Intelligent fault diagnosis of high-voltage circuit breakers using triangular global alignment kernel extreme...

Area of Science:

• Electrical Engineering
• Machine Learning
• Signal Processing

Background:

• Vibration-based intelligent fault diagnosis is crucial for high-voltage circuit breakers (HVCBs).
• Traditional machine learning methods struggle with diagnostic instability due to sampling asynchrony in HVCBs.
• Fluctuations in control voltage can lead to inaccurate mechanical state recognition from vibration signals.

Purpose of the Study:

• To develop a robust machine learning model for accurate mechanical state recognition in HVCBs.
• To address the challenge of sampling asynchrony in vibration-based fault diagnosis.
• To improve the stability and performance of intelligent fault diagnosis systems for HVCBs.

Main Methods:

• Introduction of an improved kernel extreme learning machine (K-ELM) termed triangular global alignment kernel (TGAK) extreme learning machine (TGAK-ELM).
• Development of the TGAK, an elastic kernel designed to consider all possible sample alignments.
• Utilizing TGAK for a flexible similarity measure between vibration signal samples.

Main Results:

• The proposed TGAK-ELM demonstrated effective fault diagnosis for a 35kV HVCB.
• TGAK-ELM achieved superior diagnostic results compared to other state-of-the-art machine learning methods.
• Experiments on eight diverse datasets from the UCR repository confirmed the broader applicability of TGAK-ELM.

Conclusions:

• The TGAK-ELM offers a significant advancement in vibration-based intelligent fault diagnosis for HVCBs.
• The method effectively mitigates issues caused by sampling asynchrony, leading to enhanced diagnostic accuracy.
• TGAK-ELM shows potential for application in various fields requiring robust time-series analysis.