Abstract
Abstrcat The rural energy transition is pivotal for achieving building decarbonization goals, yet rural dwellings in China often exhibit high energy demand and low end-use efficiency due to limited envelope quality and legacy energy-use practices. This study investigates the drivers and interaction pathways of end-use energy consumption in rural dwellings of China’s Guanzhong Plain. By conducting a comprehensive field survey of 438 households, this research collected empirical data on household characteristics, building attributes, and energy use patterns. The CatBoost machine learning model, integrated with the SHAP interpretability framework, was employed to identify key factors and analyze nonlinear interaction effects. The results demonstrate that appliance ownership exerts the greatest influence on energy consumption, followed by a significant contribution from energy structure. Important building-related factors include floor area, construction year, and roof type, while household factors such as distance to town and annual income demonstrated complex, nonlinear relationships with energy use. Factor interactions further reveal that enhanced thermal performance can mitigate increased energy use driven by increased appliance ownership. Households located near towns showed increased energy demand as income rose, whereas those in remote areas showed the opposite trend. Furthermore, newer buildings with thicker exterior walls exhibited better energy performance. Multi-story homes generally consumed less energy, though high occupancy could offset this advantage. The model demonstrated strong predictive performance (R2 = 0.90), effectively capturing complex, multidimensional energy consumption behaviors. These findings underscored the importance of tailored energy-saving strategies that integrate improvements to building envelopes, transitions to clean energy, and behavior-based interventions, adapted to local conditions.
| Original language | English |
|---|---|
| Article number | 117518 |
| Journal | Energy and Buildings |
| Volume | 362 |
| DOIs | |
| State | Published - 1 Jul 2026 |
UN SDGs
This output contributes to the following UN Sustainable Development Goals (SDGs)
-
SDG 7 Affordable and Clean Energy
Keywords
- Building energy use
- Factor interactions
- Household survey
- Machine learning interpretation
- SHAP analysis
Fingerprint
Dive into the research topics of 'Drivers and interaction pathways of end-use energy consumption in rural Chinese dwellings: Evidence from the Guanzhong Plain'. Together they form a unique fingerprint.Cite this
- APA
- Author
- BIBTEX
- Harvard
- Standard
- RIS
- Vancouver