Insights from hydronic heating systems in 259 commercial buildings

This paper analyzes hydronic heating systems in 259 buildings across 56 US organizations. The dataset comprises 120,000,000 measurements taken by building automation systems from 2014 to 2024 and for a typical building contains measured supply and return water temperature, flow rate, output power, system state and outdoor temperature spanning 2.2 years (15-minute interval). Pump and boiler data are available for smaller subsets of buildings. Analysis shows that many assumptions about these systems do not align with real-world performance. Systems operate far more frequently than expected − 81% of hours annually for the median building. Loads are lower and more skewed than expected, and oversizing is common even accounting for redundancy requirements. Load distributions indicate that relatively small equipment could serve a surprisingly large percentage of annual loads. Notably, equipment sized to 30% of the maximum measured load could serve up 84% of the total heating energy consumption for the median building. Analysis of key parameters, such as supply water temperature, show that functional resets are rare and that manual overrides are common, often persisting for months or years. Return water temperatures and other data show that boilers operate less efficiently than expected. These findings offer valuable insights for operators, designers, owners, and policymakers, highlighting that substantial energy savings opportunities are commonplace and can be captured by modifying design and operation based on how these systems actually operate. The paper includes the code used for analysis and visualization as well as an open-source dataset of 216 buildings.