Identification of Lightning Strikes to Towers Using Their Electric Field Signatures and a Machine Learning Approach

Ziqin Ding, Si Chen, Hanbo Yang, Kang Yang, Vladimir A. Rakov, Joel B. Harley, Yanan Zhu, Istvan Kereszy

Abstract: Lightning often strikes tall (≥200 m) objects and poses a significant threat to the infrastructure. Using the characteristic electric field signatures of the towerterminated lightning, we developed a machine learning (ML) approach to identify strikes to towers in large lightning datasets. The ground-truth data were acquired at the Lightning Observatory in Gainesville (LOG), Florida. The classification model used in this study is based on a supervised multilayer perceptron model, aiming to capture complex pulse patterns with the neural network architecture. The results show that tower-terminated lightning and nontowerterminated lightning can be accurately identified. We found that the model performance depends on the configuration of the training dataset. Using the local interpretable model-agnostic explanations (LIMEs), we found that our ML model is capable of capturing the key features of electric field signatures produced by lightning strikes to tall towers.

External IDs:doi:10.1109/jsen.2025.3637548