Data-driven classification model of lightning generated LF/VLF radiation electric field waveforms

Abstract:
A convolutional neural network (CNN) model is proposed for the classification of very low frequency and low frequency (VLF/LF) lightning electric field waveforms. This model adopts multi-scale convolutional kernels and shortcut connections to enhance the ability of lightning waveform classification. Based on the data recorded from five provinces in China, the proposed model achieves an accuracy of 99.6% for a four-type classification task including return strokes (RS), the intra-cloud (IC) lightning, preliminary breakdown (PB), and narrow bipolar events (NBE). Compared with classic machine learning methods and deep learning methods, the proposed model performs better in classification accuracy and convergence speed. Using the knowledge distillation method, a model suitable for low-computing-power platforms is obtained. The distilled model takes only 59ms for single classification and has a classification accuracy of 99.0%, demonstrating reliable application of the proposed model on low-computing-power platforms.