Weather normalization

Weather normalization adjusts energy usage or other weather-sensitive data to remove the effect of unusually hot, cold or wet conditions, so that a comparison across years or sites reflects real changes in behavior or efficiency rather than the weather that happened to occur. A mild winter should not be reported as an efficiency win.

The typical approach fits a regression analysis of historical usage against degree days or another weather variable, then recalculates what usage would have been under long-term average, or "normal," weather conditions, holding everything else constant. This differs from raw seasonality adjustment, which removes a repeating calendar pattern, whereas weather normalization removes the effect of a specific period's actual weather relative to a climatological baseline, and the two are often applied together.

Weather normalization underpins utility rate cases, building energy-efficiency reporting, and corporate emissions tracking, where regulators and auditors expect year-over-year comparisons to be normalized so improvements are not masked or manufactured by weather variation. Common pitfalls include using too short a historical baseline to define "normal" weather, which makes the baseline itself unstable, and applying a single-site regression model to a portfolio of buildings with very different weather sensitivity feeding load forecasting assumptions that do not hold across the portfolio.