•  
  •  
 

Keywords

gas insulated switchgear, particle discharge, mechanical fault, ultrasound, vibration, random forestidentification

Abstract

Metal particle discharge and mechanical faults inside gas insulated switchgear (GIS) are key factors affecting its safety performance. These faults can be detected and identified using the ultrasonic and vibration signals generated by them. Simulated experiments of four types of particle discharge, normal operation, and flange bolt loosening were designed on a GIS platform, and the time-frequency domain characteristics of the signals were analyzed. It is found that, in the ultrasonic signals, the frequency bands of linear, spherical, and lump particle discharges are mainly concentrated in 50–60 kHz, while that of flaky particle discharge is concentrated in 15–30 kHz; in the time domain, the linear particle fluctuation frequency reaches nearly 200 times at the fastest, and the lump particle signal decays by 68.23%, showing the highest decay degree; in the vibration signals, the main frequency of normal operation is 100 Hz, and the amplitude of the 100 Hz frequency significantly decreases when the bolt is loosened. Based on the signal characteristics, the positive/negative half-cycle peak ratio, ringing times, attenuation coefficient, as well as the centroid frequency, frequency amplitude ratio, and total harmonic distortion were extracted. Combining the characteristics of ultrasonic and vibration signals and the random forest algorithm, effective identification of six operating states is successfully achieved, and the recognition accuracy is improved.

DOI

10.19781/j.issn.1673-9140.2026.03.028

First Page

307

Last Page

321

Download

Share

COinS