, and NAWMS, a whole-house pipe-level water monitoring system using vibration sensors attached to pipes. These examples demonstrate how a model-based sensor calibration framework addresses the in-situ sensor calibration problem, and fuses sensor information from various sensing modalities. The talk will end by describing how this core technology, when combined with sensing of human activity and ambient conditions, may provide even finer-grained personalized and activity level insights into our resource usage patterns.