Paper Conference

Proceedings of Building Simulation 2021: 17th Conference of IBPSA

     

Consequences-based graphical model for contextualized occupants’ activities estimation in connected buildings

Huynh Phan 1, Thomas Recht 1, Laurent Mora 1, Stéphane Ploix 2
1 I2M Bordeaux, University of Bordeaux, CNRS, Arts et Metiers Institute of Technology, Bordeaux INP, F-33400 Talence, France
2 G-SCOP, Grenoble Institute of Technology, UMR CNRS 5272, 46 Avenue Felix Viallet, 38031 Grenoble Cedex 1, France


DOI: https://doi.org/10.26868/25222708.2021.30205
Abstract: Recently, many stochastic models have been proposed to better estimate the activities of occupants with the aim of reducing the discrepancy between estimated and actual energy consumption in residential buildings. However, these models are hybrid. They mix statistical data from different contexts that make them difficult to verify for a particular household. This contribution proposes a general approach, which takes into account particular contexts, to estimate contextualized activities, which denote activities in a specific household’s contexts. Specifically, a mobile application is developed to collect the information of activities from occupants in their house. Information Gain is used to determine major features from measured data and context information. Then, a consequences-based Bayesian Network (CBBN) with an expert structure is built to estimate contextualized occupant activities. Finally, a case study is presented with three activities cooking breakfast, cooking lunch, and cooking dinner. The results show that these activities are related not only to the time of a day but also to the movements of occupants and the usages of electrical appliances, which are involved frequently during these activities’ duration.
Keywords: contextualized activity estimation, Bayesian Network, human behavior co-simulation
Pages: 3417 - 3424
Paper:
bs2021_30205