摘要
由于变电站现场局部放电检测易于受到各类干扰的影响,造成检测数据中存在噪声,使数据中信号特征不明显,因此传统的基于统计特征的模式识别方法在应对现场检测数据时识别率较低。提出了一种基于深度稀疏降噪自编码器网络的模式识别方法。对试验检测出的典型特征图谱,利用深度稀疏降噪自编码器进行主动染噪学习训练,最后得到可以有效去噪的深度特征提取模型,并利用Softmax分类器输出识别结果。利用在变电站现场实测数据对方法进行验证,并与传统的识别方法进行对比,证明方法对含有噪声的局部放电信号有更好的识别效果。
As partial discharge（ PD） detection on substation site is easily affected by various disturbances,test data are polluted with noise and do not contain obvious signal features. Therefore,the traditional pattern recognition method based on statistical features has a low recognition rate with respect to site test data. In this background,this paper proposed a pattern recognition method based on the stacked de-noising auto-encoder（ SDAE） network. The typical characteristic spectrum detected in the test received an active noising training on the basis of the SDAE. Finally,a depth character extraction model with effective de-noising was obtained and recognition results were given through Softmax classifier. Verification through measurement data obtained from substation site and comparison with the traditional recognition method proved that the proposed approach could produce better recognition effect for noise-containing partial discharge signals.
作者
郗晓光
何金
曹梦
陈荣
宋晓博
李苏雅
Xi Xiaoguang;He Jin;Cao Meng;Chen Rong;Song Xiaobo;Li Suya(State Grid Tianjin Electric Power Co.Electric Power Research Institute,Tianjin 300384,Chin)
出处
《电气自动化》
2018年第4期115-118,共4页
Electrical Automation
关键词
深度稀疏降噪自编码器
去噪
局部放电
特征提取
模式识别
stacked de-noising auto-encoder network （SDAE）
de-noising
partial discharge
feature extraction
pattern recognition
