【作者】 林伟斌；

【导师】 季天瑶； 陈晓东；

【作者基本信息】 华南理工大学 ， 电气工程（专业学位）， 2020， 硕士

【摘要】 近年来,我国电网规模不断扩大,系统间互联程度不断提高,大量的高倍率快速励磁装置投入到电网,导致低频振荡现象日益严重。低频振荡的发生极大的危害了电力系统稳定性,同时导致系统输送容量无法最大化,因此,监测和分析低频振荡模态参数对系统的安全稳定运行具有重要意义。目前低频振荡的模态参数分析方法多是基于暂态振荡信号进行参数辨识,但暂态振荡信号发生概率低,无法实时反映系统状态,因此该类方法无法辨识系统正常运行时的模态参数。相反,系统正常运行时负荷波动引起的类噪声信号易于获取,且包含系统当前模态信息。因此,本文着重研究基于类噪声信号的低频振荡模态参数辨识方法。本文介绍了电力系统低频振荡的基本理论并分析了从类噪声信号进行低频振荡参数模态辨识的可行性。随后介绍了随机子空间识别算法及盲源分离的基本原理,并对两种方法用于类噪声信号的低频振荡参数辨识的适用性进行了分析,为后续研究打下基础。针对传统随机子空间识别算法定阶困难,存在虚假模态的问题,提出一种双协方差随机子空间识别算法。通过构造两个不同维度的汉克尔矩阵剔除虚假模态,然后通过系统聚类算法进行物理模态参数提取。改进后的双协方差随机子空间识别算法能够实现物理模态自动拾取、自动定阶。实验结果表明,所提方法能够从类噪声信号中实现对低频振荡模态参数的有效精确辨识,并具有良好的抗噪性能,识别结果中不会出现虚假模态。考虑到模态辨识受到激励的随机性、测量误差、数据窗长等因素的影响,估计的模态参数不可避免存在误差,本文提出一种新的考虑不确定度的低频振荡模态参数辨识方法。利用二阶统计量盲源分离技术将含有多模态的类噪声信号分解为多个单模态信号,采用随机减量技术从单模态信号中提取自由振荡信号进行模态参数计算,并引入bootstrap计算置信区间,以此衡量模态参数的不确定度。实验结果表明,所提方法能够有效辨识低频振荡模态参数,同时能够提供模态参数估计结果的置信区间。更多还原

【Abstract】 In recent years,the scale of Chinese power grid has been continuously expanded,and the degree of interconnection between systems has been continuously improved.A large number of high-magnification fast excitation devices have been put into the power grid,resulting in increasingly low frequency oscillations(LFO).The occurrence of LFO greatly harms the stability of the power system,and at the same time,the transmission capacity of the system cannot be maximized.Therefore,monitoring and analyzing the modal parameters of LFO are of great significance to the safe and stable operation of the grid.At present,most of the modal parameter analysis methods for LFO are based on transient oscillation signals.However,transient oscillation signals have a low probability of occurrence and are difficult to obtain in real time.In contrast,ambient signals caused by load fluctuations during normal operation are easy to obtain and contain the current modal information of the system.Therefore,this paper focuses on the identification of LFO modal parameters based on ambient signals.This paper first introduces the basic theory of LFO in power system and analyses the feasibility of modal parameter identification of LFO via ambient signals.Subsequently,the basic principles of the stochastic subspace identification(SSI)algorithm and blind source separation are introduced,and the applicability of the two methods for the identification of LFO modal parameters is analyzed,laying the foundation for the following.Aiming at the difficulty of order determination and the existence of false modes in the traditional SSI,a double covariance SSI algorithm is proposed.The false modes are removed by constructing Hankel matrices of two different dimensions,and then the physical modal parameters are extracted by hierarchical clustering algorithm.The improved double covariance SSI algorithm can realize automatic picking and automatic ordering of physical modes,and can effectively and accurately identify LFO modal parameters from ambient signals.Results show that the proposed method can achieve effective and accurate identification of LFO modal parameters from ambient signals,and has good anti-noise performance.Considering that the modal identification is affected by factors such as randomness of excitation,measurement error,data window length,etc.,the estimated modal parameters inevitably have errors.This paper proposes a new method for identifying LFO modal parameters considering uncertainty.The blind source separation technique is used to decompose the multi-mode ambient signal into multiple single-mode signals.Random decrement technology is used to extract the free decay signals from the single-mode signals for modal parameter calculation,and bootstrap is introduced to calculate the confidence interval,so as to measure the uncertainty of modal parameters.The results show that the proposed method can effectively identify modal parameters of LFO,and can provide confidence intervals of modal parameters.更多还原