Robust Deep Neural Network-Based Internet of Things for Power Transformer Fault Diagnosis

Strategic Insights -

Researchers from Aswan University in Egypt have developed a robust deep neural network-based internet of things (IoT) system for diagnosing power transformer faults. The system, which utilizes a deep learning model hybridized with an IoT platform, has achieved high accuracy in diagnosing faults under imbalanced data and uncertainties. The proposed method outperforms several state-of-the-art approaches and has been validated through empirical results. Additionally, the system's ability to analyze transferred dissolved gas analysis (DGA) samples and visualize transformer faults remotely has been demonstrated.

Key Takeaways:

  • The Duval pentagon method (DPM) is one of the most accurate and reliable dissolved gas analysis (DGA) interpretation methodologies.
  • Implementing large amounts of data in DPM is still challenging and has several limitations.
  • A robust deep neural network (DNN) method for precise DGA monitoring has been introduced, which includes a synthetic minority over-sampling technique-edited nearest neighbor (SMOTE-ENN) preprocessing to eliminate noise from the imbalanced dataset.
  • A unique RobustScaler technique is employed to maintain high performance against uncertain data noise.
  • The proposed system utilizes an industrial IoT platform to analyze transferred DGA samples and visualize transformer faults remotely.
  • The proposed method achieves satisfaction in diagnosing faults for the assessment dataset, with an accuracy of 98.19 %.
  • The system has been validated through empirical results, which show that it outperforms several state-of-the-art approaches.
  • The proposed method is effective in diagnosing faults under imbalanced data and uncertainties, with a superior prediction diagnosis of the transformer faults.

Statistics:

  • 98.19 % accuracy in diagnosing faults for the assessment dataset.
  • 20 % uncertainty noise tolerated by the proposed method.
  • The proposed system utilizes a deep learning model hybridized with an IoT platform.
  • TheIoT gateway sends classification results to the cloud for visualizing the detected fault on the IoT dashboard.

Sources:

  • International Journal of Electrical Power & Energy Systems (Elsevier), 2025, 168():110731.
  • doi.org/10.1016/j.ijepes.2025.110731 (free version available at https://doi-org.sdpl.idm.oclc.org/)