Paper Conference

Proceedings of Building Simulation 2021: 17th Conference of IBPSA


Development of Synthetic Ageing Profiling for Occupant Behaviour

Yuezhong Liu, Yin-Leng Theng
Nanyang Technological University, Singapore

Abstract: Occupant behaviour is one major factor influencing energy use in residential buildings, which is also impacted by sociodemographic such as age. A better understanding of age-behaviour-building interactions enables to describe with higher accuracy the energy consumption as the ageing population increase rapidly. However, the collection of real occupant behaviour faces difficulties in privacy issues and various sensors with relatively high costs. This pilot study proposed to develop synthetic ageing profiling for occupant behaviour through the Generative Adversarial Network (GAN), which is capable of predicting electricity consumption by simulation tool (EnergyPlus) during the design process. We validated the results with 600 household profiles with two-generational cohorts (40-64 years and 65 years and above) in Singapore. The Kullback-Leibler divergence of the synthetic and real profiling are within 0.46 for a majority of parameters and clusters. The synthetic results will help the building energy performance research for an ageing population.
Keywords: Synthetic Ageing Occupant Behaviour, Generative Adversarial Network, Machine Learning, Energy Simulation
Pages: 3498 - 3504