TY - GEN
T1 - Evaluate the Energy Saving from Window Opening Behavior Through Coupling a Deep Learning Model with EnergyPlus
AU - Pandey, Pratik
AU - Dong, Bing
AU - Sharifi, Nina
AU - Malsegna, Mason Kenneth
N1 - Publisher Copyright:
© 2023, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.
PY - 2023
Y1 - 2023
N2 - Buildings consume about 40% of the electricity generated and significantly burden the energy sector. A new concept of net-zero energy building has emerged to ease this burden. The preliminary design relies heavily on the simulation data from building simulation engines like Energy Plus. Research has pointed out about a 30% difference in average between actual and simulated energy consumption. One of the primary reasons for this discrepancy is pointed towards the use of schedule-based occupant behavior (OB) during simulation. This conventional schedule should be replaced by a model that could predict the occupants’ actions reasonably. Binomial Logistic Regression and Neural Network architectures were developed and tested for their prediction accuracy of window's state in the residential dorms of a local university at Syracuse, NY. The neural network models outperformed the Binomial logistic regression model by a considerable margin when both of them were tested in entirely different homes. We found that the Deep Neural Network model with ‘Adam’ optimizer and ridge regularization parameter of 0.01 performed the best to predict the state of the windows when the learning rate was 0.01. These models were then integrated into the Energy Plus (v 9.6) using newly released E+ Python API and tested in a single zone of the campus building. Using the ANN-driven schedules, we observed that the total sensible monthly heating load requirement for the month of September is reduced by 38.56%, and the total sensible cooling load requirement is reduced by 55.52%.
AB - Buildings consume about 40% of the electricity generated and significantly burden the energy sector. A new concept of net-zero energy building has emerged to ease this burden. The preliminary design relies heavily on the simulation data from building simulation engines like Energy Plus. Research has pointed out about a 30% difference in average between actual and simulated energy consumption. One of the primary reasons for this discrepancy is pointed towards the use of schedule-based occupant behavior (OB) during simulation. This conventional schedule should be replaced by a model that could predict the occupants’ actions reasonably. Binomial Logistic Regression and Neural Network architectures were developed and tested for their prediction accuracy of window's state in the residential dorms of a local university at Syracuse, NY. The neural network models outperformed the Binomial logistic regression model by a considerable margin when both of them were tested in entirely different homes. We found that the Deep Neural Network model with ‘Adam’ optimizer and ridge regularization parameter of 0.01 performed the best to predict the state of the windows when the learning rate was 0.01. These models were then integrated into the Energy Plus (v 9.6) using newly released E+ Python API and tested in a single zone of the campus building. Using the ANN-driven schedules, we observed that the total sensible monthly heating load requirement for the month of September is reduced by 38.56%, and the total sensible cooling load requirement is reduced by 55.52%.
KW - Artificial neural network
KW - EnergyPlus Python API
KW - EnergyPlus machine learning integration
KW - Window opening schedule
UR - http://www.scopus.com/inward/record.url?scp=85172728988&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85172728988&partnerID=8YFLogxK
U2 - 10.1007/978-981-19-9822-5_102
DO - 10.1007/978-981-19-9822-5_102
M3 - Conference contribution
AN - SCOPUS:85172728988
SN - 9789811998218
T3 - Environmental Science and Engineering
SP - 959
EP - 969
BT - Proceedings of the 5th International Conference on Building Energy and Environment
A2 - Wang, Liangzhu Leon
A2 - Ge, Hua
A2 - Ouf, Mohamed
A2 - Zhai, Zhiqiang John
A2 - Qi, Dahai
A2 - Sun, Chanjuan
A2 - Wang, Dengjia
PB - Springer Science and Business Media Deutschland GmbH
T2 - 5th International Conference on Building Energy and Environment, COBEE 2022
Y2 - 25 July 2022 through 29 July 2022
ER -