V. M. Zavala, D. Skow, T. Celinski, and P. Dickinson, "Techno-Economic Evaluation of a Next-Generation Building Energy Management System," Technical Memorandum ANL/MCS-TM-313, May 2011. [pdf]
We perform a technological evaluation of BuildingIQ's next-generation energy management (EM) system and present a preliminary energy savings analysis for the Theory and Computing Sciences building at Argonne National Laboratory. The EM system uses a model predictive control framework that incorporates black-box machine learning dynamic models. The system enables the optimization of operational set-points for the HVAC system and for the building zones in a proactive manner by foreseeing weather conditions and electricity prices. In addition, it can foresee dynamic limitations arising from the thermal mass of the building, can perform proactive demand-response, and can trade off between electricity cost, comfort satisfaction, and CO[sub 2] emissions resulting from energy consumption. The black-box models are learned in real-time by using available sensor data and do not require building and HVAC topology information. This approach enables short deployment times, low installation and maintenance costs, and widespread deployment. Our preliminary studies at the TCS building indicate that significant amounts of energy can be saved by exploiting the ambient air to delay the start-up of the HVAC system and by exploiting unoccupied periods to relax the zone set-points. We analyze how to use the system to achieve energy savings of up to 45% with minimal impact on comfort conditions. We also identify several technological limitations of the EM system that restrict its performance and we propose directions of future research.