基于改进Elman神经网络的电缆早期故障分类识别方法  

Cable Early Fault Identification and Classification Based on Improved Elman Neural Network

作　　者：游旺 李文沛 胡泰山 吉慧子 李晟 李福权 YOU Wang;LI Wenpei;HU Taishan;JI Huizi;LI Sheng;LI Fuquan(Shenzhen Power Supply Bureau Co.,Ltd.,Shenzhen 518001,China;China Southern Power Grid Research Institute Co.,Ltd.,Guangzhou 510000,China;Guangdong Electric Power Design and Research Institute,Guangzhou 510000,China)

机构地区：[1]深圳供电局有限公司,广东深圳518001 [2]南方电网科学研究院有限责任公司,广东广州510000 [3]广东省电力设计研究院,广东广州510000

出　　处：《电工技术》2022年第5期154-158,161,共6页Electric Engineering

基　　金：国家自然科学基金(编号51867010);南方电网科学研究院科技咨询项目(编号SEPRI-K205033)。

摘　　要：针对电缆早期故障难以检测及现有技术识别精度低的问题,提出基于改进Elman神经网络的电缆早期故障分类与识别的方法。首先采用小波变换提取过电流信号的特征向量,将其作为Elman神经网络的输入向量,构建故障分类识别模型。为防止训练过程中出现过拟合与训练时间过长问题,利用Dropout技术对Elman神经网络进行改进。最后通过PSCAD/EMTDC搭建仿真模型进行验证,结果表明所提方法能有效识别电缆早期故障,且具有较高的准确率,与BP神经网络相比,性能提升显著。Aiming at the problems of difficult detection of early cable faults and low recognition accuracy of existing technologies,a method of cable early fault recognition and classification based on improved Elman neural network is proposed.Firstly,the wavelet transform is used to extract the feature vector of the overcurrent signal,which is used as the input vector of the Elman neural network to construct the fault recognition classification model.In order to prevent the problems of over-fitting and excessive training time during the training process,the Elman neural network is improved by using the Dropout technology.Finally,a simulation model was built through PSCAD/EMTDC to verify.The results show that the proposed method can effectively identify early cable faults and has a higher accuracy rate.Compared with the BP neural network,the performance is improved significantly.

关 键 词：电缆早期故障 DROPOUT ELMAN神经网络 小波变换 

分 类 号：TM77[电气工程—电力系统及自动化]