节点文献

基于广域量测的电力系统区域间低频振荡分析

Analysis of Interarea Electromechanical Oscillations in the Power Systems Based on Wams Data

【作者】 张鹏

【导师】 王晓茹;

【作者基本信息】 西南交通大学 , 电力系统及其自动化, 2015, 博士

【摘要】 区域间低频振荡问题是互联电力系统中存在的固有现象,弱阻尼振荡模式导致的持续振荡现象严重制约了联络线功率输送能力,不利于系统经济运行;而负阻尼振荡模式所引发的增幅振荡现象将危害系统稳定性,若发现不及时或控制不恰当,严重时甚至导致大停电事故。鉴于低频振荡问题危害的严重性,国际电子与电气工程师协会(IEEE)和国际大电网组织(CIGRE)均曾成立专门工作组对其进行研究。传统的电力系统低频振荡问题分析方法需要对系统建立详细的数学模型,将模型在某个运行点线性化后基于小信号稳定性理论进行分析。对于大规模复杂互联电网,对其准确建模本身就存在一定困难,而且由于系统运行条件多变,若模型不能及时得到更新,则不能准确反映系统的实际低频振荡情形。PMU(同步相量测量单元)在电力系统中的大量安装使得基于广域量测数据的低频振荡分析成为可能,由于实测数据真实体现了系统当前的运行状态,因此基于量测的低频振荡分析方法可有效弥补基于模型的分析方法的不足,具有广阔的应用前景。本论文基于广域量测数据分析电力系统低频振荡问题。将结构学领域用于分析建筑或机械结构振动特性的ITD (Ibrahim Time Domain)方法引入电力系统,基于大扰动后的自由振荡信号,实现对低频振荡模式参数的辨识。基于纽约/新英格兰16机系统模型对所提方法的性能进行了验证和评估。通过采用零相位偏移低通滤波器对输入信号进行预处理,有效提高了所提方法对量测噪声的鲁棒性。所采用的ITD方法基于多输入通道,与已有文献中基于单输入通道的模式辨识方法(比如Prony方法,TLS-ESPRIT方法,小波方法,以及卡尔曼滤波方法)相比,ITD方法除了能辨识振荡模式的频率和阻尼比外,还能同时辨识振荡模态,这三个参数可以完整地描述某个低频振荡模式的特征。与已有文献所采用的基于多输入通道的随机子空间方法(SSI)的对比结果表明,ITD方法与SSI方法在辨识精度方面性能相当,但由于SSI方法需要对规模庞大的矩阵进行奇异值分解,因此本文所采用的ITD方法在计算速度方面优势明显。另外,当只有少量PMU量测点数据时,只要这有限个PMU安装点的观测量对某个振荡模式具有较高的可观性,则ITD方法可利用该有限个PMU安装点的量测数据准确辨识出该振荡模式。基于中国西南某实际电网PMU实测数据,采用ITD方法辨识出该电网与其所属的大区电网之间存在的约0.32Hz的振荡模式。ITD方法本质上是基于自由振荡信号的方法,而自由振荡信号的获得依赖于大扰动的发生,因此基于自由振荡信号进行区域间低频振荡模式辨识存在一定的局限性。系统在大部分时间下处于环境激励状态,此时系统的响应是由系统中负荷的随机波动所引起的,为随机响应信号。在一定条件下,随机减量技术(RDT)和自然激励技术(NExT)可以从随机响应信号中提取其自由振荡特征,通过将这两种技术与ITD方法相结合,实现了环境激励条件下基于RDT-ITD和NExT-ITD的低频振荡模式辨识。16机系统仿真结果表明,与已有文献中的RDT-Prony方法相比,RDT-ITD方法和NExT-ITD方法对量测噪声具有更好的鲁棒性。中国西南某实际电网PMU实测数据验证结果表明,RDT-ITD方法和NExT-ITD方法均能准确辨识出该电网内部存在的约0.75-0.8Hz的低频振荡模式。已经有文献基于NExT-ERA方法在环境激励条件下进行电力系统低频振荡模式辨识,但仅仅讨论了该方法在频率和阻尼比辨识方面的性能。本文对NExT-ERA在振荡模态辨识方面的性能进行了验证评估。同时,提出基于RDT-ERA的低频振荡模式辨识方法,并基于同一仿真模型,即16机系统模型,对RDT-ERA, NExT-ERA, RDT-ITD与NExT-ITD的性能进行对比。中国西南某实际电网PMU实测数据验证结果表明,RDT-ERA方法和NExT-ERA方法均能准确辨识出该电网内部存在的约0.75-0.8Hz的低频振荡模式。在环境激励条件下进行低频振荡模式辨识时会受到两种随机因素影响,其一,环境激励具有随机特性;其二,量测噪声也具有随机特性。这两种随机因素导致单次的辨识结果具有偶然性,不具说服力。因此,本文基于蒙特卡洛仿真实验,从概率统计的角度研究了这两种随机因素对辨识结果的影响。同时,本文在设计蒙特卡洛仿真实验研究某种随机因素的影响时,均考虑到将另外一种随机因素的影响排除在外,从而避免了两种随机因素影响的混叠。

【Abstract】 Interarea electromechanically oscillations are inherent to interconnected power systems. Oscillation modes with low damping ratios may cause sustained oscillations, which will set limits to the power transferring capability of the tie lines and go against the economic operation of the power systems; while oscillation modes with negative damping ratios may cause oscillations with growing amplitudes, which may pose threat to the stability of the power system; those growing oscillations, if not observed in time or controlled properly, may even cause catastrophic blackout events. Considering the potential serious threat of the low frequency oscillations to the power systems, both the IEEE and the CIGRE have special publications focusing on this subject.The traditional way to study the interarea oscillations is to linearize the detailed state space model of the power system around an operating point and conduct eigen-analysis of the system state matrix. The model based methods are constrained by the following facts:for large interconnected power systems, it is not an easy job to obtain its detailed model accurately; even worse, the operation condition of the system is always changing, and if the model of the system is not updated in time, it cannot reflect the oscillation characteristics of the system correctly. The growing installation of phasor measurement units (PMUs) in the power systems makes it possible to analyze the interarea oscillations based on wide area measurements. Since the PMU measurements correctly reflect the current operation condition of the power systems, methods based on measurements can effectively complement that based on detailed system model. The main job of this dissertation is to analyze the interarea electromechanical oscillations based on wide area measurements.The ITD (Ibrahim Time Domain) method, which was formerly applied to analyze the vibration characteristics of the architecture or mechanism structures, was introduced to estimate the interarea modes in the power systems using free response signals. The performance of the ITD method was verified and evaluated in the 16-machine power system model, which is the reduced order model of the New York/New England power system. By passing the input signals through a zero-phase low-pass filter, the robustness of the ITD method to measurement noise can be effectively improved. The ITD method is based on multi-channel input signals, and has the ability to estimate the frequencies, damping ratios as well as the mode shapes simultaneously, and these three parameters can fully describe the characteristics of a specific mode; while methods based on single-channel input signal (such as the Prony method, the TLS-ESPERIT method, the wavelet method, and the Kalman filtering technique) can only estimate the frequencies and damping ratios. Comparison results between the ITD method and the SSI (Stochastic Subspace Identification) method show that both methods are able to estimate the interarea modes accurately, however, the ITD method consumes much less cup time than the SSI method does, because the SSI method has to perform singular value decomposition of a large matrix, which is time consuming. In the situation when only limited PMU measurements are available, as long as these available measurements have good observability for the interested modes, the ITD method can still give acceptable estimation results for these modes. The ITD method was verified using PMU data obtained from a real power grid in Southwest West China, and identified the 0.32 Hz oscillation mode between this power grid and another larger power grid.Theoretically, the ITD method is based on free responses, and its application relies on the occurrence of major disturbances in the power systems. However, for most of the time the system is operating under ambient conditions, during which the responses of the system are caused by random fluctuations of the loads, which are called random responses. Under certain assumptions, the the random decrement technique (RDT) and the natural excitation technique (NExT) are able to extract the ring-down signatures from the random responses. By combining these two techniques with the ITD method, the RDT-ITD method and the NExT-ITD method are employed to estimate the interarea modes during ambient operation conditions. Simulation results in the 16-machine power system model show that, compared with the RDT-Prony method proposed in the former literature, the RDT-ITD method and the NExT-ITD method are much more robust to measurement noise. Estimation results using real PMU measurements obtained from a real power grid in Southwest West China show that the RDT-ITD method and the NExT-ITD method are able to identify the 0.75-0.8 Hz mode existed inside this power grid.The NExT-ERA method was formerly used in the literature to estimate the interarea modes during ambient operation of the power systems, however, the former literature only focused on the estimation of the frequencies and damping ratios. In this dissertation, the ability of the NExT-ERA method in estimating the mode shapes were verified and evaluated. Meanwhile, the RDT-ERA method was proposed to estimate the interarea modes. The performance of RDT-ERA、NExT-ERA、RDT-ITD and NExT-ITD were compared in a common simulation model, which is the 16-machine power system model. Estimation results using real PMU measurements obtained from a real power grid in Southwest West China show that both the RDT-ERA method and the NExT-ERA method are able to identify the 0.75-0.8 Hz mode existed inside this power grid.Both the ambient excitations and the measurement noise are stochastic in nature, and may bring uncertainties to the estimation results during ambient operation conditions. The estimation result from one time of experiment is not convictive. Therefore, in this dissertation, the effects of these two kinds of stochastic factors are analyzed in a statistical way by conducting Monte Carlo simulations. The Monte Carlo simulations are designed in such a way that it can distinguish whether the uncertainties in the estimation results are caused by the ambient excitations or by the measurement noise. So, when studying the effect of the ambient excitations, the measurement noise is fixed for every time of experiment, and vice versa.

节点文献中: 

本文链接的文献网络图示:

本文的引文网络