电瓷避雷器Issue(2)：1-9,9.DOI:10.16188/j.isa.1003-8337.2025.02.001
基于自适应DBSCAN和XGBoost的雷电临近预测
Thundercloud Proximity Prediction Based on Adaptive DBSCAN and XGBoost
摘要
Abstract
Lightning is a serious threat to the safe and stable operation of power systems and transmission lines.In order to improve the lightning prediction accuracy,a method is proposed to identify,track and predict thunderclouds based on historical and real-time data from lightning location systems.The pro-posed multi-density adaptive DBSCAN clustering can generate neighbourhood radius and density thre-shold parameters according to the spatial distribution characteristics of the lightning dataset itself,redu-cing the influence of human subjective setting parameters on the clustering results.The real-time light-ning location data are clustered to form thunderclouds,the thundercloud centre of mass is calculated,the thundercloud centre of mass is used to replace the position of the thundercloud,and XGBoost is intro-duced to learn the historical thundercloud movement pattern to dynamically predict the thundercloud movement trend,and the inverse distance weighting method is used to calculate the thundercloud range.A strong thunderstorm activity in 2018 is used for simulation analysis.The results show that the method has accurate prediction results,with an average offset error within 2 km,a prediction accuracy of over 80%,and a false alarm rate below 40%.It provides a reference basis for proactive lightning prevention and lightning warning.关键词
雷电定位/临近预测/自适应DBSCAN/XGBoostKey words
lightning localization/proximity prediction/adaptive DBSCAN/XGBoost引用本文复制引用
薄轶帅,刘立群,李沛智..基于自适应DBSCAN和XGBoost的雷电临近预测[J].电瓷避雷器,2025,(2):1-9,9.基金项目
山西省基础研究计划面上项目(编号:202203021221153). Project supported by Basic Research Program General Project of Shanxi Province(No.202203021221153). （编号:202203021221153）
