A Novel Mechanical Fault Feature Selection and Diagnosis Approach for High-Voltage Circuit Breakers Using Features...

Area of Science:

• Electrical Engineering
• Power Systems Engineering
• Mechanical Engineering

Background:

• High-voltage circuit breakers (HVCBs) are critical for power system stability.
• Mechanical failures in HVCBs compromise reliability.
• Existing fault diagnosis methods suffer from complex and inefficient feature extraction.

Purpose of the Study:

• To propose a novel approach for mechanical fault feature selection and diagnosis in HVCBs.
• To improve the efficiency of feature extraction without relying on signal processing.
• To enhance the accuracy and reliability of HVCB fault diagnosis.

Main Methods:

• Vibration signals from HVCBs were segmented based on time scale and signal amplitude changes.
• Ensemble learning was used for feature extraction, constructing a feature vector.
• Random Forest (RF) determined feature importance (Gini importance), followed by Sequential Forward Selection (SFS) for optimal subset identification.
• Regularized Fisher's Criterion (RFC) assessed classification ability.
• A hierarchical hybrid classifier (One-Class Support Vector Machine (OCSVM) and RF) performed fault diagnosis.

Main Results:

• The proposed method demonstrated high feature extraction efficiency.
• The approach achieved high recognition accuracy for HVCB mechanical faults using vibration signals.
• Unknown fault types, even from untrained samples, were effectively identified.

Conclusions:

• The novel method significantly improves feature extraction efficiency for HVCB mechanical fault diagnosis.
• The approach offers high accuracy in recognizing both known and unknown fault types.
• This technique enhances the overall reliability and safety of power systems by enabling robust HVCB fault diagnosis.