分段傅里叶神经网络的低频振荡模式识别方法

针对低频振荡带宽较窄、主导模式较少的特点，提出了分段傅里叶神经网络的主导模式识别方法。采用分段傅里叶系数以求得振荡阻尼特性；为克服傅里叶系数直接求解的困难，采用有限神经元的正交基神经网络模型进行求解。根据分段傅里叶系数识别振荡主导模式的频率和衰减因子；再根据其与衰减时间窗的关系得到振荡幅值。该方法既保留了傅里叶算法抗噪性好的特点，又利用神经网络训练，进一步提高了抗噪性和可靠性，并通过算例仿真得到了证明。该研究对电力系统低频振荡的在线动态识别具有实际意义。...

Full description

Saved in:
Bibliographic Details
Published in电力系统保护与控制 Vol. 40; no. 15; pp. 40 - 45
Main Author 竺炜 马建伟 曾喆昭 周益华
Format Journal Article
LanguageChinese
Published 长沙理工大学电气与信息工程学院,湖南长沙410114 2012
中国电力科学研究院,北京100192%长沙理工大学电气与信息工程学院,湖南长沙,410114
Subjects
Prony
主导模式
低频振荡
模式识别
神经网络
主导模式
低频振荡
神经网络
Prony
模式识别
Online AccessGet full text
ISSN1674-3415
DOI10.3969/j.issn.1674-3415.2012.15.008

Cover

More Information
Summary:针对低频振荡带宽较窄、主导模式较少的特点，提出了分段傅里叶神经网络的主导模式识别方法。采用分段傅里叶系数以求得振荡阻尼特性；为克服傅里叶系数直接求解的困难，采用有限神经元的正交基神经网络模型进行求解。根据分段傅里叶系数识别振荡主导模式的频率和衰减因子；再根据其与衰减时间窗的关系得到振荡幅值。该方法既保留了傅里叶算法抗噪性好的特点，又利用神经网络训练，进一步提高了抗噪性和可靠性，并通过算例仿真得到了证明。该研究对电力系统低频振荡的在线动态识别具有实际意义。
Bibliography:Prony; neural network; dominant mode; low frequency oscillation; mode recognition
41-1401/TM
Because of narrow band width and fewer dominant modes of low-frequency oscillations signal, segmental Fourier neural network algorithm is proposed for identifying the dominant mode. The damping factor can be obtained by segmental Fourier coefficients. The orthogonal basis neural network is used to overcome the difficulty of working Fourier coefficient out directly. Frequency of dominant mode can be calculated out by Fourier coefficients; then the amplitude can be calculated by the relationship between the damping time window and Fourier coefficients. This method has good anti-noise ability as Fourier algorithm does. Moreover, due to the use of neural network, the noise immunity and reliability are improved, which are proved by simulation examples. It has practical significance for low-frequency oscillation online dynamic identification.
ZHU Wei, MA Jian-wei, ZENG Zhe-zhao, ZHOU Yi-hua （1. College of Electrical ＆ Informat
ISSN:1674-3415
DOI:10.3969/j.issn.1674-3415.2012.15.008