Reliance on typical weather data misrepresents cooling and heating energy use: Insights from 23 years of building energy simulations across the U.S.

Accurate weather data are critical for simulating building energy use and assessing power-grid demand. Typical meteorological year (TMY3) datasets are widely used for this purpose but represent long-term average conditions assembled from different years, limiting their ability to capture interannual variability and extreme events that often drive peak loads. Actual meteorological year (AMY) data provide continuous, year-specific weather records and thus offer a more realistic depiction of variability and extremes. However, their application has been constrained by limited duration, spatial coverage, and the coarse resolution of many long-term products. In this study, we compare residential building energy consumption across more than 500 U.S. urban locations using TMY3 data and 23 years of AMY data enabled by the Historical Comprehensive Hourly Urban Weather Database (CHUWD-H v1.1). AMY-based simulations reveal substantial year-to-year variability and consistently higher peak loads than TMY3-based results. Relative to the 23-year AMY simulations, TMY3 underestimates cooling energy demand by 11.7 ± 7.5% and overestimates heating demand by 13.6 ± 16.5% on average. These findings demonstrate that reliance on TMY3 can systematically misrepresent both energy demand magnitude and extremes, and underscore the necessity of long-term, urban-resolved AMY datasets for robust building energy assessments and climate-resilient power-system planning.