Mapping Spatiotemporal Disparities in Residential Electricity Inequality Using Machine Learning.

Area of Science:

• Energy policy
• Environmental science
• Data science

Background:

• Decarbonizing the energy sector through electrification is essential.
• Energy unaffordability may increase without adequate safeguards.
• Understanding neighborhood-scale residential electricity inequality is crucial.

Purpose of the Study:

• To develop a high-resolution, spatiotemporally explicit machine learning (ML) framework to predict residential electricity consumption.
• To construct an electricity affordability gap (EAG) metric to identify energy-vulnerable communities.
• To analyze spatial and temporal disparities in energy affordability.

Main Methods:

• Utilized publicly available, multisourced remote sensing and census data.
• Developed a machine learning (ML) framework for predicting tract-level monthly electricity consumption (2013-2020).
• Constructed the electricity affordability gap (EAG) metric (electricity bills vs. 3% household income).

Main Results:

• The ML framework improved electricity consumption data resolution, achieving an R² of 0.82 compared to LEAD data.
• An estimated annual $16.18 billion economic burden impacts electricity bill affordability.
• Monthly EAG is 2-3 times higher in summer/winter, and rural residents face up to 1.7 times higher burdens than urban counterparts.

Conclusions:

• The developed framework enhances the understanding of residential electricity inequality.
• Significant seasonal and urban-rural disparities in energy affordability were identified.
• Insights can inform equitable electrification strategies and energy justice efforts by addressing spatiotemporal mismatches.