This study examines how commercial activity patterns and regional characteristics shape electricity consumption in commercial buildings across urban centers in Busan, Republic of Korea. Building-level monthly electricity use for 2024 was collected from the national building-energy database and log-transformed to address heteroskedasticity. Commercial activity was operationalized at the building level by constructing ratios of business categories derived from permit-based industry records. To enhance interpretability and mitigate multicollinearity, approximately 90 detailed business types were first consolidated into eight functional groups and then reduced through k-means clustering into four activity clusters: neighborhood services, accommodation-office, restaurants-retail, and entertainment-medical-education. The analytical framework compared ordinary least squares (OLS) regression with hierarchical linear modeling (HLM), where buildings (Level 1) were nested within ten urban commercial centers (Level 2). The null model yielded an intraclass correlation coefficient of 0.083, indicating meaningful between-center variation and supporting a multilevel approach. Results from the preferred random-intercept model show that building height and floor area ratio are positively associated with electricity consumption, while reinforced-concrete structure is linked to lower consumption. Relative to the neighborhood-services cluster, all other activity clusters exhibit higher electricity use, with the entertainment - medical education cluster showing the strongest effect, suggesting the importance of longer operating hours and standby loads. These findings highlight that commercial-building electricity demand is jointly determined by building attributes and functional commercial activity patterns, underscoring the need for place-based energy-demand management tailored to urban center types.
Park et al. (Sat,) studied this question.