Extreme weather events, such as heatwaves and cold snaps, significantly impact energy infrastructure and cause power outages and price spikes. The 2021 Texas freeze (Winter Storm Uri) and the 2020 California heatwave are prime examples of how unexpected surges in heating and cooling demand, respectively, can overwhelm the grid. Currently, peak electricity demand in the contiguous US (CONUS) occurs during summer due to air conditioning. However, deep decarbonization pathways typically involve electrifying building heating, shifting peak demand to winter. This study addresses the critical question of how climate variability and change will affect peak heating and cooling demands in this electrified future. While climate change generally leads to increased temperatures, it also alters the frequency and intensity of extreme weather events, making it crucial to understand how these changes affect both average and peak energy demand. This is especially important for reliable grid operation and managing the associated costs. The study focuses on historical trends to inform near-term operational decisions and investments, acknowledging the uncertainty of long-term technological and socioeconomic factors.

Existing literature points to a general increase in global surface temperatures due to anthropogenic climate change. A simple shift in temperature distribution would imply decreased heating and increased cooling demand. However, the literature also highlights the increasing frequency and magnitude of heatwaves. Understanding changes in cold extremes is less straightforward, with ongoing research investigating the link between Arctic amplification and mid-latitude winter temperature extremes. Previous studies have examined the impact of electrification on electricity load, projecting a shift towards winter peaking systems in many parts of the US. This study builds upon this existing research by providing a spatially explicit, long-term analysis of trends in heating and cooling demand, considering both average annual and peak demands.

This study employs a retrospective analysis of trends in heating and cooling demand using temperature-based proxies over 72 years (1950–2021). The analysis utilizes hourly temperature data from the ERA-5 reanalysis dataset. Inferred demand for heating and cooling is calculated as the difference between hourly temperature and a threshold temperature of 65 °F, representing the degrees a building must be heated or cooled to reach thermal comfort. Total thermal demand is the sum of heating and cooling demand. The analysis quantifies changes in both annual mean and annual maximum (peak) energy demands. To account for population distribution, trends are aggregated spatially, weighting each grid cell by its 2020 population. The study focuses on major electric grid systems, using Florida and the Midcontinent Independent System Operator (MISO) as representative examples of hot and cold regions. Statistical analysis includes the Mann-Kendall trend test and Thiel-Sen slope estimation to assess the significance of trends. Field significance tests account for the multiple comparisons across grid cells. The analysis also examines thermal load factors (ratio of average to peak demand) as a measure of grid operational efficiency.

The study reveals robust increases in annual mean cooling demand and decreases in annual mean heating demand across most of the CONUS. The largest increases in cooling demand are observed in central Colorado (~1%/year), while the largest decreases in heating demand are found in southern Florida (below -1%/year). Total thermal demand shows negative trends across most of the CONUS due to longer winters. However, in southern states, increased cooling demand outweighs decreased heating demand. Aggregating findings at the grid level, Florida shows a statistically significant increase in total thermal demand, while MISO shows a significant decrease, representing the trends in hot and cold regions. Analyzing peak demands (72-hour events), increasing trends in peak cooling demand are observed across large portions of the CONUS (median trend: 0.16%/year). Decreasing trends in peak heating demand are found almost everywhere (median trend: -0.1%/year). Peak total thermal load also shows decreasing trends, except in southernmost regions. There is no systematic shift in the seasonality of peak events. In Florida, peak cooling demand is increasing, and peak heating demand has recently declined; MISO shows high inter-annual variability in peak heating demand. Thermal load factors (average demand/peak demand) show increasing trends in the southern US, with significant increases in southernmost regions. Northern regions and parts of the West show decreasing load factors. This suggests decreasing capacity utilization rates in the North. Lastly, peak winter heating demand is highly variable, posing a challenge for grid operations.

The findings demonstrate significant shifts in thermal loads under a warming climate, impacting regional energy economics. The faster decrease in average heating demand compared to peak heating demand leads to decreasing load factors in northern regions, where heating dominates. Conversely, southern regions, dominated by cooling, experience increasing load factors. These contrasting trends highlight the implications of widespread electrification of heating: increasing costs in the north to maintain reliability and decreasing costs in the south. The study's results pertain to temperature-induced changes, not accounting for humidity, housing stock, occupancy, or heating/cooling technology. Extrapolation to regions outside the CONUS, especially the global south, needs caution due to differences in air conditioning adoption.

This study reveals significant changes in thermal loads across the CONUS driven by climate change. Average winter heating demand is decreasing while average summer cooling demand is increasing, but peak demands show different trends. The high inter-annual variability in peak heating demand underscores the need for improved prediction and reserve capacity management in winter. Future work could incorporate non-temperature factors and further investigate climate model projections of future climate conditions. This research provides crucial insights for grid operators and policymakers in planning for a future characterized by increased electrification and climate change impacts.

The study relies on temperature-based proxies for energy demand, neglecting factors like humidity, building characteristics, occupancy, and the efficiency of heating and cooling technologies. The spatial resolution of the data may also limit the analysis of smaller-scale variations. Extrapolating the findings beyond the CONUS may require adjustments due to differences in energy consumption patterns and climatic conditions.