摘要
为了实现低压串联故障电弧的有效诊断,基于ULI699标准搭建了交流电压为220 V、频率为50 Hz的串联故障电弧实验平台,并对不同负载回路正常工作电流以及串联故障电弧电流进行数据采集,提出基于小波包能量熵的低压串联故障电弧诊断方法.通过对电流信号进行4层小波包分解,提取小波包能量熵作为特征向量描述故障电弧电流信号在不同频段的能量分布.采用主元分析(PCA)法提取特征向量的主元作为BP神经网络的输入,实现样本最优压缩以简化神经网络结构.仿真结果表明,该方法故障诊断准确率较高,能够有效地识别串联故障电弧.
In order to effectively diagnose the low voltage series arc fault, the experimental platform for series arc fault under AC voltage of 220 V and frequency of 50 Hz was established based on the UL1699 standard. The data collection for both normal working current and series arc fault current under different load loops was performed, and the diagnosis method for low voltage series arc fault based on wavelet packet-energy entropy was proposed. Through the four-layer wavelet packet decomposition of current signals, the wavelet packet-energy entropy was extracted as the feature vectors to describe the energy distribution of arc fault current signals under different frequency bands. The principal components of feature vectors were extracted as the input of BP neural network with the principal component analysis （PCA） method. Therefore, the optimum compression of samples was realized to simplify the structure of neural network. The simulated results show that the proposed method has high accurate rate of fault diagnosis, and can effectively identify the series arc fault.
出处
《沈阳工业大学学报》
EI
CAS
北大核心
2013年第6期606-612,共7页
Journal of Shenyang University of Technology
基金
国家自然科学基金资助项目(50877048)
辽宁省教育厅优秀人才支持计划资助项目(LR2011002)
关键词
故障电弧
BP神经网络
故障诊断
电气火灾
主元分析
小波包
小波包能量熵
特征提取
arc fault
BP neural network
fault diagnosis
electrical fire
principal component analysis
wavelet packet
wavelet packet-energy entropy
feature extraction
