Sign in Subscribe
Technical institutes

Ultrasonic Temperature Measurement Method for Oil-Immersed Power Transformers

Strategic Insights

— 1 min read

Researchers at Wuhan University of Technology have introduced a novel ultrasonic temperature measurement method for oil-immersed power transformers, which tackles the challenges of monitoring internal temperatures. The method combines machine learning algorithms with ultrasonic sensing to accurately diagnose internal transformer temperatures. The researchers have established an experimental platform to collect acoustic data and filter out interference noise, using feature extraction and dimensionality reduction techniques to build a transformer temperature identification model.

Key Takeaways:

  • The ultrasonic temperature measurement method introduced in this research uses a combination of machine learning algorithms and ultrasonic sensing to accurately diagnose internal transformer temperatures.
  • The method involves establishing an experimental platform to collect acoustic data at various temperatures, analyzing interference noise from transformer core vibrations, and filtering out magnetically induced noise.
  • The researchers used three machine learning algorithms: random forest (RF), k-nearest neighbor (KNN), and support vector machine (SVM) to build a transformer temperature identification model.
  • The models achieved recognition accuracies of 88.57%, 94.29%, and 91.43%, respectively, indicating the proposed method's ability to accurately diagnose internal transformer temperatures.
  • The researchers have optimized feature dimensionality using two feature dimensionality reduction methods: RF, and correlation-based feature selection (CFS).
  • The study has significant implications for engineering practice, particularly in the diagnosis and maintenance of oil-immersed power transformers.
  • The research was funded by the National Natural Science Foundation of China (NSFC) and the Key Research and Development Program of Shandong Province, China.

Statistics:

  • 88.57% recognition accuracy achieved by the random forest (RF) model
  • 94.29% recognition accuracy achieved by the k-nearest neighbor (KNN) model
  • 91.43% recognition accuracy achieved by the support vector machine (SVM) model
  • 2 feature dimensionality reduction methods used: RF and correlation-based feature selection (CFS)
  • 3 machine learning algorithms used: random forest (RF), k-nearest neighbor (KNN), and support vector machine (SVM)

Sources:

  • A Novel Acoustic Temperature Measurement Technology of Transformers Based On Ultrasonic Sensing. Ieee Sensors Journal, 2025;25(15):29890-29901.
  • Institute of Electrical and Electronics Engineers - www.ieee.org/;
  • Ieee Sensors Journal - ieeexplore.ieee.org/xpl/RecentIssue.jsp?punumber=7361

Read more