TY - GEN
T1 - Collaborative recommender systems for building automation
AU - LeMay, Michael
AU - Haas, Jason J.
AU - Gunter, Carl
PY - 2009/4/3
Y1 - 2009/4/3
N2 - Building Automation Systems (BASs) can save building owners money by reducing energy consumption while simultaneously preserving occupant comfort. There are algorithms that optimize this tradeoff, such as detecting which appliances are turned on without requiring expensive status detectors to be attached to each appliance. However, better ways are needed to determine which algorithms are best-suited to a particular building. This paper explores the idea of allowing building managers to automatically communicate among themselves and exchange ratings of individual monitoring and control algorithms in such a way that each building manager can then obtain predicted ratings for all algorithms that he has not yet tried personally. We allow individual algorithms to be replaced by using a blackboard architecture to loosen the coupling between them. We propose a recommender system that operates on a database of contributed ratings to predict ratings of untried algorithms. To explore this approach, we developed a prototype that seamlessly interacts with both emulated physical buildings and buildings simulated in software and we implemented several of the control algorithms described in previous works. We demonstrate a recommender system that selects between algorithms in various types of buildings.
AB - Building Automation Systems (BASs) can save building owners money by reducing energy consumption while simultaneously preserving occupant comfort. There are algorithms that optimize this tradeoff, such as detecting which appliances are turned on without requiring expensive status detectors to be attached to each appliance. However, better ways are needed to determine which algorithms are best-suited to a particular building. This paper explores the idea of allowing building managers to automatically communicate among themselves and exchange ratings of individual monitoring and control algorithms in such a way that each building manager can then obtain predicted ratings for all algorithms that he has not yet tried personally. We allow individual algorithms to be replaced by using a blackboard architecture to loosen the coupling between them. We propose a recommender system that operates on a database of contributed ratings to predict ratings of untried algorithms. To explore this approach, we developed a prototype that seamlessly interacts with both emulated physical buildings and buildings simulated in software and we implemented several of the control algorithms described in previous works. We demonstrate a recommender system that selects between algorithms in various types of buildings.
UR - http://www.scopus.com/inward/record.url?scp=78650760810&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=78650760810&partnerID=8YFLogxK
U2 - 10.1109/HICSS.2009.114
DO - 10.1109/HICSS.2009.114
M3 - Conference contribution
AN - SCOPUS:78650760810
SN - 9780769534503
T3 - Proceedings of the 42nd Annual Hawaii International Conference on System Sciences, HICSS
BT - Proceedings of the 42nd Annual Hawaii International Conference on System Sciences, HICSS
T2 - 42nd Annual Hawaii International Conference on System Sciences, HICSS
Y2 - 5 January 2009 through 9 January 2009
ER -