Building occupancy prediction using machine learning and existing measurement data
The research aims to develop prediction models of the building usage and their spaces using machine learning and typical existing data streams from building automation systems (BAS).
The occupant number and usage rate varies significantly between office buildings. Moreover, the space usage divides unevenly within buildings. The building usage information rarely exists without to mention the usage forecast. A building usage highly affects in user comfort, indoor environment quality, maintenance costs and energy consumption. Usage information and the forecast can help to understand user needs better, improve user satisfaction, increase space efficiency and allocate maintenance resources.
Even now, numerous buildings are of sufficiently modern type that include several sensors measuring indoor conditions and the HVAC device operation. By using this data and machine learning, it is possible to estimate the space usage. Furthermore, usage information enables creating a prediction model that uses occupant history and other relevant data. These methods facilitate to predict usage with greatly less cost and more private than with, for example, separate camera based systems that purpose is in occupant calculation.
The research aims to develop machine learning and data based prediction model for building usage. The research studies the viability of BAS data for building usage prediction and the prediction model accuracy. The study will be conducted in real building environment with the data collected from real office buildings.