A systematic approach to occupancy modeling in ambient sensor-rich buildings

摘要: With ever-rising energy demand and diminishing sources of inexpensive resources, conservation has become an increasingly important topic. Building heating, ventilation, air conditioning (HVAC) systems are considered to be a prime target for due their significant contribution commercial buildings' consumption in the US. Knowing building's occupancy plays crucial role implementing demand-response HVAC controls, with corresponding potential reduction consumption, especially office buildings. This paper evaluates modeling (both binary detection multi-class estimation) using twelve ambient sensor variables. Performance six machine-learning techniques is evaluated both single-occupancy multi-occupancy offices. Of six, decision-tree technique yielded best overall accuracy (i.e. 96.0% 98.2%) root mean square error (RMSE) 0.109 0.156). The each individual variable via information gain. It found that CO2, door status, light variables have contributions final results. observed generally increases as number sensors increases. also examines possibility building global model, explores reasons low performance estimation. Lastly, model used estimate visualize accumulative room thermal zone usage test-bed three months. results reveal effective vacancy accounts substantial portion operational hours, varying from 19.8% 29.8% average 23.3%, which bears savings. Furthermore, authors simulated months DesignBuilder EnergyPlus, compared occupancy-based controls authors' occupancy-modeling conventional currently implemented building. demonstrate 20% gas 18% electricity could effectively saved if control implemented.