基于卷积神经网络的绝缘子异常检测方法  

sulator Fault Identification Method Based on Convolutional Beural Network

作　　者：李自乾 庄国峰 夏旺 LI Ziqian;ZHUANG Guofeng;XIA Wang

机构地区：[1]国网平凉供电公司,甘肃平凉744000 [2]国网甘肃省电力公司,兰州730000

出　　处：《科技创新与应用》2024年第27期37-40,共4页Technology Innovation and Application

摘　　要：输电网络是电网“大动脉”,而绝缘子运行状态直接影响整条线路的运行安全和可靠性。为了快速评估绝缘子运行状态,有效提升运行可靠性。传统的故障识别主要依赖人工经验,识别的准确性依赖工程师经验。为此,该文提出一种基于卷积神经网络的电网故障识别方法,相较直接使用故障数据的传统方法,先将故障波形转换为图片,充分发挥卷积神经网络强大的图片处理能力。实验结果表明,该文所训练的卷积神经网络具有很好的准确率,正确率可达到98%。The rapid identification of power grid faults can shorten the outage time and effectively improve reliability.Traditional fault identification mainly relies on manual experience,and the accuracy of identification depends on the experience of engineers.Therefore,this article proposes a power grid fault identification method based on Convolutional Neural Network(CNN).Compared to traditional methods that directly use fault data,this method first converts the fault waveform into images,fully leveraging the powerful image processing ability of Convolutional Neural Networks.The experimental results show that the Convolutional Neural Network trained in this paper has good accuracy,with an accuracy rate of 98%.

关 键 词：故障类型 深度学习 类型识别 图片转换 卷积 

分 类 号：TM216[一般工业技术—材料科学与工程]