Prediction study on power frequency breakdown voltage of ring structure short air gaps

Abstract

External insulation characteristic is the key issues in the design and operation process of high voltage transportation and transfer facilities, whose fundamental base is the breakdown characteristic of air gap. Applying artificial intelligence algorithm using machine learning to design external insulation of transmission line is one of the new directions in the study of external insulation. Therefore, a new method using the breakdown voltage of typical electrode gaps to predict the breakdown voltage of complex gaps is proposed, which is based on electric field features and support vector machine (SVM). The electric field distribution was calculated by finite element method and the electric field features were extracted to characterize the gap geometries. The prediction model was established by SVM, and the electric field features were set as input parameters of the model, while the output was whether the gap would breakdown under a given voltage. Applying this method, sphere gap and rod-plane gap is chosen as training samples to predict the breakdown voltage of ring-plane and ring-ring gaps and the calculated results are in good agreement with the experimental values by comparison. This method provides a new way for getting the breakdown voltage of complex air gaps. So it is conducive to decrease test work and reduce experimental cost.