Using Robust Locally Weighted Regression with Adaptive Bandwidth to Predict Occupant Comfort

One of the main consumers of energy in buildings are the HVAC systems intended to maintain the internal environment for the comfort and safety of the occupants. Occupant satisfaction, is influenced by many different factors, including air temperature, radiant temperature, humidity, the outdoor environment, activity levels and clothing. Occupant thermal comfort is traditionally measured by the Predicted Mean Vote (PMV) metric, which estimates the expected response of the occupants on a seven point scale. PMV is a statistical measure, which holds for large populations. For small groups, however, the actual thermal comfort could be significantly different, and so energy may be wasted trying to achieve unwanted conditions. In this paper, we apply Locally Weighted Regression with Adaptive Bandwidth (LRAB) to learn individual occupant preferences based on historical reports. As an initial investigation, we attempt to do this based on just one input parameter, the internal air temperature. Using publicly available datasets, we demonstrate that this technique can be significantly more accurate in predicting individual comfort than PMV, relies on easily obtainable input data, and is much faster to compute. It is therefore a promising technique to be used as input to adpative HVAC control systems.