Scholarlink

<p>As power transmission networks become more sophisticated, driven by the integration of distributed energy resources and real-time data acquisition, traditional fault diagnosis methods face significant challenges in terms of accuracy, speed, and scalability. Pursuing of the current systems and information analyses based on adequate machine learning algorithms, such as Random Forest (RF), are insufficient to reveal the complexity and constantly evolving characteristics of modern grids. For these reasons, this research introduces the AI-Augmented Fault Diagnosis Framework rooted in a three-tier deep learning structure that integrates GNN for capturing the spatial characteristics of the system, Transformer for adequate consideration of the temporal aspect, and Reinforcement Learning for flexibility. Whereas the GNN deals with spatial interactions of components like transformers, transmission lines, and nodes and the Transformer model works as a time series data for fault pattern modeling. In incorporating RL into the decision making process, it brings in the variable aspect of detection parameters and thresholds for adjustment with real-time interaction. The effectiveness of the proposed framework is highly rewarding, in comparison to previous studies: Accuracy of 99.2% Fault Detection Rate (FDR) of 98.5% Mean Time to Detection (MTTD) of 1.6s Mean Time to Repair (MTTR) of 7.1&#xa0;min. The proposed framework was tested using MATLAB and Simulink, and the results revealed its suitability for application in today’s advanced power systems.</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

AI-Augmented Fault Diagnosis Framework for Real-Time Transmission Network Monitoring

  • 期刊论文

  • 发表日期：

    2026年4月30日

查看全文
  • Jiaxin Lin,
  • Wenyuan Wu,
  • Bo Hu,
  • Qiang Qin,
  • Hanye Huang

摘要

As power transmission networks become more sophisticated, driven by the integration of distributed energy resources and real-time data acquisition, traditional fault diagnosis methods face significant challenges in terms of accuracy, speed, and scalability. Pursuing of the current systems and information analyses based on adequate machine learning algorithms, such as Random Forest (RF), are insufficient to reveal the complexity and constantly evolving characteristics of modern grids. For these reasons, this research introduces the AI-Augmented Fault Diagnosis Framework rooted in a three-tier deep learning structure that integrates GNN for capturing the spatial characteristics of the system, Transformer for adequate consideration of the temporal aspect, and Reinforcement Learning for flexibility. Whereas the GNN deals with spatial interactions of components like transformers, transmission lines, and nodes and the Transformer model works as a time series data for fault pattern modeling. In incorporating RL into the decision making process, it brings in the variable aspect of detection parameters and thresholds for adjustment with real-time interaction. The effectiveness of the proposed framework is highly rewarding, in comparison to previous studies: Accuracy of 99.2% Fault Detection Rate (FDR) of 98.5% Mean Time to Detection (MTTD) of 1.6s Mean Time to Repair (MTTR) of 7.1 min. The proposed framework was tested using MATLAB and Simulink, and the results revealed its suitability for application in today’s advanced power systems.