A data mining approach to evaluating the impact of occupancy levels on energy performance: A case study of a university educational building

Educational buildings, including Classrooms, Study Halls, Research Laboratories, and departmental administrative offices are substantial contributors to energy consumption due to intensive utilization throughout the academic year. However, a considerable portion of this energy is wasted, primarily attributes to the inadequate awareness among occupants regarding energy-saving practices and the uninterrupted operation of building energy systems, irrespective of actual occupancy levels or demand patterns. This paper investigated the impact of occupant density on energy consumption for air conditioning, with a focus on gas consumption by a gas heat pump (GHP) system in an educational facility. The daily gas consumption per occupant-hour was calculated for the analysed space throughout the academic year. Hourly occupancy data was primarily gathered through field surveys, corresponding to specific space usage patterns to ensure accuracy. A hybrid model combining Gaussian Mixture Model (GMM)-based clustering with Bayesian Hidden Markov Models (HMMs) was developed to estimate hourly occupant numbers, grounded in real-time electricity consumption as a key variable. As a result, the analysis of air conditioning-related energy consumption indicated that optimizing load scheduling during low-occupancy periods could contribute to reducing the overall energy consumption of the educational building.