Urban form play an important role in influencing building energy use. Previous empirical studies on urban form and building energy mostly represented urban form as separate indicators of form components and socio-economic conditions. However, urban form is a complex system that consists of different components and interactions. Thus, simply considering urban form as an aggregation of their components and properties often suffers from strong correlations among these factors and the confounding effect as in previous studies. To better understand the energy performance of complex urban form, this study identifies urban form typologies in Seoul using unsupervised learning and examines the energy performance of those typologies. The Gaussian mixture model (GMM), a widely used clustering method, is adopted to identify 16 urban form typologies based on 10 common urban form and population distribution factors. Electricity use and gas use in different typologies in 2019 are collected and compared using the Brown Forsythe ANOVA and Games Howell post hoc test. Results suggest that urban form typologies reflect energy efficiency to a certain degree, and the difference in energy efficiency across urban form typologies is more significant in gas than in electricity because of occupant behavior differences in their use. Among the urban form typologies, high-rise high-density urban form type is the most efficient, and the high-rise mixed type is the least efficient. The findings from this research could help urban planners and designers to understand the relationship between complex urban form and building energy better to support planning and design toward an energy-efficient urban form.
Keywords urban morphology, clustering, building energy consumption, Gaussian mixture model, urban planning