Feature Identification for Non-Intrusively Extracting Occupant Energy-Use Information in Office Buildings
Detailed energy-use information of office buildings’ occupants is necessary to implement proper simulation/intervention techniques. However, acquiring accurate occupant-specific energy consumption in office buildings at low cost is currently a challenging task since existing intrusive load monitoring (ILM) technologies require a large capital investment to provide high-resolution electricity usage data for individual occupants. On the other hand, non-intrusive load monitoring (NILM) approaches have been proven as more cost effective and flexible approaches to provide energy-use information of individual appliances. Therefore, extending the concept of NILM to individual occupants would be beneficial. This paper proposes two occupancy-related energy-consuming features, delay interval and magnitude of power changes and evaluates their significances for extracting occupant-specific power changes in a non-intrusive manner. The proposed features were examined through implementing a logistic regression model as a predictor on aggregate energy load data collected from an office building. Hypotheses tests also confirmed that both features are statistically significant to non-intrusively derive individual occupants’ energy-use information. As the main contribution of this study, these features could be utilized in developing sophisticated NILM-based approaches to monitor individual occupant energy-consuming behavior.