TY - JOUR
T1 - Physical and JIT model based hybrid modeling approach for building thermal load prediction
AU - Iino, Yutaka
AU - Murai, Masahiko
AU - Murayama, Dai
AU - Motoyama, Ichiro
PY - 2011
Y1 - 2011
N2 - Energy conservation in building fields is one of the key issues in environmental point of view as well as that of industrial, transportation and residential fields. The half of the total energy consumption in a building is occupied by HVAC (Heating, Ventilating and Air Conditioning) systems. In order to realize energy conservation of HVAC system, a thermal load prediction model for building is required. This paper propose a hybrid modeling approach with physical and Just-in-Time (JIT) model for building thermal load prediction. The proposed method has features and benefits such as, (1) it is applicable to the case in which past operation data for load prediction model learning is poor, (2) it has a self checking function, which always supervises if the data driven load prediction and the physical based one are consistent or not, so it can find if something is wrong in load prediction procedure, (3) it has ability to adjust load prediction in real-time against sudden change of model parameters and environmental conditions. The proposed method is evaluated with real operation data of an existing building, and the improvement of load prediction performance is illustrated.
AB - Energy conservation in building fields is one of the key issues in environmental point of view as well as that of industrial, transportation and residential fields. The half of the total energy consumption in a building is occupied by HVAC (Heating, Ventilating and Air Conditioning) systems. In order to realize energy conservation of HVAC system, a thermal load prediction model for building is required. This paper propose a hybrid modeling approach with physical and Just-in-Time (JIT) model for building thermal load prediction. The proposed method has features and benefits such as, (1) it is applicable to the case in which past operation data for load prediction model learning is poor, (2) it has a self checking function, which always supervises if the data driven load prediction and the physical based one are consistent or not, so it can find if something is wrong in load prediction procedure, (3) it has ability to adjust load prediction in real-time against sudden change of model parameters and environmental conditions. The proposed method is evaluated with real operation data of an existing building, and the improvement of load prediction performance is illustrated.
KW - Just in time modeling
KW - Physical model
KW - Thermal load prediction
UR - http://www.scopus.com/inward/record.url?scp=80052766050&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=80052766050&partnerID=8YFLogxK
U2 - 10.1541/ieejeiss.131.1431
DO - 10.1541/ieejeiss.131.1431
M3 - Article
AN - SCOPUS:80052766050
VL - 131
SP - 1431
EP - 1438
JO - IEEJ Transactions on Electronics, Information and Systems
JF - IEEJ Transactions on Electronics, Information and Systems
SN - 0385-4221
IS - 8
ER -