Bongchan Jeong, Heewon Choi, Jihyun Yoo, J.S. Park
Languages: English | Pages: 6 pp
Bibliographic info:
39th AIVC Conference "Smart Ventilation for Buildings", Antibes Juan-Les-Pins, France, 18-19 September 2018

Occupants control indoor environments to meet their individual needs for comfort. The control of window is the most common natural ventilation method influencing indoor environment as well as the energy use of the buildings to maintain a suitable environment. Therefore a better understanding of window control behaviour of the occupants has significant implication to enhance occupant comfort with minimal energy consumption. The objective of this study was to identify an appropriate algorithm and variables to develop a predictive model for window control. A longitudinal field measurement was performed for 10 months in 23 residential houses. Outdoor and indoor environmental conditions and window status were continuously monitored for the period. To identify an appropriate modelling algorithm, the logistic regression which is a traditional statistical method for binary data and three popular machine learning models, k-Nearest Neighbours (KNN), Random Forest (RF) and Artificial Neural Networks (ANN) were applied and compared. The result of this study reveals that the machine learning algorithms outperforms the traditional statistical regression model. Another contribution of this study is that variables influencing occupants to control window were varied in each season and from person to person. Thus, these results show the improvement of prediction with the use of machine learning-based control system.