多光谱信息下的油纸绝缘局部放电深度学习融合诊断方法  

A deep learning fusion diagnosis method of partial discharge in oil-paper insulation based on multi-spectral information

作　　者：胡蝶 李晓枫 姜晓峰 阳瑞霖 刘阳[2] 董明[2] HU Die;LI Xiaofeng;JIANG Xiaofeng;YANG Ruilin;LIU Yang;DONG Ming(Wuling Power Corporation,Changsha 410004,China;State Key Laboratory of Electrical Insulation and Power Equipment(Xi’an Jiaotong University),Xi’an 710049,China)

机构地区：[1]五凌电力有限公司,湖南长沙410004 [2]电工材料电气绝缘全国重点实验室(西安交通大学),陕西西安710049

出　　处：《电工电能新技术》2024年第10期85-92,共8页Advanced Technology of Electrical Engineering and Energy

基　　金：长沙市科技计划项目(kh2201277)。

摘　　要：为了获得更丰富的放电特征信息,提高局部放电诊断的效率,提出了一种多光谱信息下的油纸绝缘局部放电深度学习融合方法。首先,基于微型光学传感器件和局部放电光谱分布,构建了局部放电多光谱同步检测平台,通过实验获得了四种放电类型的多光谱数据;然后构建卷积神经网络模型,将不同光谱段的局部放电数据作为模型不同通道的输入,利用通道级融合提取多光谱信号中的有效信息,对油纸绝缘局部放电类型进行准确辨识。结果表明:不同放电类型的多光谱信息可以作为模式识别的有效特征;引入多光谱信息后,本文所提方法的识别准确率可以达到98%以上,相比于仅使用脉冲电流信号有着明显的提升;相比于统计特征参数分析法和深度神经网络,提出的方法对多光谱信息融合的效果更好,识别的准确率更高。In order to obtain more abundant discharge characteristic information and improve the efficiency of partial discharge diagnosis,a deep learning fusion method of partial discharge in oil-paper insulation based on multi-spectral information is proposed.Firstly,based on the micro-optical sensor and the spectral distribution of partial discharge,a multi-spectral synchronous detection platform for partial discharge is constructed,and the multi-spectral data of four discharge types are obtained through experiments.Then the convolution neural network model is constructed,and the partial discharge data of different spectral sections are used as the input of different channels of the model.The effective information in the multispectral signal is extracted by channel level fusion,and the partial discharge types of oil-paper insulation are accurately identified.The results show that the multi-spectral information of different discharge types can be used as an effective feature of pattern recognition;by the introduction of multi-spectral information,the recognition accuracy of the proposed method can reach more than 98%,which is significantly improved compared with that of only using pulse current signals;compared with statistical characteristic parameter analysis and deep neural network,and the proposed method has better effect on multi-spectral information fusion and higher recognition accuracy.

关 键 词：油纸绝缘 局部放电 深度学习 多光谱信息 模式识别 

分 类 号：TM855[电气工程—高电压与绝缘技术]