摘要
故障电弧是引发电气火灾事故的主要原因之一。本文将自组织映射神经网络引入故障电弧研究领域,进行不同负荷情况下的故障电弧聚类分析。首先参照美国UL1699标准进行实验采集电流数据,然后利用自组织映射神经网络实现可视化聚类,并结合k均值法确定聚类结果。根据聚类结果分析故障电弧,对比故障与正常时的差异所在,提取故障电弧的典型特征。最后总结出故障电弧电流通常具有电流短时为零、正负半周差异大、幅值变化大等特征,为故障电弧保护技术提供参考。
Arc fault is one of the prime reasons causing electrical fire accidents.This paper carries out cluster analysis of arc faults under different loads through using self-organizing map neural network.Firstly,data are collected in the experiments based on UL1699.Then clustering is visualized using SOM neural network and K-means cluster.According to the clustering results,some typical characters of arc fault are found,such as short-time zero current,asymmetrical waveform and amplitude variation,which can provide a reference for arc fault protection.
出处
《仪器仪表学报》
EI
CAS
CSCD
北大核心
2010年第3期571-576,共6页
Chinese Journal of Scientific Instrument
关键词
故障电弧
自组织映射
聚类分析
电气火灾
arc fault
self-organizing map
cluster analysis
electrical fire
