
基本信息:
- 专利标题: 基于声纹识别和神经网络的变压器/电抗器故障诊断方法
- 专利标题(英):Transformer/reactor fault diagnosis method based on voiceprint recognition and neural network
- 申请号:CN201910690335.3 申请日:2019-07-29
- 公开(公告)号:CN110534118A 公开(公告)日:2019-12-03
- 发明人: 徐海青 , 季坤 , 赵峰 , 陈是同 , 徐唯耀 , 秦浩 , 王文清 , 王维佳 , 吴立刚 , 付成成 , 浦正国 , 梁翀 , 廖逍 , 张晨晨 , 张天奇 , 余江斌 , 胡心颖 , 韩涛
- 申请人: 安徽继远软件有限公司 , 国网信息通信产业集团有限公司 , 国网安徽省电力有限公司 , 国家电网有限公司
- 申请人地址: 安徽省合肥市高新区习友路1800号
- 专利权人: 安徽继远软件有限公司,国网信息通信产业集团有限公司,国网安徽省电力有限公司,国家电网有限公司
- 当前专利权人: 安徽继远软件有限公司,国网信息通信产业集团有限公司,国网安徽省电力有限公司,国家电网有限公司
- 当前专利权人地址: 安徽省合肥市高新区习友路1800号
- 代理机构: 合肥维可专利代理事务所
- 代理人: 吴明华
- 主分类号: G10L17/26
- IPC分类号: G10L17/26 ; G10L17/18 ; G10L21/10
The invention discloses a transformer/reactor fault diagnosis method based on voiceprint recognition and a neural network. The method includes the steps of: sampling voice signals generated during operation of a transformer/electric reactor; dividing the voice signal into a plurality of voice segments, directly converting the voice segments into a spectrogram, processing the spectrogram by adopting a neural network formed by connecting a CNN network and an LSTM network in series, and finally performing transformer fault diagnosis based on a processing result of the neural network. According tothe invention, the characteristics that the CNN network is high in image processing capability and the LSTM network is easy to model the time series data are utilized; the network advantages of the CNN network and the LSTM are combined; according to the transformer/reactor fault diagnosis method based on voiceprint recognition, the accuracy of fault diagnosis of the transformer/reactor is high, whether the transformer/reactor has an internal fault or not can be recognized, the labor intensity of maintenance personnel can be effectively reduced, and the intelligent level of transformer/reactordetection is improved.
公开/授权文献:
- CN110534118B 基于声纹识别和神经网络的变压器/电抗器故障诊断方法 公开/授权日:2021-10-08
IPC结构图谱:
G | 物理 |
--G10 | 乐器;声学 |
----G10L | 语言分析或合成;语言识别 |
------G10L17/00 | 讲话者辨认或验证 |
--------G10L17/26 | .特殊语音特征的识别,例如测谎器的使用;动物声音识别 |