摘要
为了有效地解决配电网故障类型辨识困难的问题,提出了一种基于相电压图像生成的配电网故障类型辨识方法。首先,提取故障前与故障后一个周期的三相电压信号,并通过模变换矩阵将获取的三相电压信号转换为线模分量;其次,利用格拉姆角和场以及格拉姆角差场将电压信号的线模分量转换为含有丰富故障特征的图像,并将这两种特征图进行空间域图像融合,融合后图像的点线面等信息可以充分反映当前系统的运行工况;最后,将融合后的图像输入到卷积神经网络中,通过Softmax函数输出配电网故障类型。实验结果表明,所提方法在高阻接故障时,依然能够准确地诊断出配电网故障类型,且诊断方法具有较高的鲁棒性。
To effectively address the difficulty in identifying the fault type in distribution networks,a fault type identification method based on phase voltage images is proposed.First,the three-phase voltage signals before and after a fault are extracted for one cycle,and the acquired three-phase voltage signals are converted into line-mode components through a modal transformation matrix.Next,the line-mode components of the voltage signal are transformed into images rich in fault characteristics using Gram angle and field,as well as Gram angle difference field.These two feature maps are then fused in the spatial domain.The fused image,which contains information such as points,lines,and surfaces,can fully reflect the operating conditions of the system.Finally,the fused image is input into a convolutional neural network,and the fault type of the distribution network is output through the Softmax function.The experimental results show that the proposed method can accurately diagnose the fault type of the distribution network under complex conditions such as high-resistance grounding,noisy signals and unsynchronized sampling times,showcasing the high robustness of the diagnostic method.
作者
蒋涛
刘刚
吕东飞
郭赢
李子浩
JIANG Tao;LIU Gang;LYU Dong-fei;GUO Ying;LI Zi-hao(State Grid Zibo Power Supply Company,Zibo 255020)
出处
《制造业自动化》
2025年第8期90-98,共9页
Manufacturing Automation
基金
国网山东省电力公司科技项目(520603240009)。
关键词
故障类型辨识
格拉姆角场
卷积神经网络
图像融合
fault identification
Gramian angular field
convolutional neural network
image fusion
