A Fault Diagnosis Method for Switchgear Based on Acoustic Signatures of Partial Discharge Sounds

High-voltage switchgear plays a critical role in power systems, and its failure can lead to severe electrical accidents. To improve the accuracy and reliability of switchgear fault detection, a fault acoustic signature recognition method based on acoustic patterns is proposed. First, the four common types of partial discharge in switchgear are analyzed, and corresponding experimental models are designed based on the discharge mechanisms. Using a high-sensitivity microphone, discharge sound signals are collected on the experimental platform, along with environmental noise, to construct a high-quality acoustic signature dataset. To enhance the quality of partial discharge sound signals, the collected sound data undergoes preprocessing. To address the noise interference in the actual operating environment, the FastICA algorithm is employed for noise reduction, improving the signal-to-noise ratio. Fault recognition of switchgear discharge sounds is achieved based on the ECAPA-TDNN network, with an identification accuracy of 98.89%.