[Back]

F. Tahmasebi, A. Mahdavi:
"Generation and evaluation of embedded probabilistic occupancy models for predictive building systems control";
Talk: ECPPM 2014 eWork and eBusiness in Architecture, Engineering and Construction, Wien, Österreich; 2014-09-17 - 2014-09-19; in: "Proceedings of the 10th European Conference on Product and Process Modelling (ECPPM2014), Vienna, Austria, 17-19 September 2014", A. Mahdavi, B. Martens, R.J. Scherer (ed.); Taylor & Francis - Balkema, 1/1/Boca Raton|London|New York|Leiden (2014), ISBN: 978-1-138-02710-7; 675 - 680.

Knowledge of occupants' presence and behavior in buildings and associated predictive models
are of central importance to the implementation efforts concerning predictive building systems control strategies.
Specifically, prediction of occupants' presence in office buildings represents a necessary condition for
predicting their interactions with building systems. Implementation of occupancy prediction models in existing
buildings can benefit from available occupancy monitoring data. Actual occupancy data can be compared
with model predictions on an ongoing basis, thus improving model reliability. In the present contribution, we
examine various options to process occupancy monitoring data toward developing probabilistic occupancy
models. These options can be described in terms of a number of related questions: What temporal horizon of
past occupancy information (days, weeks, months) shall be considered for model development? Would it be
advantageous to differentially treat individual week days? Shall model identification occur in fixed intervals
or in terms of dynamically receding horizons? To explore these questions on an empirical basis, we selected a
university campus office area, which is equipped with a monitoring infrastructure and includes a number of
open and closed offices. For this case study, the above mentioned options for occupancy presence prediction
were implemented and compared with actual occupancy information. The results facilitate a discussion of the
potential and limitations of probabilistic occupancy models intended for incorporation in the control logic of
existing buildings.

none - see english version