基于图注意力网络与聚类的用户画像研究  

User Portrait Construction Based on Graph Attention Network Combined With Clustering Algorithm Model

作　　者：刘励耘 许孝梅 Liu Liyun;Xu Xiaomei(Yunnan Power Grid Co.,Ltd.,Honghe Power Supply Bureau,Honghe 651400,Yunnan,China)

机构地区：[1]云南电网有限责任公司红河供电局,云南红河651400

出　　处：《云南电力技术》2025年第1期68-75,85,共9页Yunnan Electric Power

基　　金：中国南方电网有限责任公司科技项目:面向多维度充电行为分析的“车-桩-网”一体化动态运行监控技术的研究(YNKJXM20230343)。

摘　　要：本文针对电动汽车用户充电行为的差异性对充电设施布局带来的挑战,提出了一种适用电动汽车充电场景的基于图注意力网络与聚类的用户画像模型。利用某地电动汽车充电数据,通过数据预处理、特征选择、特征重构以及用户画像模型构建等步骤,将充电用户划分为7个类别。结果表明,不同用户群体在充电频率、充电效率等方面表现出显著差异。本文的结论可为充电基础设施的优化布局、个性化充电服务及市场营销策略的制定提供理论支持。In this paper,a user portrait model based on graph attention network and clustering is proposed for electric vehicle charging scenarios in response to the challenges posed by the differences in charging behaviors of electric vehicle users to the layout of charging facilities.The paper utilizes electric vehicle charging data from a certain place to classify charging users into seven categories through data preprocessing,feature selection,feature reconstruction and user portrait model construction.The results show that different user groups exhibit significant differences in charging frequency and charging efficiency.The conclusions of this paper can provide theoretical support for the optimal layout of charging infrastructure,personalized charging service and the formulation of marketing strategies.

关 键 词：电动汽车 用户画像 GAT模型 聚类分析 

分 类 号：TM74[电气工程—电力系统及自动化]