摘要
随着园区负荷预测的重要性日益凸显,为满足电网公司和用电企业之间的数据安全和隐私保护需求,文章提出了一种基于压缩感知的纵向联邦学习园区负荷预测方法。该方法利用压缩感知技术对梯度进行降维压缩,有效减少了传输的数据量。实验结果显示,相较于传统的联邦学习方法,文章的方法在保持模型准确率的同时显著降低了通信消耗。此外,该方法还有效保护了数据的隐私性,为电网公司和用电企业提供了更安全的合作环境。
With the increasing importance of park load forecasting,in order to satisfy the data security and privacy protection needs between power enterprises and users,this paper proposes a vertical federated learning park load forecasting method based on compressed sensing.The method utilizes the compressed sensing technique to downscale the gradient,which effectively reduces the amount of transmitted data.Experimental results show that compared with the traditional federated learning method,the method in this paper significantly reduces the communication consumption while maintaining the model accuracy.In addition,the method effectively protects the privacy of the data and provides a safer cooperation environment between power enterprises and users.
作者
杨珂
朱洪斌
李达
张闻彬
杨挺
覃小兵
YANG Ke;ZHU Hongbin;LI Da;ZHANG Wenbin;YANG Ting;QIN Xiaobing(State Grid Digital Technology Holding Co.,Ltd.,Xicheng District,Beijing 100053,China;State Grid Blockchain Technology(Beijing)Co.,Ltd.,Xicheng District,Beijing 100053,China;Big Data Center,State Grid Corporation of China,Xicheng District,Beijing 100052,China;State Grid Blockchain Application Technology Laboratory,Xicheng District,Beijing 100053,China;Information and Communication Company,State Grid Shandong Electric Power Company,Jinan 250001,Shandong Province,China;School of Electrical and Automation Information Engineering,Tianjin University,Nankai District,Tianjin 300072,China)
出处
《电力信息与通信技术》
2024年第5期36-42,共7页
Electric Power Information and Communication Technology
基金
国家电网有限公司总部科技项目资助“能源大数据多方安全融合应用隐私计算技术研究”(5108-202218280A-2-393-XG)。
关键词
负荷预测
纵向联邦学习
压缩感知
load forecasting
vertical federated learning
compressed sensing