Electric Field Feature Extraction and Breakdown Voltage Prediction of Sphere Gaps with Quasi-uniform Field

Abstract

The distribution of electric field (EF) has a strong correlation with air insulation strength. In order to obtain air gap breakdown voltage (BV) through simulating calculation instead of experimental measurement, various parameters related to the electrostatic field distribution are used to characterize the sphere gap structure, which are defined on the shortest path between two sphere electrodes. These features of different sphere gaps are extracted from EF simulation results and input to a support vector machine (SVM) model optimized by the grid search algorithm, so as to establish their correlations with the output BVs. The golden section search method is used to predict the BV iteratively. Taking 9 sphere gaps, with the diameter in the range of 15 to 50 cm, as training samples, the SVM model is employed for BV prediction of 107 quasi-uniform sphere gaps with different diameters ranging from 75 to 150 cm and gap distances from 2 to 75 cm. For the predicted results, the average relative error is only 0.97% compared with the test values presented in IEC 60052. The results demonstrate the effectiveness of the EF features and the prediction model. This research offers a reference for air gap BV calculation by numerical approaches.