基于卷积神经网络的输电线路区内外故障判断及故障选相方法研究

Research on Internal and External Fault Diagnosis and Fault-selection of Transmission Line Based on Convolutional Neural Network

在分析深度学习模型之一——卷积神经网络(convolutional neural network,CNN)的结构及原理的基础上,提出基于卷积神经网络的输电线路区内外故障判断及故障选相新方法。提出了采用两个softmax分类器的CNN网络结构,用同一CNN网络同时解决了区内外故障判断和故障选相两类非独立分类问题,实现了两种非独立分类问题的权值共享。数字仿真实验和实际现场故障数据测试结果表明:文中构建的数学模型,能同时实现区内外故障判断和故障选相,对采样率要求低,不需要整定任何参数,不受系统频率、故障位置、负荷电流、过渡电阻等因素的影响,结果准确可靠。

The structure and principle of one of the deep learning models, convolutional neural network (CNN), were analyzed in this paper. A new method of identification between internal and external faults and fault-selection for transmission line based on CNN was proposed. A CNN network architecture using two softmax classifiers was proposed, and the two kinds of non-independent classification problems were solved by the same network, and the shared weights of the two non-independent classification problems were realized. The digital simulation experiments and on-site recorded fault data test results show that, the mathematical model constructed in this paper can solve the problem of the internal and external fault diagnosis and fault-selection. Besides, this method does not require a high sampling rate and does not need to set various parameters. Also, it is hardly affected by the system frequency, fault location, load current, transition resistance etc. The results are accurate and reliable. © 2016 Chin. Soc. for Elec. Eng.