基于数据驱动及无监督分析的电缆故障检测与诊断方法

Cable fault detection and diagnosis based on data-driven and unsupervised analysis

在电缆故障检测与诊断中,常规的诊断方法随着电缆运行时间的增加,其内部时间窗的标准差越来越高,导致该方法的可靠性越来越差。为了解决这一问题,提出基于数据驱动及无监督分析的电缆故障检测与诊断方法。通过重采样采集电缆工作原始数据,去除其直流偏移量,针对数据中含有的不同类型噪声,采用阈值法将噪声去除,采用无监督分析技术检测电缆故障,完成检测后,利用数据驱动技术构建特征参数储备池,计算原数据的特征贡献率,诊断电缆故障类型。实验结果表明,设计的基于数据驱动及无监督分析的诊断方法检测率在95%以上,误检率在0.5以下,样本数据更完整,该诊断方法的可靠性得到了提高。

In cable fault detection and diagnosis,with the increase of cable running time,the standard deviation of the internal time window becomes higher and higher,which leads to the deterioration of reliability of the conventional diagnosis method.In order to solve this problem,a cable fault detection and diagnosis method based on data-driven and unsupervised analysis is proposed.Via resampling collecting original data cable work,remove the dc offset,in view of the data contained in the different types of noise,threshold value method is used to remove noise,using unsupervised detection cable fault analysis technology,complete detection,using the data characteristic parameters of drive technology reserve pool,to calculate contribution of original data characteristics,diagnostic cable fault type.The experimental results show that the detection rate of the designed diagnostic method based on data-driven and unsupervised analysis is more than 95%,and the error detection rate is less than 0.5.The sample data is more complete,and the reliability of the diagnostic method is improved.