Identification of HVAC Energy Use Patterns in Historic Buildings Using Change-Point Models and Cluster Analysis

Historic buildings account for a significant proportion of the existing building stock, yet their HVAC systems often operate inefficiently and consume substantial energy. To identify typical HVAC energy use patterns and support energy management, this study used a calibrated EnergyPlus v22.2.0 model with outdoor air temperature as the independent variable and HVAC electricity, sensible heating energy, and sensible cooling energy as the dependent variables within a change-point analysis framework. Latin hypercube sampling (LHS) was applied to five variables within predefined ranges, and 2800 parametric variants were generated from four archetype models. Base load, change-point temperature, and heating and cooling slopes were extracted, and three clustering methods—GMM, K-means, and HDBSCAN—were compared. HVAC electricity was primarily characterized by a 5P change-point model, whereas sensible heating energy and sensible cooling energy were stably represented by 3P heating and 3P cooling models, respectively. K-means performed best overall for HVAC electricity pattern recognition, and both sensible heating energy and sensible cooling energy showed advantages for K-means over HDBSCAN. Robustness analysis indicated high stability for all three energy variables, with sensible heating energy showing the best performance. These findings provide a basis for HVAC operation diagnosis and energy-saving optimization in historic buildings.