Introducing a fast and efficient machine learning model for lightning localization via Lightning-Induced Voltages on Transmission Lines

Lightning localization is crucial for a wide range of applications. Conventional lightning localization methods typically use at least four antennas as sensors for detecting lightning electromagnetic radiation. One of the complications of these methods is the need for sensor synchronization. Most papers on lightning localization highlight the issues of complexity and synchronization requirements. In this paper, we propose a fast learning model using the XGBoost algorithm for lightning localization. This method utilizes two sensors to receive signals from the induced voltages on transmission lines, eliminating the need for synchronization. By employing the principal component analysis (PCA) algorithm, the input dimensions of the model are reduced, which decreases model complexity, speeds up calculations, reduces hardware resource usage, and enhances accuracy. By training the model with varying numbers of principal components, we can identify the smallest input dimensions that maintain model accuracy. The final model is evaluated using a noisy test dataset. The evaluation results show that the model achieves accuracy (R²) of over 99%. Additionally, studies indicate that the model's accuracy depends on the configuration of the transmission lines and the position of the sensors.