Random Convolutional Kernel Transform with Empirical Mode Decomposition for Classification of Insulators from Power...

Area of Science:

• Electrical Engineering
• Signal Processing
• Machine Learning

Background:

• Reliable electrical energy supply depends on functional insulators.
• Ultrasound time series data from insulators can identify faults.
• Existing methods require enhancement for accurate insulator fault classification.

Purpose of the Study:

• To propose a novel strategy for insulator fault diagnosis.
• To evaluate the efficacy of combining empirical mode decomposition (EMD) methods with Rocket algorithms.
• To improve the accuracy and reliability of electrical insulator fault detection.

Main Methods:

• Utilized the Random Convolutional Kernel Transform (Rocket) algorithm for feature extraction from time series ultrasound data.
• Integrated three EMD methods: Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN), Empirical Wavelet Transform (EWT), and Variational Mode Decomposition (VMD).
• Applied logistic regression classifiers to diagnose insulator faults using the extracted features.

Main Results:

• The combination of EMD methods and MiniRocket significantly boosted logistic regression accuracy.
• Achieved high diagnostic accuracies: 0.992 with CEEMDAN, 0.995 with EWT, and 0.980 with VMD.
• Demonstrated superior performance compared to baseline methods without EMD.

Conclusions:

• Empirical mode decomposition methods, when integrated with MiniRocket, offer a powerful approach for insulator fault diagnosis.
• The proposed strategy enhances the accuracy and dependability of power system monitoring.
• This research highlights the potential for improved safety and reliability in electrical infrastructure.