河北工业科技2025，Vol.42Issue(6)：510-516,565,8.DOI:10.7535/hbgykj.2025yx06002
基于神经网络自适应训练的输电电缆局放检测技术
Partial discharge detection technology for transmission cables based on adaptive training of neural networks
摘要
Abstract
To enhance the accuracy and reliability of partial discharge(PD)detection in transmission cables and address the issue of detection deviation caused by the inability of traditional methods to learn the characteristics of different discharge modes,a transmission cable PD detection technology based on neural network adaptive training was proposed.Firstly,constraint conditions were generated based on the adaptive wavelet transform points.The Sigmoid function was output according to the cluster transfer relationship of the wavelet function.The feature detection fitness was obtained through adaptive training of the neural network,and the cable partial discharge feature extraction model was constructed;Secondly,empirical mode decomposition was carried out on the extracted partial discharge features to obtain multi-scale energy features;Finally,based on the multi-scale energy features obtained from decomposition,the mean sequence of the partial discharge detection sequence samples was calculated to achieve partial discharge detection of transmission cables.The results show that the output pulse amplitudes of the proposed cable partial discharge feature extraction model are stable within the range of 100～170 pC under different boost conditions,effectively capturing the time-frequency characteristics of the partial discharge signal.Empirical mode decomposition methods can successfully separate essential mode components from complex partial discharge signals.In the detection of three types of defects on cable insulation,namely insulation scratches,tip discharge and moisture,the fitting degree of the partial discharge quantity aggregation range and the preset range of the proposed technology reach over 95%,effectively solving the problem of signal recognition under aliasing interference.Compared with the traditional methods,the proposed technology can more precisely match the preset partial discharge quantity range,and has higher detection accuracy and reliability.This research can provide a high-precision and highly reliable solution for partial discharge detection of transmission cables,providing effective technical support for ensuring the safe and stable operation of the power system.关键词
电力电子技术/神经网络/自适应训练/输电电缆/局放检测/自适应小波变换Key words
power electronics technology/neural network/adaptive training/transmission cables/partial discharge testing/adaptive wavelet transform分类
信息技术与安全科学引用本文复制引用
佘强,靳继勇,陈俊德..基于神经网络自适应训练的输电电缆局放检测技术[J].河北工业科技,2025,42(6):510-516,565,8.基金项目
国网青海省电力公司科技项目(522801240004) （522801240004）
