Paper Conference

Proceedings of Building Simulation 2021: 17th Conference of IBPSA


Finding occupant reference profiles using self-organising maps for thermal performance of buildings

Aline Schaefer, Enedir Ghisi
Universidade Federal de Santa Catarina (UFSC), Brazil

Abstract: Data mining techniques are a great ally for obtaining profiles to represent occupant's behaviour in buildings. Self-Organising Map (SOM) is an efficient tool that can be used for that purpose. This study aimed to find occupant reference profiles using SOM for building performance simulations purposes. The method consists of using SOM, computer simulations and data analysis. SOM best performance was obtained with an 8x15 hexagonal mesh, reaching 100% accuracy. Three profiles were obtained. The systematic procedure proposed proved to be suitable for clusters formation and may be replicated by other researchers using other data samples.
Keywords: Occupant reference profile, cluster analysis, SOM Network, Building thermal performance, social housing
Pages: 3505 - 3512