Intelligent Prediction of Dielectric Strength for Long Air Gaps at High Altitudes Based on Regularization-Logistic Regression

Abstract

Air discharge originates from the interaction between electric field (EF) and the meteorological environment, and the dielectric strength of a long air gap is affected by both EF distribution and atmospheric parameters. A regularization-logistic regression (R-LR) model is proposed for the accurate calculation of long air gap discharge voltage at high altitudes. Using 4 meteorological features and 9 EF features as inputs for the R-LR model, the R-LR model was trained using discharge test data from rod-plane air gaps at altitudes of 55–4300 m. The trained model was used to predict the dielectric strength of rod-plane air gaps at an altitude of 5000 m. The prediction results of the R-LR model were compared with those of random forest (RF), support vector classifier (SVC) and k-nearest neighbor (KNN) models. The results showed that the accuracy and generalization of the R-LR model were better than those of the other models. The MAPE of the R-LR model on the test set was 1.32%. The model was validated by using the rod-plane air gap discharge test data under different meteorological environments in a plain area. The predicted discharge voltage values were basically consistent with the test values, further demonstrating the effectiveness and generalizability of the model. This study can offer a reference for predicting the dielectric strength of air gaps under different meteorological environments in high altitude areas. © 2025 Institute of Electrical Engineers of Japan and Wiley Periodicals LLC.