Novel Diagnosis Method for GIS Mechanical Defects Based on an Improved Lightweight CNN Model With Load Adaptive Matching

Mechanical defects of GAS-insulated metalenclosed switchgear (GIS) equipment seriously threaten power grid security, but ON-site complex operating conditions create great challenges for defect diagnosis. Therefore, this article proposes a novel diagnosis and state assessmentmethod for GIS mechanical defects under varying currents. First, a time-frequency analysis method of GIS vibration signals based on neighboring mode noise suppression was proposed, and a defect type and severity diagnosis model with load adaptive matching was designed with the improved SqueezeNet. Then, vibration datasets with different severities and currents of typical defects were established based on 110 kV GIS mechanical vibration platform. Finally, model validation and comparison analysis are carried out. Results show that the proposed method effectively mines feature and achieves accurate multiobjective diagnoses of the defect type and severity under varying currents, which is more accurate than traditional methods. The smaller model and faster computing speed are more suitable for edge deployment application.