Detection and Location of Incipient Cable Fault Section Based on Traveling Wave Feature and Stacked Automatic Encoder
|更新时间:2026-02-13
|
Detection and Location of Incipient Cable Fault Section Based on Traveling Wave Feature and Stacked Automatic Encoder
Vol. 42, Issue 4, Pages: 72-79(2019)
作者机构:
深圳供电局有限公司,
作者简介:
基金信息:
DOI:
CLC:TM755
Published:2019
稿件说明:
移动端阅览
Li Chengsheng. Detection and Location of Incipient Cable Fault Section Based on Traveling Wave Feature and Stacked Automatic Encoder[J]. 2019, 42(4): 72-79.
DOI:
Li Chengsheng. Detection and Location of Incipient Cable Fault Section Based on Traveling Wave Feature and Stacked Automatic Encoder[J]. 2019, 42(4): 72-79.DOI:
Detection and Location of Incipient Cable Fault Section Based on Traveling Wave Feature and Stacked Automatic Encoder
电缆由于不易受天气影响及可靠性高等优点,在配电网中的使用越加广泛。但是,随着电缆运行年限的增加, 电缆会因局部放电等影响而发生故障。电缆早期故障作为永久性故障的前兆,对其进行识别和定位是提高电网安全 性的重要方法。由于过电流扰动信号在时频域下包含丰富的暂态信息,提出了行波特征量与堆叠自动编码器相结合 的电缆早期故障定位方法。首先利用 S 变换对电流信号进行处理,提取行波等特征量,并将其作为输入数据对堆叠自 动编码器进行训练,利用训练好的网络,正确定位早期故障发生的区段。通过在 PSCAD/EMTDC 中搭建仿真模型获 得仿真数据,结果表明,所提的方法具有较高的准确率和较好的泛化性能。
Abstract
Cables are more and more widely used in distribution networks due to their high reliability and not easily affected by weather. However,with the increase of service life of the cable,the cable will fail due to partial discharge and other effects. Incipient cable fault is a precursor to permanent faults. It is an important way to improve grid security for identifying and locating incipient cable fault. Because the overcurrent disturbance signal contains rich transient information in the time - frequency domain,a method of locating incipient cable fault section is proposed using traveling wave feature and stacked automatic encoder ( SAE) together. Firstly,the current signal is processed by S - transform,and the features are extracted,which are used as the input of stack automatic encoder. The well - trained network can correctly locate the incipient cable fault section. The simulation data is obtained by constructing the simulation model in PSCAD/EMTDC. The results show that the proposed method has higher accuracy and better generalization performance.