Change-point model-based clustering for urban building energy analysis

Achieving carbon neutrality by 2050 requires evaluating and retrofitting existing buildings. However, despite the numerous studies on energy analytics, they usually focus on energy consumption patterns and motifs rather than encompassing various energy usage characteristics. This study proposes a novel symbolic hierarchical clustering for building energy analytics at the city level. It utilizes change-point model (CPM) parameters to represent building energy usage, performance, occupant behavioral characteristics. The clustering method based on the CPM parameters defines energy performance signatures (EPS) for determining their energy characteristics and as symbolic data transformation. In a case study conducted in Gangwon, South Korea, five different energy performance signatures (EPSs 1–5) showing their unique energy characteristics were determined for commercial buildings. EPS1 to 3 were classified as signatures with good performance (65.5% of all buildings) while EPS4 and 5 were classified as signatures with bad performance (34.5%). Using this EPS symbolic data, an EPS map was visualized and analyzed from various perspectives. For example, buildings that showed a continuous or overall decline in envelope performance over five years were among the oldest buildings (construction completion date closer to 1978; 7.9%). Despite poor envelope performance, buildings with lower energy usage showed a tendency for occupants to delay heating (28.4%). The proposed method can contribute to the data-driven building energy analytics in providing detailed insights into energy usage patterns, building energy performance, and occupant behavioral characteristics at the city level. The effectiveness of open-source energy data for urban building energy analysis would be improved through the proposed method.