DOE OSTI · 1651384
Sensor-Based Energy Modeling (sBEM) D4: Final report
Abstract
Four residential demonstration homes have been built as part of the ZEBRAlliance and fully instrumented with 50+ sensors collecting 1-minute data and 190+ sensors collecting 15-minute data (over 250 sensors total) with the first year’s worth of data for two houses collected in January, 2011. This sensor data will be mined to develop predictive models of component-level and whole-building energy consumption by utilizing machine-learning techniques in a way that circumvents problems with typical white-box modeling approaches (faulty/time-consuming input building models, energy simulation assumptions, and contractor variability in as-built results). Such a system would constitute an objective, quantitative framework whereby raw sensor data is utilized to determine both a prediction, and thus a variance (i.e. within 10% of predicted 80% of the time), for energy consumption by a residential building. Furthermore, analysis will be conducted to determine the predictive accuracy for a smaller number of sensors.
Explore related subjects
Keep this discovery
Explore connections, maps & timelines
Edwards, Richard E., Parker, Lynne Edwards, New, Joshua Ryan. 2011-09-30. Sensor-Based Energy Modeling (sBEM) D4: Final report. https://doi.org/10.2172/1651384
Cite the original work for its findings. Save a collection to share your selection of sources.