A multi-sensor based occupancy estimation model for supporting demand driven HVAC operations

Heating, ventilation, and air conditioning (HVAC) is a major energy consumer in buildings, and implementing demand driven HVAC operations is a way to reduce HVAC related energy consumption. This relies on the availability of occupancy information, which determines peak/off-hour modes that impact cooling/heating loads of HVAC systems. This research proposes an occupancy estimation model that is built on a combination of non-intrusive sensors that can detect indoor temperature, humidity, CO2 concentration, light, sound and motion. Sensor data is processed in real time using a radial basis function (RBF) neural network to estimate the number of occupants. Field tests carried out in two shared lab spaces for 20 consecutive days report an overall detection rate of 87.62% for self-estimation and 64.83% for cross-estimation. The results indicate the ability of the proposed system to monitor the occupancy information of multi-occupancy spaces in real time, supporting demand driven HVAC operations.