Flashover Identification of Aged Silicone Rubber Insulators Using Partial Discharge Signal Processing and Machine Learning Random Forest Algorithm

Abstract

Due to a variety of factors including light weight, effortlessness of moving and installation, towering automatic power to heaviness percentage, destruction resistance, aesthetic appearance, and better padding piece, silicone rubber outdoor insulators have replaced ceramic units in distribution networks. The flashover analysis of long-term aged silicone rubber insulators is the subject of this research. The investigation is base on the limited release approach, which is a very effective and efficient method that is nearly universally used in high-voltage transmission and distribution systems. On aged polymeric insulators, laboratory experiments are conducted under ac voltage, at various humidity levels, and at various infectivity level with NACL impurity. The insulator was thermally aged for 100 days in the laboratory within the heat cavity.This approach simulated the conditions encountered in real-time field applications. The partial discharge signals are collected using a PD monitoring device that can capture PD waveforms as well as phase resolved patterns.The time domain and frequency domain features of PD pulses are investigated in order to make a comparison between new and old polymeric insulators. To detect insulator flashover due to pollution, a machine learning-based Random forest (RF) classifier technique is used.The results show that the performance of outdoor polymeric insulators degrades with age, and that flashover of silicone rubber insulators may be detected using partial discharge pattern analysis and an RF classifier model.