基于电动出租车数据的充电桩选址聚类方法比较  

Comparison of Clustering Methods for Charging Station Site Selection Based on Electrical Taxi Data

作　　者：甄西媛 高超[1] 李向华 冀杰[2] ZHEN Xiyuan;GAO Chao;LI Xianghua;JI Jie(Northwestern Polytechnical University,Xi’an 710072,China;Southwest University,Chongqing 400715,China)

机构地区：[1]西北工业大学,西安710072 [2]西南大学,重庆400715

出　　处：《汽车工程学报》2023年第4期564-573,共10页Chinese Journal of Automotive Engineering

基　　金：国家科技部重点研发计划项目(2022YFE0112300):突发公共卫生事件下的医疗资源供给与配置模式研究。

摘　　要：为有效降低出租车运营企业及经营者的经济成本,通过分析出租车的卫星轨迹数据,比较和选取用于电动出租车充电桩选址规划的聚类方法。以上海市电动出租车充电站的选址规划为研究对象,分别基于孤立森林和聚类算法设计异常值检测方法,对相关时段的出租车卫星数据进行清理以及数据可视化处理;比较层次聚类(Agglomerative Clustering)、高斯混合模型(Gaussian Mixture Model,GMM)、K-means聚类、Mean-Shift聚类以及谱聚类(Spectral Clustering)5种算法的聚类效果,并选取K-means算法作为充电桩选址规划参考算法。从城市区域划分及企业运营角度确定充电桩选址方案,为未来上海市区电动出租车充电桩的数量和容量配置提供设计依据。In order to effectively reduce the economic expenses for taxi companies and users,the analysis of satellite trajectory data for electrical taxis is used to compare and select suitable clustering methods for charging pile location planning.Focusing on the location planning of electrical taxi charging stations in Shanghai,the paper designed the outlier detection method based on isolated forest and clustering algorithms to clean up the taxi satellite data within the relevant time period,followed by data visualization processing.The clustering effects of five algorithms,including Agglomerative hierarchical clustering,GMM Gaussian mixed clustering,K-means clustering,Mean-shift clustering and Spectrum clustering were evaluated and compared.And the K-means algorithm was selected as the reference algorithm for charging pile location planning.From the perspectives of urban zoning and business operations,a site selection strategy is determined,which provides a foundation for the design and planning of the quantity and capacity allocation of electric taxi charging piles in Shanghai for the future.

关 键 词：电动出租车 充电桩选址 异常值检测 聚类方法 可视化 

分 类 号：U491.8[交通运输工程—交通运输规划与管理]