Dynamic assessment and optimization of rural household heating energy transition in developing regions with interpretable machine learning

The low-carbon transition of household heating energy is crucial for mitigating global climate change, improving human well-being, promoting economic development, and achieving the Sustainable Development Goals (SDGs). It is therefore critical to explore the household heating energy transition (HET) to leverage these interconnections and strategically navigate the dynamic energy landscape. This study innovatively develops a dynamic assessment framework for household HET from a micro-level perspective, integrating three distinct stages, namely “Facilitation transition,” “Initial transition,” and “Stable transition.” Drawing on a dataset comprising 1858 rural households in the ecologically fragile region of Northwest China, with a total of 53,882 observations, it employs the interpretable machine learning (ML) model to identify the key drivers of household HET. It characterizes distinct energy-use profiles, quantifies nonlinear threshold mechanisms and interaction effects among multidimensional drivers, and designs precision-targeted intervention strategies for household HET. The key findings highlight that ML models exhibit significant advantages in identifying the key drivers of household HET. Among the drivers, policy perception and satisfaction (PPS), household income (HI), social networks (SNs), and family size demonstrate significant multiparameter, nonlinear, and threshold effects on household HET. Notably, SNs exhibits a pronounced context-dependent bimodal effect, providing novel insights into the heterogeneous impacts on household HET, and SNs, PPS, and HI show significant coupled and synergistic effects on household HET. The proposed multiscale household HET optimization framework offers actionable guidance for designing precision-targeted intervention strategies and robust policy frameworks aimed at promoting low-carbon energy transitions and sustainable development in developing and less-developed regions globally.