Occupancy Probability Estimation in High-Rise Residential Buildings using Composite Electricity and Heating Use

Determining occupancy status is crucial for effective facility control, particularly for optimizing thermal comfort and energy efficiency. However, occupancy detection based solely on baseline electrical energy values tends to be less accurate in residential buildings that do not rely on electrical heating during winter. This study aims to assess occupancy status and extract occupant schedules by analyzing energy consumption data from high-rise residential buildings. We collected energy data from actual apartments during the winter season and employed Fast Fourier Transform for correlation analysis with electricity usage. The results revealed that heating energy had the strongest correlation with electrical energy, yielding a correlation coefficient of 0.72. For data exhibiting low occupancy probabilities, we applied unsupervised learning models, specifically the Gaussian Mixture Model and Hidden Markov Model, to adjust existing occupancy probabilities using available heating energy data, ultimately calculating refined final occupancy probabilities.