Research on Partial Discharge Pattern Recognition and Fault Diagnosis of Switchgear Based on Simplified Features in Power Internet of Things

This paper has provided a method for partial discharge pattern recognition and fault diagnosis of switchgear based on simplified features in power internet of things (PIOT). An experimental platform was built, and four typical defect models were designed for simulation of partial discharge in switchgear. Then, transient earth voltage (TEV) was measured and twenty-one features were extracted based on phase and time-series. On this basis, the method of maximum variance analysis was applied for selecting features with high correlation to form the simplified feature set for reducing data transmission pressure in PIOT. Finally, the algorithm of k-nearest neighbor (KNN) was adopted to verify the effectiveness of simplified features for partial discharge pattern recognition. Results show that the average accuracy of pattern recognition based on simplified features is over 90%, and this method can be used for practical application which can reduce the pressure of data transmission and improve the operating efficiency of the monitoring system in PIOT.