Experimental-based models for predicting the flashover voltage of polluted SiR insulators using leakage current characteristics

This study aims to predict the flashover voltage (FOV) of silicone rubber (SiR) insulators. Accordingly, it benefits from two methods, including an artificial neural network (ANN) model and a FOV gradient fitting model. The FOV and leakage current (LC) tests are carried out on one un-aged and three aged specimens under uniform and longitudinal non-uniform pollution circumstances. The proposed ANN model is designed based on equivalent salt deposit density (ESDD), pollution non-uniformity degree, aging time, LC first harmonic magnitude I-1, and total harmonic distortion. Furthermore, a FOV gradient model is proposed based on ESDD, I-1, and creepage distance. To validate the proposed models, the FOV and LC tests are conducted on two different types of SiR insulators. Then, the predicted FOV from the ANN model and calculated FOV from the voltage gradient model are obtained. The results indicate that the relative errors of the ANN model and FOV gradient fitting model are <6.4 and 7.3%, respectively.