Heterogeneity in Temperature Sensitivity of Building Energy Use for Space Heating

District heating is a cornerstone in many decarbonization strategies for urban building stocks, but its efficiency depends on the thermal performance of connected buildings. In this study, we evaluate the temperature sensitivity of heating energy use for buildings that are part of the district heating network in Leipzig, Germany. We analyze smart meter readings from 3427 residential, public, and commercial buildings, comprising over 80 million observations between January 2022 and August 2024. Our findings suggest that a 1°C decrease in outdoor temperature is associated with a median increase in heating energy use of 0.75 W/m 2 . However, this sensitivity varies substantially with 15% of buildings having an effect that is larger than 2 W/m 2 . Single-family homes exhibit the highest sensitivity, whereas heating energy use of large multi-family and public buildings is less responsive to temperature variation. We further show that temperature sensitivity can be estimated from static building characteristics through supervised Machine Learning. Gradient-boosted decision trees achieve the highest predictive performance and the most important predictors are building height, area, and volume. Finally, we quantify retrofitting potential by simulating improvements in temperature sensitivity coefficients. Our findings suggest that a 50% reduction in the sensitivity to outdoor temperature is associated with an expected reduction of up to 20% in annual heating energy use for the buildings with absolute temperature sensitivity above 2 W /( m 2 K ). Our approach illustrates how smart meter readings can be utilized to identify candidate buildings for energy retrofitting.