面向联邦学习的油电混合电动汽车站点推荐算法研究  

Research on hybrid electric vehicle station recommendation algorithm based on federated learning

作　　者：蒋灵慧 冯霞[1] 崔凯平 王亚茹 JIANG Linghui;FENG Xia;CUI Kaiping;WANG Yaru(School of Automotive and Transportation Engineering,Jiangsu University,Zhenjiang 212013,China;School of Data Science,City University of Macao,Macao 999078,China)

机构地区：[1]江苏大学汽车与交通工程学院,江苏镇江212013 [2]澳门城市大学数据科学学院,中国澳门999078

出　　处：《重庆理工大学学报(自然科学)》2025年第2期113-119,共7页Journal of Chongqing University of Technology:Natural Science

基　　金：国家自然科学基金项目(62272203)。

摘　　要：作为一种低碳、环保的交通工具,油电混合电动汽车(hybrid electric vehicle,HEV)发展迅速。为保障推荐过程中用户的隐私安全,提出了一种面向纵向联邦学习算法的HEV站点推荐算法。通过本地训练、中央聚合的模型训练机制,在保证用户隐私数据安全的前提下,更新局部训练模型。将区块链技术与云计算相结合,通过使用加密算法和分布式存储,提供一个安全可信的云服务网络,负责传输本地计算的训练参数。利用去中心化的数据聚合器取代易出现单点故障的集中式架构,创建了一个灵活且可扩展的云网络。实验结果表明,具有10个云节点的分散式算法比传统的集中式算法快5.2 s。可见,基于纵向联邦学习的推荐算法既能保证推荐的精准性,也能充分调动闲置站点,有效提高推荐效率。As a means of low-carbon and environmentally friendly transportation,Hybrid Electric Vehicle(HEV)is developing rapidly.To guarantee the privacy and security of users in the recommendation process,we propose a HEV station recommendation algorithm based on vertical federated learning algorithm.Through the model training mechanism of local training and central aggregation,the local training model is updated under the premise of user privacy data security.Blockchain technology is integrated with cloud computing to provide a secure and trustworthy cloud service network responsible for transmitting locally computed training parameters through the use of cryptographic algorithms and distributed storage.A flexible and scalable cloud network is created by using a decentralized data aggregator instead of a centralized architecture which is prone to a single point of failure.Our experimental results show the decentralized algorithm with 10 cloud nodes is 5.2 s faster than the traditional centralized algorithm.Also,the recommendation algorithm based on longitudinal federated learning not only ensures the accuracy of the recommendation,but also fully mobilizes the idle sites to effectively improve the recommendation efficiency.

关 键 词：油电混合电动汽车 推荐算法 纵向联邦学习 充电站推荐 云计算 区块链 

分 类 号：TM910.6[电气工程—电力电子与电力传动]