Data-driven analysis of influential factors on residential energy end-use in the US

Decarbonizing the residential buildings sector is important to realize a net-zero future. However, little research comprehensively explores how residential energy by end-uses, i.e., space heating, space cooling, water heating, and appliances, vary across different climate, building, and occupant characteristics. Therefore, based on Residential Energy Consumption Survey 2015, this study investigates how these energy end-uses correlated with different influential factors, i.e., climate, demographics, housing and appliance features. Recursively random forest regression and partial dependency plots were employed to analyze and quantify the impact of influencing factors on end-use energy usage. The developed models illustrate nonlinear and varying impacts of influential factors on different end-uses, with R-square from 0.481 (appliance) to 0.885 (space cooling). Specifically, housing size and climate are determinant factors of space heating/cooling, while family size and total appliance number determine water heating and appliance usage, respectively. Use frequency, size, and vintage of common appliances (e.g., refrigerator, laundry) affect appliance energy consumption. Furthermore, one-way ANOVA confirms the statistically significant differences in energy consumption across varying household income groups, i.e., households with higher incomes consume more energy. Houses with older vintages use more total energy, especially in space heating, while less in cooling due to lower penetration of air conditioners. Finally, the analysis proves that residential end-use electrification, mainly space heating and water heating, is beneficial for energy efficiency. The findings contribute to understanding influential factors on different end-uses in households, which from a holistic level informs the potential pathway for building decarbonization, especially targeting at specific end-uses or occupant groups