Spatial coordination of electric vehicle adoption within the urban built environment using machine learning

Abstract

Adopting electric vehicles (EVs) aims to reduce urban carbon emissions from the demand side of transportation. However, there is limited understanding of how the urban built environment influences the process of EV adoption. As EV technology and the policy landscape continue to evolve, comprehending these dynamically developing relationships is crucial for promoting EVs and facilitating their coordination with existing urban infrastructure. This study proposes a data-driven machine learning framework to explore the nonlinear relationships and threshold effects between EV adoption and various urban environmental variables. By integrating GIS spatial analysis, we enhance the interpretability of the model results. Taking Beijing, China, as a case study, the findings confirm several factors previously identified as influencing EV adoption, such as charging infrastructure, population density, and economic development. Additionally, this study reveals regional insights that differ from existing research. Factors such as proximity to urban centers and the distance to, and accessibility of, public charging stations are no longer the most influential variables in the city. Compared to linear assumptions, a comprehensive analysis that includes multiple nonlinear variables enhances our understanding of the coordination of EVs within existing urban systems. The threshold effects from the results suggest a current 1.5-km radius for new charging infrastructure development around residential areas, while it is also important to consider other coordinating thresholds, such as the distance from the city center (10 km or 30 km), land-use mix entropy (around 0.7), and areas with low building coverage and low bus stop density. This study calls for more attention to the continuously evolving dynamic nonlinear relationship between EVs and various built environment variables.