Recognizing energy-related activities using sensors commonly installed in office buildings

Automated control based on user activities and preferences could reduce energy consumption of office buildings. In this paper, we investigated generalisation properties of an office activity recognition approach using sensors that are frequently installed in modern and refurbished office buildings. In particular, per-desk passive infrared (PIR) sensors and power plug meters were considered in an evaluation study including more than 100 hours of data from both, a single-person room and a three-user multi-person office room. Layered hidden Markov models (LHMM) were used for the recognition. Results showed that 30 hours and 50 hours of training data were needed to achieve robust recognition of desk activities and estimate people count, respectively. The recognition can be performed independent of a particular occupant desk. In further simulations considering different energy profiles, we show how energy consumption due to lighting and office appliances is related to occupant behaviour.