摘要
为了进行航空交流串联电弧故障检测,该文在115V/400Hz的航空供电条件下按照美标UL 1699以及SAE AS 5692进行了点接触试验、振动试验和截断试验。利用工控机和数据采集卡提取出电流的正常信号和电弧故障信号,根据基于相空间重构的方法和波形比较法计算出电流信号的计盒维数、信息维数、均值比及其标准差和峰峰值。结果发现,发生了电弧故障后,以上特征值相比于正常情况均产生了较大程度的改变。将上述特征值组成电弧故障样本,作为遗传算法优化的BP神经网络的输入数据,将线路是否正常作为输出,进行电弧故障识别。分析结果表明,该方法的识别效率在96%以上。
Point contact test,guillotine test and vibration test,which are based on the American standards containing UL 1699 and SAE AS 5692,were finished under the condition of 115V/400Hz aircraft power supply in order to detect the alternating series arc faults in aviation.The normal and the arc fault current signals were extracted using the industrial personal computer and the data acquisition card,then the box counting dimensions,information dimensions,mean ratios and their standard deviations and peak-to-peak values of the current signals were calculated based on phase space reconstruction and wave comparison.The results show that the huge changes occur in the above characteristic quantities compered with the normal situation during the arc fault.The sample of the arc fault which is constructed by the above characteristic quantities are deemed as the inputs of the BP neural network optimized by genetic algorithm.Whether the circuit normal or not is deemed as the outputs of the BP neural network in order that identify the arc fault.The results show that the recognition efficiency of this method is above 96%.
作者
崔芮华
曹欢
Cui Ruihua;Cao Huan(Electrical Research Institute Hebei University of Technology,Tianjin,300130,China)
出处
《电工技术学报》
EI
CSCD
北大核心
2020年第S01期243-250,共8页
Transactions of China Electrotechnical Society
基金
河北省自然科学基金青年基金（E2015202143）
河北省教育厅青年基金（QN2014148）资助项目
关键词
相空间重构
相平面图
计盒维数
信息维数
遗传算法
BP
神经网络
Phase space reconstruction
phase plane
box counting dimension
information dimension
genetic algorithm
BP neural network
