基于双通道卷积神经网络的配网线路实时运行状态检测研究  

Research on Real time Operation Status Detection of Distribution Network Lines Based on Dual Channel Convolution Neural Network

作　　者：丘雪娇 程礼隽 QIU Xuejiao;CHENG Lijun(State Grid Fujian Electric Power Co.,Ltd.,Longyan Power Supply Company,Longyan 364000,China;School of Electrical and Electronic Engineering,North China Electric Power University,Beijing 102206,China)

机构地区：[1]国网福建省电力有限公司龙岩供电公司,福建龙岩364000 [2]华北电力大学电气与电子工程学院,北京102206

出　　处：《通信电源技术》2025年第6期77-81,共5页Telecom Power Technology

摘　　要：在配网线路实时运行状态检测中,无人机获取的图像常受天气、光线等因素影响,导致图像质量不佳,且现有图像处理方法无法全面提取复杂的图像特征,导致对配网线路状态的识别不够准确。因此,设计一种基于双通道卷积神经网络的配网线路实时运行状态检测方法。首先,利用无人机获取配网线路运行状态图像,并进行预处理以提高图像质量。其次,采用双通道卷积神经网络提取图像特征,结合一维和二维卷积神经网络全面捕获特征信息,二维卷积用于提取空间特征,一维卷积用于提取时间序列特征,从而提高对复杂图像信息的理解和识别能力。利用双通道卷积神经网络模型定义复杂损失函数来优化网络参数,从而实现状态识别。再次,设计线路状态监测终端,通过计算电流突变量和传播时间定位故障区段,并采用综合感官检查法和先进测距技术进一步锁定故障点。最后,开发一套用户友好的App软件,实时展示线路运行状态和设备信息。实验结果表明,设计方法在配网线路9 km处准确诊断出瞬时电流异常,成功定位故障,展现出较高的运行状态检测与故障定位准确性。In the real-time operation state detection of distribution lines,the images obtained by unmanned aerial vehicles are often affected by weather,light and other factors,which leads to poor image quality,and the existing image processing methods can not fully extract complex image features,which leads to inaccurate identification of distribution lines.Therefore,a real-time operation state detection method of distribution lines based on dual-channel convolutional neural network is designed.Firstly,the unmanned aerial vehicle is used to obtain the running state image of distribution lines,and the image is preprocessed to improve the image quality.Secondly,two-channel convolutional neural network is used to extract image features,and one-dimensional and two-dimensional convolutional neural networks are combined to capture feature information comprehensively,two-dimensional convolution is used to extract spatial features and one-dimensional convolution is used to extract time series features,thus improving the understanding and recognition ability of complex image information.The complex loss function is defined by using the twochannel convolutional neural network model to optimize the network parameters,so as to realize state identification.Thirdly,the line condition monitoring terminal is designed,and the fault section is located by calculating the current mutation and propagation time,and the fault point is further locked by comprehensive sensory inspection and advanced ranging technology.Finally,a set of user-friendly App software is developed to display the line running status and equipment information in real time.The experimental results show that the design method can accurately diagnose the instantaneous current anomaly at 9 km of the distribution line,successfully locate the fault,and show high accuracy of operation state detection and fault location.

关 键 词：双通道 卷积神经网络 配网线路 运行状态 故障检测 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]