基于生成对抗网络的电缆局部放电异常自动监测设计

Design of Automatic Monitor for Cable Partial Discharge Anomaly Based on Generative Adversarial Network

电力设备运行过程中,电缆绝缘损伤、内部缺陷、外部环境等因素容易造成电缆局部放电异常,进而引发电力事故和断电故障。为了及时、有效地监测电缆局部放电异常,在生成对抗网络环境下完成电缆局部放电异常自动监测。根据电缆内部结构分析电缆局部放电原因,利用生成对抗网络重构插补缺失数据,获取完整电缆运行数据。建立随机矩阵,获取电缆运行数据的概率密度函数,提取特征向量,构建特征指标矩阵对特征向量实施奇异值分解,辨识电缆局部放电状态,实现电缆局部放电异常的自动监测。实验结果表明:所提方法在提取电缆局部放电信号脉冲时波形振动幅度小且波形完整;电缆局部放电位置定位与实际位置一致;电缆局部放电位置的相对误差低于1%。

During the operation of power equipment,cable insulation damage,internal defects,external environment and other factors are easy to cause local abnormal discharge of cable,and then cause power accidents and power failure.In order to timely and effectively mo⁃nitor local discharge,automatic monitoring of local discharge in the generated network environment is completed.According to the inter⁃nal structure of the cable,the local discharge reasons of the cable are analyzed,and the missing data are inserted to obtain the complete cable operation data.The random matrix is established to obtain the probability density function of the cable operation data,the feature vector is extracted,and the feature index matrix is constructed to implement the singular value decomposition of the feature vector,iden⁃tify the local discharge state of the cable,and realize the automatic monitoring of the cable local discharge anomaly.The experimental results show that the waveform vibration amplitude of the proposed method is small and the waveform is complete when extracting the pulse of cable partial discharge signal.The positioning of partial discharge position of cable is consistent with the actual position The relative error of partial discharge position of cable is less than 1%.