Rapid Archetype-Based City-Scale Building Modeling and Parameterization

Due to the challenges associated with obtaining specific building-level non-geometry information, the building archetype method has been proposed as a means of modeling building energy use on an urban scale. This method involves the classification of buildings and the characterization of archetypes. Traditional classification methods select certain building features that are known to have a significant impact on energy usage (such as building age and use type) and classify similar buildings into one category based on these features. However, these methods are subject to researcher bias. Recent progress in machine learning techniques offers new insights into this field. This paper proposes a method for developing building archetypes based on machine learning, and presents a case study of New York City. Building features are extracted from open-source datasets available on official websites. Well-known machine learning algorithms are employed to classify buildings based on these features. Each building category is assigned a template based on related ASHRAE standards to characterize the non-geometric parameters required for energy modeling input. The validity of this approach will be tested by simulating the total energy use of buildings using the proposed archetypes via the EnergyPlus engine and OpenStudio.