摘要
宽频振荡模式具有时变性和时空分布特性,对振荡分类方法在准确性、自适应性等方面提出了更高的要求。为此,该文提出一种基于格拉姆差场(gram difference field,GADF)和卷积神经网络(convolutional neural network,CNN)相结合的宽频振荡分类方法。首先,利用GADF将宽频振荡一维时间序列转换为二维特征图,保留了数据对时间的依赖性和数据间存在的潜在联系特征。然后,通过CNN对GADF特征图自适应地完成宽频振荡模态特征的检测和分类。仿真和实测数据分析结果表明,GADF-CNN方法可以有效检测宽频振荡类型,具有更高的分类检测准确率和自适应性。
Frequency oscillation mode has time-varying and spatio-temporal distribution characteristics, which requires higher accuracy and adaptability of the classifications.Therefore, this paper proposes a broadband oscillation detection based on the combination of the Gram difference field(GADF) and the convolutional neural network(CNN).Firstly, the GADF is used to transform the 1d time series of the broadband oscillations into the 2d feature graphs, preserving the time-dependence of the data and the potential relationship between the data. Then, the CNN adaptively detects and classifies the broadband oscillation modal features in the GADF feature graphs. Simulation and measured data analysis results show that the GADF-CNN method can effectively detect the broadband oscillation types, and has higher classification detection accuracy and adaptability.
作者
赵妍
唐文石
聂永辉
王泽通
ZHAO Yan;TANG Wenshi;NIE Yonghui;WANG Zetong(School of Power Transmission and Distribution Technology,Northeast Electric Power University,Jilin 132012,Jilin Province,China;School of Electrical Engineering,Northeast Electric Power University,Jilin 132012,Jilin Province,China;Academic Affairs Office,Northeast Electric Power University,Jilin 132012,Jilin Province,China)
出处
《电网技术》
EI
CSCD
北大核心
2022年第11期4364-4372,共9页
Power System Technology
关键词
宽频振荡分类
格拉姆角差场
卷积神经网络
深度学习
broadband oscillation classification
Gram angle difference field
convolutional neural network
deep learning
