Experimental Data Analysis of Positive Streamer-Leader Dynamics in Long Air Gaps Under Slow Front Impulse Voltages Using Machine Learning

Abstract

Knowledge of the electrical discharge characteristics under various voltage conditions is crucial to designing safer and more efficient high-voltage insulation systems. This study presents positive streamer-leader dynamics in the 10-meter rod-plane air gap under slow front positive impulse voltage having a rise time of 1000 microseconds. The realization aims to improve the knowledge of long-gap discharge behavior, which is one of the key aspects in insulation design under high-voltage engineering. The voltage and the current waveforms obtained during experiments were analyzed using a machine-learning-based polynomial regression approach. Besides such analysis, image processing was applied to high-speed camera footage to determine arc lengths for different breakdown stages. Down-sampling was applied to cope with raw data, and the regression models were evaluated in terms of mean squared error (MSE) and R-squared values. The Polynomial regression analysis showed high accuracy in terms of MSE and R-squared values. The image-based analysis demonstrated that a final jump length of nearly 10 m substantiates full leader development to the plane electrode. The results indicate that machine learning and image analysis can accurately model and quantify discharge development in long air gaps. © 2025 IEEE.