电力信息与通信技术2026，Vol.24Issue(1)：72-78,7.DOI:10.16543/j.2095-641x.electric.power.ict.2026.01.08
基于wk-GDNN模型的虚假数据注入攻击检测研究
Research on False Data Injection Attack Detection Based on wk-GDNN Model
摘要
Abstract
The false data injection attack(FDIA)has a significant impact on the security of the power grid system.Currently,deep learning still has shortcomings in dealing with data processing of power grid topology structure information and capturing long-term dependency relationships.To further improve the accuracy and robustness of the current detection model for false data injection attacks in smart grids,this paper introduces the Wiener-Kinchin(wk)theorem to process the data in frequency domain,and innovatively proposes a graph frequency domain convolutional neural network detection model based on Decoder optimization(wk GDNN,Wiener-Khinchin guided dual-domain neural network).The wk-GDNN model converts the time feature information hidden in the data into frequency domain information.Secondly,it combines the power grid topology perception ability of GCN,and optimizes the spatiotemporal feature extraction through the context information extraction ability of decoder,which improves the detection accuracy and verifies the effectiveness based on IEEE-14/118 node system simulation.The experimental results showed that the F1 scores of the model were 0.9798 and 0.9761,respectively,with an average improvement of 6.67%compared to the comparison model.The results indicate that the frequency domain preprocessing based on the wk theorem and the subsequent frequency domain graph convolution co decoding provide a new paradigm for joint modeling of FDIA detection at multiple scales from time domain to frequency domain,and from nodes to systems.关键词
智能电网/虚假数据注入攻击/图卷积网络(GCN)/时空特征/频谱卷积Key words
smart grid/false data injection attack/graph convolutional network(GCN)/spatiotemporal features/spectral convolutional分类
信息技术与安全科学引用本文复制引用
曾洋,李秀芹..基于wk-GDNN模型的虚假数据注入攻击检测研究[J].电力信息与通信技术,2026,24(1):72-78,7.基金项目
河南省科技攻关计划项目"水下无线传感器网络覆盖保持与三维覆盖空洞修复方法研究"(242102210213). （242102210213）
