Scholarlink

<p>False data injection attacks (FDIAs) can disrupt state estimation and operational decisions, posing serious threats to the security of smart grids. This paper proposes a Dual-Channel Convolutional Neural Network (DC-CNN) enhanced with an attention mechanism to address the problem of FDIA localization. The framework incorporates a Standardization and Normalization Technique (STN-Tech) to preprocess measurement data, effectively reducing noise and outlier interference. By replacing traditional pooling layers with a dual-channel mechanism consisting of Max pooling and Attention pooling, DC-CNN captures both local and global features, enabling precise attack identification. Experiments conducted on the IEEE 14-Bus, IEEE 57-Bus, and IEEE 118-Bus bus systems demonstrate that the proposed method exhibits high positioning accuracy and robustness under various noise conditions. With an accuracy of 99.36% and an F1-score of 99.3% (IEEE 14-Bus), the F1-score is 5.6% higher than that of GCN; under high-noise conditions (<InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(\sigma\)</EquationSource> </InlineEquation>&#xa0;= 0.6), the F1-score remains above 97%; the accuracy rate for identifying the presence of attacks is 98.6% (IEEE 118-Bus). Overall, DC-CNN provides a reliable solution for FDIA detection and localization, significantly enhancing the cybersecurity of smart grids.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A Dual-Channel CNN for Precise Localization of False Data Injection Attacks in Smart Grids

  • 期刊论文

  • 发表日期：

    2025年8月30日

查看全文
  • Wenlong Jiang,
  • Kangrui Lan,
  • Yunfei Li,
  • Xiaoxin Zhang,
  • Yaru Li,
  • Guoqing Zhang,
  • Wengen Gao

摘要

False data injection attacks (FDIAs) can disrupt state estimation and operational decisions, posing serious threats to the security of smart grids. This paper proposes a Dual-Channel Convolutional Neural Network (DC-CNN) enhanced with an attention mechanism to address the problem of FDIA localization. The framework incorporates a Standardization and Normalization Technique (STN-Tech) to preprocess measurement data, effectively reducing noise and outlier interference. By replacing traditional pooling layers with a dual-channel mechanism consisting of Max pooling and Attention pooling, DC-CNN captures both local and global features, enabling precise attack identification. Experiments conducted on the IEEE 14-Bus, IEEE 57-Bus, and IEEE 118-Bus bus systems demonstrate that the proposed method exhibits high positioning accuracy and robustness under various noise conditions. With an accuracy of 99.36% and an F1-score of 99.3% (IEEE 14-Bus), the F1-score is 5.6% higher than that of GCN; under high-noise conditions ( \(\sigma\)  = 0.6), the F1-score remains above 97%; the accuracy rate for identifying the presence of attacks is 98.6% (IEEE 118-Bus). Overall, DC-CNN provides a reliable solution for FDIA detection and localization, significantly enhancing the cybersecurity of smart grids.