Abstract In this paper, the wavelet packet transform (WPT) is applied to analyze the time-frequency features of three kinds of lightning events (the cloud pulse, the return stroke and the stepped leader) and a new identification method based on WPT is proposed. The magnetic-field data used in this paper were recorded by Nanjing Lightning Location Network in 2018. Firstly, the wavelet spectra of three typical lightning events are given and it is found that spectral ranges of these three events are different. The predominant radiation frequency of the stepped leader is the highest and that of the return strokes is the lowest. A total of 232 cloud pulses, 876 return strokes and 373 stepped leaders are analyzed by WPT in order to investigate their behavior in time-frequency domain. The statistical result shows that the cloud pulse, return stroke and stepped leader radiate predominantly in the frequency range 9–56 kHz, 2–14 kHz and 52–236 kHz, respectively. According to the energy distribution characteristics, three indices are proposed to distinguish the different lightning events. It is found that the recognition rate of return stroke is 91% and that of stepped leader is up to 93%.

中文翻译：

---

基于磁场信号小波包变换区分不同雷电事件

摘要 本文应用小波包变换（WPT）分析了三种雷击事件（云脉冲、回击和阶梯先导）的时频特征，提出了一种基于WPT的识别新方法。 . 本文使用的磁场数据是南京闪电定位网2018年记录的。首先给出了三个典型闪电事件的小波谱，发现这三个事件的谱范围不同。阶梯先导的主要辐射频率最高，回程的主要辐射频率最低。WPT 分析了总共 232 个云脉冲、876 个回击和 373 个阶梯前导，以研究它们在时频域中的行为。统计结果表明，云脉冲，回击和阶梯前导分别主要在 9-56 kHz、2-14 kHz 和 52-236 kHz 的频率范围内辐射。根据能量分布特点，提出了三个指标来区分不同的闪电事件。发现回击识别率为91%，梯级领导识别率高达93%。