Research and Application of Partial Discharge Identification Method of Switchgear Based on Joint Feature Collocation GBDT

The audio signal in the running process of the switchgear contains a lot of operating state information, which is of great significance to the monitoring of the internal insulation fault and abnormal state. In order to accurately identify partial discharge signals and non-partial discharge signals from audio signals collected in the complex environment of switching station, a method of feature extraction and recognition of partial discharge signals based on joint feature and gradient lifting decision tree (GBDT) is proposed. Firstly, the information distribution of partial discharge signal and non-partial discharge signal in time domain, frequency domain and time-frequency domain is analyzed, and corresponding features are extracted to form joint features. Then, GBDT is used to train the model and complete the recognition. Finally, the recognition method is verified, and various combinations are compared with common features and support vector machine (SVM). The recognition accuracy of combined features with GBDT reaches 100%. Based on this method, an online monitoring system is designed, and the audio recording and partial discharge identification results are displayed in the WeChat mini program. The research method and the system designed in this paper can be used for the online analysis and detection of the audio signal of the switchgear, which has certain engineering application value.