Mapping Spatiotemporal Disparities in Residential Electricity Inequality Using Machine Learning.

The move toward electrification is critical for decarbonizing the energy sector but may exacerbate energy unaffordability without proper safeguards. Addressing this challenge requires capturing neighborhood-scale dynamics to uncover the blind spots in residential electricity inequality. Based on publicly available, multisourced remote sensing and census data, we develop a high-resolution, spatiotemporally explicit machine learning (ML) framework to predict tract-level monthly electricity consumption across the conterminous U.S. from 2013-2020. We then construct the electricity affordability gap (EAG) metric, defined as the gap between electricity bills and 3% of household income, to better identify energy-vulnerable communities over space and time. The results show that our framework largely improves the resolution of electricity consumption data while achieving an of 0.82 compared to the Low-Income Energy Affordability Data (LEAD). We estimate an annual $16.18 billion economic burden on the ability to afford electricity bills, exceeding current federal appropriations in alleviating energy difficulties. We also observe pronounced seasonal and urban-rural disparities, with monthly EAG in summer and winter being 2-3 times greater than other seasons and rural residents facing burdens up to 1.7 times higher than their urban counterparts. These insights inform equitable electrification by addressing spatiotemporal mismatches and multiple jurisdictional challenges in energy justice efforts.