Mechanical Fault Diagnosis of GIS Disconnector Based on Synchrosqueezing Wavelet Transform and Stacked Autoencoder
J. Q. Ma, F. H. Wang
Department of Electrical Engineering
Shanghai Jiaotong University
ma136candy@163.com
Purpose/Aim
With the advantage of small footprint, high reliability, long maintenance period, and low maintenance cost, Gas Insulated Switchgear (GIS) has been widely used in the high-voltage power grid. Hence, it is essential to ensure the secure and reliable operation of GIS and the power grid. Influenced by the complicated mechanical structure, mechanical defects are becoming the main types of faults in GIS and have caused a lot of accidents. As one of the essential equipment of GIS, disconnector has the function of switching lines and forming an isolation port. The abnormal mechanical condition of the disconnector will directly lead to the interruption of its opening and closing operation and the delay of power outage and transmission. To carefully recognize the mechanical condition of GIS disconnector, this paper presents a fault diagnosis method consisting of synchrosqueezing wavelet transform and stacked autoencoder based on the vibration signals originating from the operation of GIS disconnector.
Experimental methods
The vibration signals of GIS shell during disconnector operations were tested by the vibration acceleration sensor located in the GIS shell. The typical mechanical faults including the body defect, transmission defect, and mechanism defect were simulated. The corresponding vibration signals during disconnector operations were also collected.
Results/discussion
The time-frequency distributions of vibration signals obtained from synchrosqueezing wavelet transform for different mechanical conditions of GIS disconnector have obvious differences, and the relevant frequencies and amplitudes are strongly differentiated. The dataset is built based on the time-frequency features of vibration signals. Then the stacked autoencoder model is built to identify the different mechanical conditions of GIS disconnector.
Conclusions
The vibration signals during the operation of GIS are closely related to its mechanical condition. The time-frequency distribution of vibration signals of GIS disconnector has better energy aggregation. The proposed fault diagnosis model based on stacked autoencoder can effectively recognize the mechanical conditions of GIS disconnector with high accuracy.