一种配网停电故障分类和源定位的深度学习方法  

A deep learning method for power failure classification and source location in distribution network

作　　者：卫思明 戴人杰 徐修能 WEI Siming;DAI Renjie;XU Xiuneng(State Grid Shanghai Municipal Electric Power Company,Shanghai 201600,China)

机构地区：[1]国网上海市电力公司,上海201600

出　　处：《电子设计工程》2024年第10期145-148,153,共5页Electronic Design Engineering

基　　金：国网上海松江供电公司2022年度群众性创新科技项目(520935220007)。

摘　　要：配网停电故障时产生的高斯噪声造成数据失稳,影响故障源头定位的精度,为此,提出基于深度学习的配网停电故障分类和源定位方法。构建配网停电故障二元模型,分析配网停电故障数据失稳特征。将所得到的规格化数据用矩阵格式输入到深度神经网络中,获取负载率、三相不平衡度分类指标。分析停电事件信息池和停电故障研判之间的关系,获取故障分类结果。使用深度学习的有监督训练算法,结合S注入法寻找源故障位置。由实验结果可知,该方法故障电流实部在0~0.5 A范围内变化,故障电流虚部在0~1.0 A范围内变化,与实际变化范围一致,说明使用所研究方法能够获取精准的故障定位信号。Gauss noise caused by power failure in distribution network causes data instability and affects the accuracy of fault source location.Therefore,a method based on deep learning for power failure classification and source location in distribution network is proposed.The binary model of distribution network outage fault is constructed to analyze the instability characteristics of distribution network outage fault data.The normalized data are input into the deep neural network in matrix format to obtain the classification indexes of load rate and three⁃phase unbalance.Analyze the relationship between outage event information pool and outage fault research and judgment,and obtain the fault classification results.The supervised training algorithm based on deep learning and S injection method are used to find the source fault location.It can be seen from the experimental results that the real part of the fault current changes within the range of 0~0.5 A,and the imaginary part of the fault current changes within the range of 0~1.0 A,which is consistent with the actual change range,indicating that the method can obtain accurate fault location signals.

关 键 词：配网停电故障 故障分类 源定位 深度学习 

分 类 号：TN07[电子电信—物理电子学] TP391[自动化与计算机技术—计算机应用技术]