基于FastICA技术与TLS-ESPRIT方法的电力系统低频振荡模态辨识  

Power System Low Frequency Oscillation Modal Identification Based on the FastICA Technique and TLS-ESPRIT Algorithm

作　　者：张程[1,2] 刘佳静 匡宇 邱炳林 ZHANG Cheng;LIU Jiajing;KUANG Yu;QIU Binglin(School of Electronic,Electrical Engineering and Physics,Fujian University of Technology,Fuzhou 350118,China;Fujian Colleges and Universities Engineering Research Center of Smart Grid Simulation&Analysis and Integrated Control,Fuzhou 350118,China)

机构地区：[1]福建工程学院电子电气与物理学院,福州350118 [2]智能电网仿真分析与综合控制福建省高校工程研究中心,福州350118

出　　处：《高电压技术》2021年第6期2214-2222,共9页High Voltage Engineering

基　　金：国家自然科学基金(51977039);福建工程学院科研启动基金(GY-Z18060)。

摘　　要：对于目前电力系统中低频振荡参数辨识中的噪声干扰和精度问题,提出了一种新的提取低频振荡模态参数的方法,将快速独立分量分析技术(fast independent component analysis,Fast ICA)和总体最小二乘-旋转不变技术(total least squares-estimation of signal parameters via rotational invariance technique, TLS-ESPRIT)联合起来。首先运用FastICA技术对含有噪声的电力系统低频振荡广域测量信号进行预处理而达到降噪效果,而后将处理后的信号作为新的输入信号利用TLS-ESPRIT算法进行估计辨识,从而得到各个模态特征参数。通过对理想信号、EPRI-36机系统和电网实测信号仿真验证了所提方法的有效可行性,不但能够有效抑制噪声并准确地辨识低频振荡参数,而且在抗干扰性和提取精度上与传统辨识方法相比来说是有一定优势的。The noise interference and accuracy of low-frequency oscillation parameter identification in the power system are discussed, and a new method for extracting the modal parameters of low frequency oscillation is put forward. The Fast ICA(Fast Independent Component Analysis) is combined with the total least squares-estimation of signal parameters via rotational invariance technique(TLS-ESPRIT). Firstly, the FastICA technology is employed to pre-process the low-frequency oscillation wide-area measurement signal of power system containing noise so as to achieve noise reduction effect.Then, the TLS-ESPRIT algorithm is employed to estimate and identify the filtered signal to obtain each modal parameter. Finally, the validity and feasibility of FastICA-TLS-ESPRIT method are verified by simulation of ideal signal and EPRI-36 machine system and grid measure signal, and it is that this method not only can be adopted to effectively suppress noise and accurately identify low-frequency oscillation parameters, but also has certain advantages in anti-interference and extraction accuracy compared with traditional identification methods.

关 键 词：电力系统 低频振荡 快速独立分量分析 TLS-ESPRIT算法 模态辨识 噪声干扰 

分 类 号：TM712[电气工程—电力系统及自动化]