Abstract: While most electricity systems are designed to handle peak demand during summer months, pathways to deep decarbonization generally electrify building heating, thus increasing electricity demand during winter. A key question is how climate variability and change will affect peak heating and cooling demand in an electrified future. We conduct a spatially explicit analysis of trends in temperature-based proxies of electricity demand over the past 70 years. Average annual demand for heating (cooling) decreases (increases) over most of the contiguous US. However, while climate change drives robust increases in peak cooling demand, trends in peak heating demand are generally smaller and less robust. Because the distribution of temperature exhibits a long left tail, severe cold snaps dominate the extremes of thermal demand. As building heating electrifies, system operators must account for these events to ensure reliability.
Introduction
The transition to net-zero emissions relies heavily on the electrification of building heating systems, typically through the adoption of electric heat pumps. While current U.S. power grids are largely "summer-peaking"—designed to handle the surge in air conditioning during heatwaves—this shift will likely turn many regions into "winter-peaking" systems. A critical challenge for grid reliability is that infrastructure must be sized to survive extreme "peak" events, not just average conditions. This study investigates whether historical warming trends are reducing these dangerous winter peaks as effectively as they are reducing average heating needs.
Methodology and Data
The authors performed a spatially explicit retrospective analysis of temperature-based proxies for energy demand across the contiguous United States (CONUS) from 1950 to 2021.
- Proxies: "Inferred demand" was calculated based on the deviation of hourly temperatures from a 65°F (18.33°C) comfort threshold.
- Datasets: The study utilized the ERA5 Reanalysis dataset for hourly surface temperatures and GPWv4 population data to weight grid cells by density.
- Spatial Aggregation: Results were aggregated into 20 population-weighted "eGRID" sub-regions to align with actual U.S. electrical territory boundaries.
- Peak Definition: "Peak" events were defined as the annual maximum intensity of demand over a rolling 72-hour window, reflecting the duration of extreme weather events like the 2021 Winter Storm Uri.
Key Findings
The Disparity Between Average and Peak Trends
Over the last 70 years, average annual heating demand has decreased across most of the U.S., while average cooling demand has increased. These spatial trends are robust and reflect the general warming of the planet.
However, the trends for "peak" demand tell a different story. While peak cooling demand has increased significantly (median trend of 0.16% per year), trends in peak heating demand are much smaller and often statistically insignificant. In many northern regions, the "worst-case" winter event today remains nearly as severe as it was in the 1950s.
The "Long Left Tail" Paradox
The study identifies a critical statistical insight: the distribution of temperature exhibits a "long left tail." While the "bulk" of the distribution is shifting toward warmer temperatures (reducing average heating demand), the extreme cold snaps at the very edge of the tail are not shifting at the same rate.
This leads to a decrease in the "load factor" (the ratio of average to peak demand) in northern regions. As these areas electrify, they will face a dangerous paradox: they must maintain massive, expensive reserve capacities that sit idle most of the year just to prevent blackouts during rare, severe polar vortex events that show no sign of disappearing.
Conclusion and Future Work
Grid operators cannot assume that average warming will automatically mitigate winter peak risks. Reliability in an electrified future requires accounting for the persistence of extreme cold. While this study provides a vital historical baseline, it relies on temperature as a proxy and does not account for humidity, wind chill, or shifting population migrations. Future work should apply this framework to climate model projections (CMIP6) to determine if these historical "tail" behaviors will persist under more extreme future warming scenarios.
Summary updated