A data-driven framework for urban building operational energy use modeling

Accurate measurement and analysis of urban energy use is an essential step in development of low-carbon cities. However, there is a limited number of methods and tools for energy use modeling and prediction at urban or neighborhood scales. This article proposes a bottom-up datadriven framework for urban energy use modeling (UEUM) which localizes energy performance measurements and considers urban context. The framework addresses the urban building operational energy estimation through the use of disaggregated energy use data and allows for an accurate urban energy performance measurement at building-level. A machine learning approach is applied to mathematically associate building characteristics and urban context attributes; i.e., building height, as an urban intensity metric, and sprawl indices representing compactness and connectivity of neighborhoods with urban building operational energy use intensity (EUI). Once the mathematical relationship is identified, the model predicts the energy consumption of individual buildings that represent a particular end-user. Chicago as a pilot case study was selected to test the framework. Several algorithms are tested and then the improved model was used to predict energy use for around 820,000 buildings in the city. The framework has the potential to aid designers, planners, and policymakers in a better understanding of the existing urban energy use profile, and the environmental impacts of alternative scenarios of urban development.