Characterization and Machine Learning Prediction of Dielectric Strength on Coated Glass Insulators

This paper investigates the impact of SiO2/polyurea concentration and type of coating layer on the dielectric strength (DS) of high-voltage coated glass insulators in the presence of contamination and wet. Using the cold spring method, a superhydrophobic coating consisting of polyurea (Pu) loaded with Nano- and Micro-Titanium Dioxide (SiO2) with varying weight percentages (wt%) was applied to the insulator surface. The main aim of this paper is to use three typical machine learning (ML) algorithms including radial basis function artificial neural network (RBF-ANN), support vector machine (SVM), and random forest (RF) to predict the DS under the coating effect. The data collected from the experiment was entered into the proposed algorithms, and the Colab tool was utilized to create and train the ML algorithms. Experimental results revealed that micro-coating, full coating, and two layers gave the maximum DS. The RP approach introduces high-performance prediction with correlation coefficient R2/Root Mean Square Error (RMSE) reaching 0.971/1.9kV mm-1.