摘要
低压串联电弧电流为非平稳信号,故障特征区分度低且具有随机性,给电弧故障特征提取和准确检测带来困难,提出了基于功率谱密度与随机配置网络的低压串联电弧故障检测方法。首先,搭建了串联电弧故障发生平台,采集不同负载类型的电流数据,构建数据集。其次,采用功率谱密度对电流信号执行随机信号分析,实现对电流信号的定量化频域特征描述,增强故障电流与正常电流特征的区分度。然后,采用随机配置网络构建串联电弧故障检测模型,将功率谱密度特征用于随机配置网络的自适应训练学习,提升网络训练效率和模型故障检测能力。在本文构建的电流数据集上,串联电弧故障检测的平均准确率达到96.1567%,证明了方法的有效性。
Low-voltage series arc current is a non-stationary signal with low fault feature differentiation and randomness,which brings difficulties to arc fault feature extraction and accurate detection.A low voltage series arc fault detection method based on power spectral density and stochastic configuration network was proposed.First,a series arc fault generating platform was built,current data of different load types were collected,and data sets are constructed.Secondly,the power spectral density was used to perform random signal analysis on the current signal,so as to realize the quantized frequency domain feature description of the current signal,and enhance the distinction between fault current and normal current features.Then,a series arc fault detection model was constructed using a stochastic configuration network,and the power spectral density features were used for adaptive training of the stochastic configuration network to improve the efficiency of network training and the ability of model fault detection.On our current data set,the average accuracy of series arc fault detection reaches 96.1567%,which proves the effectiveness of the method.
作者
李金杰
邹国锋
魏良玉
王玮
傅桂霞
LI Jin-jie;ZOU Guo-feng;WEI Liang-yu;WANG Wei;FU Gui-xia(School of Electrical and Electronic Engineering,Shandong University of Technology,Zibo 255049,China)
出处
《科学技术与工程》
北大核心
2023年第34期14587-14595,共9页
Science Technology and Engineering
基金
山东省自然科学基金(ZR2022MF307)
国家自然科学基金(52077221)。
关键词
串联电弧检测
功率谱密度
随机配置网络
频域特征提取
自适应学习
series arc detection
power spectral density
stochastic configuration network
frequency domain feature extraction
adaptive learning
