商業ビルエネルギー運用における暖房・換気・空調関連炭素排出ドライバーの代理変数による特定

Unofficial AI-generated summary based on the public title and abstract. Not an official translation.

🔬研究者:運用混乱期における排出ドライバーのフェーズ依存性を、代理変数と段階的回帰で捉える手法を提供する。

🏢実務担当者:データが限られた商業ビルで、月次エネルギー・気象データからHVAC排出の変動要因を事後的に把握する際の参考になる。

🏛政策担当者:建物単位の炭素会計・報告制度設計において、運用段階の排出変動をどう扱うかの論点を提示する。

Abstract Energy conservation and carbon reduction in commercial buildings are essential for achieving carbon neutrality targets, but carbon emission drivers may change under operational disruptions. This study investigates phase-dependent HVAC-related carbon emission drivers in a large shopping mall using 23 monthly observations of energy consumption and meteorological data from February 2021 to December 2022. To avoid mechanical overlap between total electricity consumption and HVAC-related carbon emissions, non-HVAC electricity consumption was constructed as a proxy for operational intensity. Four progressive regression models were developed, including a temperature-only baseline model, a proxy-augmented model, a phase-dummy model, and a phase-interaction model. The results show that the phase-interaction model achieved the best in-sample explanatory performance, with the adjusted R² increasing from 0.3705 to 0.9641 and the RMSE decreasing from 231.2226 to 49.7036 compared with the temperature-only model. The marginal effect of temperature increased from 10.3703 in the restricted phase to 26.0975 in the recovery phase, indicating a stronger temperature–HVAC carbon emission association during the recovery phase. Robustness analyses using residual diagnostics, HC3 robust standard errors, leave-one-month-out sensitivity analysis, leave-one-out cross-validation, expanding-window time-series validation, and phase-stratified bootstrap resampling supported the stability of the temperature–phase interaction, although the simpler phase-dummy model showed better strictly chronological predictive performance than the interaction model. In contrast, the non-HVAC electricity proxy and its phase interaction were not statistically significant under HC3 robust inference, indicating that the proxy-related associations were less robust and should be interpreted cautiously. The proposed framework is intended for retrospective carbon accounting and exploratory driver identification in data-constrained commercial buildings, rather than real-time HVAC control or peak-load forecasting.

• openalex https://doi.org/10.1088/2631-8695/aeb182first seen 2026-10-10 05:17:50