Cao, Ziyi, Chen, Lili, Luo, Zhiwen  ORCID: https://orcid.org/0000-0002-2082-3958, Zeng, Fanzhe, Yang, Xinyan and Wang, Kai
2026.
A hybrid model for predicting the operation status and power of residential air conditioning based on coupled indoor-outdoor hygrothermal variations.
Energy and Buildings
372
, 118269.
10.1016/j.enbuild.2026.118269
|
Preview |
PDF
- Accepted Post-Print Version
Available under License Creative Commons Attribution. Download (1MB) | Preview |
Abstract
With the ongoing climate change, building cooling energy demand is projected to rise. At the same time, it would be further exacerbated by the urban-induced warming, as well as the more frequent extreme heat events. Enhancing our understanding and prediction capability of the building cooling demand can improve urban energy security and climate resilience. However, significant challenges remain in representing occupant-driven residential air conditioning (AC) operation, which responds dynamically to outdoor environmental conditions yet is difficult to observe directly. To address these, we developed a hybrid modelling framework that infers residential AC operation and energy use from indoor-outdoor hygrothermal dynamics, without requiring historical energy consumption or explicit occupancy data. By integrating machine learning (XGBoost and the PELT algorithm) with physics-based RC building energy balance model, the approach detects the AC operational status from characteristic temperature and humidity patterns, and estimates AC power from the cooling load. In doing so, occupant behaviour is treated as an emergent signal embedded within environmental responses, rather than an external input requiring direct measurement. The model was validated against the data from a residential building in Beijing. It achieved an AC-on F1-score of 81.2%, which is the harmonic mean of precision and recall and a balanced classification metric for the imbalanced dataset, and a median error of 27.09% in estimated AC power. This work enables the integration of occupant behaviour into building energy modelling in data-sparse contexts, particularly in residential buildings where AC operation is primarily human-driven. By linking outdoor climate conditions to indoor responses and behavioural patterns, the proposed framework supports improved assessment of future cooling demand and indoor overheating risks under climate change.
| Item Type: | Article |
|---|---|
| Date Type: | Publication |
| Status: | Published |
| Schools: | Schools > Architecture |
| Additional Information: | RRS policy applied |
| Publisher: | Elsevier |
| ISSN: | 0378-7788 |
| Date of First Compliant Deposit: | 29 September 2026 |
| Date of Acceptance: | 13 September 2026 |
| Last Modified: | 29 Sep 2026 14:15 |
| URI: | https://orca.cardiff.ac.uk/id/eprint/189876 |
Actions (repository staff only)
 |
Edit Item |
Altmetric
Altmetric