交通运输系统工程与信息

交通运输系统工程与信息 ›› 2025, Vol. 25 ›› Issue (6): 276-284.DOI: 10.16097/j.cnki.1009-6744.2025.06.025

• 系统工程理论与方法 • 上一篇    下一篇

建成环境对电动汽车充电行为的非线性影响及交互效应

吴静娴a,b ，管厚杰a,b ，李潇a,b ，赵靖*a,b   

  1. 上海理工大学，a.管理学院；b.智慧城市交通研究院，上海200093)
  • 收稿日期:2025-07-22 修回日期:2025-09-12 接受日期:2025-09-18 出版日期:2025-12-25 发布日期:2025-12-24
  • 作者简介:吴静娴(1987—)，女，江苏盐城人，讲师，博士。
  • 基金资助:
    国家自然科学基金(52502393)；上海市哲学社会科学规划课题 (2022ZGL008)。

Nonlinear and Interactive Effects of Built Environment on Electric Vehicle Charging Behavior

WU Jingxiana,b, GUAN Houjiea,b, LI Xiaoa,b, ZHAO Jing*a,b   

  1. a. Business School; b. Smart Urban Mobility Institute, University of Shanghai for Science and Technology, Shanghai 200093, China
  • Received:2025-07-22 Revised:2025-09-12 Accepted:2025-09-18 Online:2025-12-25 Published:2025-12-24
  • Supported by:
    National Natural Science Foundation of China(52502393)；Shanghai Philosophy and Social Science Planning Project (2022ZGL008)。

PDF

摘要： 为探究建成环境对电动汽车充电行为的非线性影响及交互作用，本文以上海市为例，结合电动汽车充电订单数据和多源建成环境数据，针对工作日与非工作日两种充电场景，构建基于梯度提升决策树(Gradient Boosting Decision Trees, GBDT)的充电强度模型。借助SHAP(SHapley Additive exPlanations)解释器，分解模型输出，量化各建成环境因素的边际贡献，识别不同场景下影响汽车充电行为的关键因素，并解析其与电动汽车充电量强度间的非线性关联。对比线性回归模型，基于GBDT的充电强度模型拟合效果更优，工作日与非工作日场景下的模型拟合优度分别为0.333和0.573。结果显示，无论是工作日还是非工作日，主次干路密度、市中心邻近度和企业密度是影响公共场站电动汽车充电的核心要素，且均呈现显著的非线性特征和阈值效应。各变量的影响机制各异：主次干路密度与企业密度对电动汽车充电量强度呈正向促进作用，而市中心邻近度与公交站点密度则表现为抑制作用。此外，主、次干路密度等变量间也存在明显的交互效应。

关键词: 城市交通, 非线性作用, 梯度提升决策树, 充电行为, 建成环境

Abstract: This study aims to investigate the nonlinear effects and interactions of built environment on the charging behavior of electric vehicles (EVs). Using the data on the EV charging order and multi-source built environment, an empirical study was made in Shanghai. To account for the differences between weekday and non-weekday charging patterns, two Gradient Boosting Decision Tree (GBDT) models were developed. Model interpretations was performed using the SHAP (SHapley Additive exPlanations) framework, which decomposes the output of model into the marginal contributions of built environment factors. The key determinants of EV charging behavior for each period were identified, and their nonlinear relationships with EV charging intensity were analyzed. Compared to the linear regression models, the GBDT-based approach achieved superior performance, with R2 values of 0.333 and 0.573 for weekdays and non-weekdays scenarios, respectively. The results indicate that, the density of main and secondary roads, proximity to the city center, and enterprise density are the core factors which affect EV charging at public stations, and exhibit significant nonlinear and threshold effects, whenever on weekdays and non-weekdays. The influence mechanisms of variables are different. The density of main and secondary roads and enterprise density positively promote the intensity of EV charging, whereas the proximity to the city center and bus stop density exhibit inhibitory effects. Additionally, the significant interactive effects were observed among variables such as the density of main and secondary roads.

Key words: urban traffic, nonlinear effect, Gradient Boosting Decision Tree, charging behavior, built environment

中图分类号: