Clustering analysis for indoor temperature and energy pattern identification through long-term monitoring data in a university classroom

Approximately 10% of Italy’s building stock is owned by local public authorities, which are generally characterized by a low level of digitalization. The lack of a common repository and data leads to poor energy efficiency, management, and Indoor Environmental Quality (IEQ). The present research, based on a post-occupancy evaluation using real-time data collection essential for verifying building performance and operation, has a twofold objective: the first is the assessment over time of the IEQ and energy efficiency under real operation conditions; the second is the characterization, through clustering methods, of indoor air temperatures and energy needs. The assessment has been performed in a representative classroom of Politecnico di Milano University using a sensor network for IEQ and energy monitoring. The data collection shows that, during the winter season, the temperature falls outside comfort class II for 50% of the occupied hours. In summer, overheating is detected for 35% of the occupied hours. The clustering analysis successfully identified daily operational patterns for both key variables, air temperature and energy. Four clusters corresponding to the winter, summer, and intermediate seasons were identified from indoor temperature data, yielding a Silhouette Score of 0.488. Concurrently, three clusters corresponding to heating, cooling, and ventilation-only modes were identified with a Silhouette Score of 0.8015. The present work confirms that continuous monitoring and clustering analysis represent a valuable methodology for pattern identification within large operational datasets, enabling the quantification of typical operational modes, thereby establishing a foundation for advanced diagnostics and appropriate control strategies.