A segmented benchmarking framework to mitigate scale bias in operational energy assessment of apartment complexes using open government data

초록

Reducing energy consumption in buildings is critical to achieving carbon neutrality; however, current performance evaluation frameworks in Korea rely predominantly on design-stage theoretical assessments, which fail to capture operational realities. This study developed a data-driven energy-benchmarking framework for apartment complexes using open government data (OGD) to transition the evaluation of building energy performance from design-based simulation to operational-stage empirical assessment. Unlike in conventional energy-use intensity (EUI)-based approaches, the annual total primary-energy consumption is adopted as the dependent variable directly capturing the scale-driven characteristics of residential complexes within a peer-group comparison framework. To control the scale-dependent structural bias inherent in aggregate energy metrics, apartment complexes are systematically segmented into four groups based on legally mandated community facility standards, which serve as proxies for complex scale and operational complexity. For each segmented group, a multiple linear regression (MLR) model was developed using physical and behavioral proxy variables derived from OGD. Group-specific models achieved a coefficient of determination (R2) ranging from 0.82 to 0.88, with a CV (RMSE) of approximately 10%. K-fold cross-validation, in which all preprocessing steps were independently repeated within each training fold, yielded a mean R2 ranging from 0.59 to 0.79, confirming reliable generalizability. A comparative analysis between the global and segmented models revealed that the unsegmented global model produces systematic grade distortions across complex scales. The proposed segmentation effectively mitigates this bias, enabling more objective and equitable operational energy assessment regardless of complex size. These findings demonstrate the practical utility of OGD-based segmented benchmarking as a scalable, transparent, and operationally grounded framework for residential building energy management and policy.

키워드