VLF至MF频段雷电电场变化波形分类数据集  

A dataset of lightning electric field change waveforms ranging from VLF to MF frequency band

作　　者：王宇[1,2] 肖力郎 贺恒鑫 傅中[4] 程晨 谷山强 陈维江[5] WANG Yu;XIAO Lilang;HE Hengxin;FU Zhong;CHENG Chen;GU Shanqiang;CHEN Weijiang(State Grid Electric Power Research Institute Wuhan NARI Limited Liability Company,Wuhan 430206,P.R.China;State Grid Electric Power Research Institute Co.,Ltd.,Nanjing 211199,P.R.China;School of Electrical and Electronic Engineering,Huazhong University of Science and Technology,Wuhan 430032,P.R.China;State Grid Anhui Electric Power Co.,Ltd.,Electric Power Research Institute,Hefei 230022,P.R.China;State Grid Corporation of China,Beijing 100031,P.R.China)

机构地区：[1]国网电力科学研究院武汉南瑞有限责任公司,武汉430206 [2]国网电力科学研究院有限公司,南京211199 [3]华中科技大学电气与电子工程学院,武汉430032 [4]国网安徽省电力有限公司电力科学研究院,合肥230022 [5]国家电网有限公司,北京100031

出　　处：《中国科学数据（中英文网络版）》2024年第4期407-416,共10页China Scientific Data

基　　金：国家电网有限公司总部科技项目(5500-202120583A-0-5-SF)。

摘　　要：雷电作为一种复杂的大气放电现象,其放电过程中会产生多种类型的电场变化波形。对这些波形进行准确地识别不仅是提升雷电监测系统性能的关键,也直接关联到对雷电物理过程的深入理解。本研究基于2022至2023年间于安徽合肥市部署的13套雷电电场变化测量设备,综合收集了合肥及周边地区雷暴活动期间的电场变化波形数据。数据时间跨度覆盖3月至10月合肥主要雷电活动季节,空间跨度涉及整个城市及其周边地区,确保了数据的代表性与全面性。在数据质量控制方面,采用了严格的数据筛选与预处理流程,确保了数据集的准确性和可靠性。采用先进的卷积神经网络(CNN)模型,对波形数据进行了系统分类,成功辨识出了四种雷电放电过程中的电场变化波形,具体包括地闪回击过程(RS,15290例)、初始击穿过程(PB,3946例)、双极性窄脉冲事件(NBE,3919例)以及一般云闪过程(IC,4111例)。分类模型的准确率超过99%,展现了高效的识别能力。本研究创建了一份大样本量、高质量分类的雷电电场变化波形分类数据集,本数据集不仅能够支持开发更为先进的雷电波形分类算法,促进雷电监测、核爆电磁脉冲监测等技术的进步,也可为深入理解雷电物理过程提供重要的数据基础。Lightning,a complex atmospheric discharge phenomenon,generates various electric field change waveforms during its discharge process.The accurate identification of waveforms is crucial for enhancing the performance of lightning monitoring systems and having a deeper understanding of the physical processes behind lightning.Based on 13 sets of lightning electric field change measuring equipment deployed in Hefei,Anhui province,from 2022 to 2023,we collected electric field change waveform data during thunderstorm activities in Hefei and its surrounding areas for the study.The data span from March to October,covering the main lightning activity season in Hefei and extending to the entire city and its surrounding areas,ensuring the representativeness and comprehensiveness of the data.To ensure the accuracy and reliability of the dataset,strict data screening and preprocessing procedures were implemented.An advanced convolutional neural network(CNN)model was used to systematically classify the waveform data,successfully identifying four types of electric field change waveforms during the lightning discharge process,including Return Stroke(RS,15290 cases),Preliminary Breakdown(PB,3946 cases),Narrow Bipolar Events(NBE,3919 cases),and Intra-cloud Discharge(IC,4111 cases).The classification model achieved an accuracy exceeding 99%,demonstrating its efficient recognition capabilities.This paper has established a large-scale,high-quality classified dataset of lightning electric field change waveforms.This dataset can support the development of more advanced lightning waveform classification algorithms and promotes the advancement of technologies in lightning monitoring and nuclear explosion electromagnetic pulse monitoring.Moreover,it is expected to provide an important data foundation for a deeper understanding of the physical processes behind lightning.

关 键 词：雷电电场变化波形 地闪 云闪 初始击穿过程 双极性窄脉冲事件 

分 类 号：P427.3[天文地球—大气科学及气象学] P413