Driving forces and typologies behind household energy consumption disparities in China: A machine learning-based approach.

Establishing an intuitive link between driving factors of household energy consumption activities and inequalities is important for the understanding of household heterogeneity in energy consumption behaviours. This paper proposes a novel typology framework based on machine learning approaches and data from 3637 Chinese households in 2014 from 85 cities. Activity-based energy consumption was measured, highlighting inequalities across activities, regions and household types. The results showed significant energy consumption disparities between urban/rural and north/south households, especially in cooking, space heating and vehicle activities. By identifying driving factors of energy consumption, a new household typology classified samples into 6 (all), 6 (urban) and 7 (rural) types. Within these types, households with similar demographic structures, lifestyles and energy consumption habits were clustered. Demographic structure, region, and primary energy demand were used as the basis for the typology. The findings demonstrated how household lifestyle differences explained the cause and underlying driving factors of urban-rural energy consumption inequalities and provided suggestions for city-by-city and type-by-type measurements to support effective low-carbon transformation in cities. • Activity-based household energy consumption (HEC) is measured and calibrated. • Household energy consumption inequalities exist in urban/rural and the north/south groups. • A four-step machine-learning approach well classifies households into reasonable clusters. • Activity-based variables are driving factors for HECs in a LASSO model. • Inequality is narrowed in household clusters and type-by-type policies are needed. [ABSTRACT FROM AUTHOR]