基于相量测量单元的智能电网断路故障定位研究

发布时间：2018-07-10 17:36

本文选题：传输线断路 + 相量测量单元　；参考：《复旦大学》2014年硕士论文

【摘要】：近年来,国内外电力系统发生了多次由连锁故障导致的大规模停电事故。这些由微小扰动引发的电力系统连锁故障会导致电网大而积崩溃的灾难性后果。此类事故的频繁发生,引发了许多关于级联故障的研究。预防级联故障事故的关键之一就在于快速准确地定位初始故障的发生,包括单传输线故障和多传输线故障。因此,在大规模智能电网中,迅速并准确地定位电力线断路故障非常重要。由于相量测量单元(Phasor Measurement Unit, PMU)的广泛使用,使得通过PMU来解决传输线断路故障定位问题成为可能。直接利用PMU检测传输断线断路故障成为一个新兴的研究方向。本论文将断路定位问题转化为图上特定边的寻找问题开展研究,基于断路前与断路后两次由PMU上测量到的实时电压相角信息,通过构建优化模型来完成电网中的断路故障定位。首先,从直流潮流计算原理出发,借助于图理论,把电网等效于一个图模型。在此纂础上以简化的直流潮流计算方程为基础,通过电网拓扑的加权拉普拉斯矩阵,进行系统建模。同时,结合传输线断路故障的稀疏性本质,从稀疏信号处理的角度研究传统的传输线断路故障定位问题,提出了一种利用全局PMU相角信息检测传输线断路故障的算法。其次,针对PMU的成本考虑,目前PMU在电网中还仅限于局部枢纽节点及关键输电断面进行配置。本论文提出只利用PMU的局部观测来定位断路故障的算法,即根据PMU的布置与否,将电网中节点分为可量测的内部系统和不可量测的外部系统,仅通过可量测的内部系统节点的电压相角来估计全网(包括外部系统)断路的传输线断路故障。通过将加权拉普拉斯矩阵进行分块,得到一个仅使用部分观测进行全局线路检测的方程,结合压缩感知的正交匹配追踪算法,进而实现对外部故障线路的实时准确定位。IEEE 118节点模型的仿真结果表明该算法在未提高计算复杂度的前提下,仅使用部分观测的数据就达到了较为满意的估计准确率。最后,考虑到多条断路情况下直接求解引入的高复杂度及低准确率问题,本论文给出了基于随机采样的传输线断路故障定位模型,避免了采用贪婪算法直接求解带来的误差累积。算法从最大化似然概率角度引入概率模型,假设每一条传输线都服从伯努利的先验分布,进而利用随机样本对此概率进行有效性迭代。IEEE 118节点模型的仿真结果表明所提出的方法在不需要多次迭代的前提下,对于多断路情况定位性能有显著提升。
[Abstract]:In recent years, power systems at home and abroad have occurred a number of cascading failures caused by large-scale power outages. These cascading faults of power system caused by small disturbances can lead to the catastrophic result of power grid collapse. The frequent occurrence of such accidents has led to a lot of research on cascading faults. One of the keys to prevent cascading faults is to locate the initial faults quickly and accurately, including single transmission line faults and multiple transmission line failures. Therefore, it is very important to locate power line fault quickly and accurately in large scale smart grid. Due to the wide use of Phasor Measurement Unit (PMU), it is possible to solve the problem of fault location by PMU. It is a new research direction to use PMU directly to detect transmission-break fault. In this paper, the problem of location of open circuit is transformed into the problem of finding specific edges on the graph. Based on the information of the phase angle of real time voltage measured by PMU before and after the break, the optimal model is constructed to locate the fault in the power network. Firstly, based on the principle of DC power flow calculation, the grid is equivalent to a graph model by means of graph theory. Based on the simplified DC power flow calculation equation, the system is modeled by the weighted Laplace matrix of power network topology. At the same time, according to the sparse nature of transmission line fault, the traditional problem of transmission line fault location is studied from the point of view of sparse signal processing, and an algorithm is proposed to detect transmission line fault by using global PMU phase angle information. Secondly, considering the cost of PMU, PMU is only limited to local hub nodes and key transmission sections. In this paper, only local observation of PMU is used to locate open circuit fault. According to the layout of PMU, nodes in power network are divided into measurable internal system and unmeasurable external system. The transmission-line fault of the whole network (including external system) is estimated only by measuring the voltage phase angle of the internal system node. By dividing the weighted Laplace matrix into blocks, an equation for global line detection using only partial observations is obtained, combined with a compressed perceptual orthogonal matching tracking algorithm. The simulation results of the IEEE118-bus model for external fault lines show that the proposed algorithm achieves satisfactory estimation accuracy by using only some of the observed data without increasing the computational complexity. Finally, considering the problem of high complexity and low accuracy caused by direct solution in the case of multiple open circuits, this paper presents a transmission line fault location model based on random sampling. The error accumulation caused by the greedy algorithm is avoided. The algorithm introduces a probabilistic model from the point of view of maximum likelihood probability, assuming that every transmission line follows the prior distribution of Bernoulli. The simulation results of the IEEE118-bus model using random samples show that the proposed method can significantly improve the localization performance of multi-break situations without multiple iterations.
【学位授予单位】：复旦大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TM75

【参考文献】