电力系统恢复中的网络重构与分区恢复策略

发布时间：2018-05-11 18:04

本文选题：电力系统恢复 + 网络重构　；参考：《浙江大学》2017年硕士论文

【摘要】：近几年国内外发生了多起大面积停电事故,如2003年8·14美加大停电、2012年印度大停电、2015年广东湛江电网受台风影响引起的大面积停电事故、2015年土耳其"3.31"大停电事故等为电力系统的安全稳定运行敲响了警钟。一旦发生大面积停电,如果没有事先制定的恢复预案,就会造成重大的经济损失。因此,有必要深入而系统地研究电力系统的事故处理及完全或部分停电后的恢复问题,这对减少故障影响和停电损失,确保事故发生后能安全、快速、智能地恢复供电,具有十分重要的理论和实际意义。大停电后的电力系统恢复可分为黑启动、网络重构和负荷恢复三个阶段。网络重构阶段的主要任务是尽快为失电机组送电并逐步建立起一个稳定的网架结构,为下一阶段全面恢复负荷打下坚实的基础。网络重构阶段的系统恢复策略总体上可分为两类:串行恢复和并行恢复。串行恢复策略在大多数发电机并网前接力恢复各厂站;并行恢复策略将系统分成几个子系统先各自独立恢复,待各子系统恢复完成后再通过并网来实现整个系统的恢复。在此背景下,本文对基于加权网络节点重要度的网络重构策略以及电力系统分区恢复问题展开研究,并取得了一定的研究成果:1、首先针对介数法、节点删除法、节点收缩法等传统方法在评价加权网络节点重要度时存在的不足,提出了节点重要度评价矩阵方法,综合考虑了节点位置和邻接点贡献信息。在此基础上,发展了一种改进的网络重构双层优化模型。上层模型以最大化系统可用发电容量为目标,以非黑启动机组获得启动功率的时间为优化变量;下层模型以恢复路径平均节点重要度最大为目标来确定发电机节点的恢复路径。在下层模型中,通过调节系数改变线路权重中线路电容与操作时间的比重,以避免恢复路径所需时间过长而导致待恢复机组无法尽快恢复的问题。2、提出了基于半监督谱聚类的黑启动分区策略。首先,采用节点间的电气距离这个因素为线路权重赋值,在此基础上建立电力系统无向加权图,进而依据比例割集准则建立黑启动分区策略。该策略由两步构成:第一步建立待恢复机组的分组模型,其以最大化系统发电量和最小化恢复线路总电容为目标,从而保证机组的安全快速启动;第二步以得到的机组分组信息为基础,利用半监督谱聚类算法求解黑启动分区模型。为了克服传统k-means算法对初始聚类中心敏感的缺点,在利用半监督谱聚类算法对特征向量进行聚类的最后一步采用了k-means++算法对特征向量进行聚类。3、提出了黑启动分区中用于网络简化的电流追踪法。首先,针对黑启动分区问题的特点构建电力系统无向加权图,线路权值定义为线路的有功潮流;然后依据多路规范割集准则建立黑启动分区问题的数学模型。之后采用三步方法求解得到黑启动分区结果:1)以最大化系统发电量为目标建立待恢复机组的分组模型,并采用遗传算法求解该模型得到机组分组信息;2)以得到的机组分组信息为基础,采用电流追踪法计算得到支路电流中各电源分量,并以此为基础对电力网络图进行简化;3)利用谱聚类算法求解针对简化后的电力网络图的黑启动分区模型,最后把在简化过程中被合并的节点还原即得到最终的黑启动分区结果。最后对本文研究工作进行了总结,并指出了接下去可以继续研究改进的方面。
[Abstract]:In recent years, there have been a number of large area blackouts at home and abroad, such as the 8. 14 American blackout in 2003, the large blackout in India in 2012, the large area blackout caused by the impact of typhoon in Zhanjiang power grid in Guangdong in 2015, and the "3.31" blackout in Turkey in 2015. If there is no pre planned recovery plan, it will cause significant economic losses. Therefore, it is necessary to systematically study the accident treatment of the power system and the recovery after a complete or partial power outage, which will reduce the impact of the fault and the loss of power outage, and ensure that the power supply can be safely, quickly and intelligently after the accident. The restoration of power system after blackout can be divided into three stages: black start, network reconfiguration and load recovery. The main task of the network reconfiguration stage is to send electricity to the power loss unit as soon as possible and gradually establish a stable network frame structure to lay a solid foundation for the next stage full recovery load. The system recovery strategy in the stage can be divided into two categories: serial recovery and parallel recovery. The serial recovery strategy restores the stations before most generators are connected to the grid. The parallel recovery strategy divides the system into several subsystems, and the recovery of the whole system is realized after the subsystems are restored. In this context, this paper studies the network reconstruction strategy based on the weight of weighted network nodes and the problem of partition restoration in power system, and obtains some research results. 1. Firstly, the shortcomings of the traditional methods, such as the mediate method, the node deletion method, the node contraction method, and so on, are proposed to evaluate the importance of the weighted network nodes. On the basis of this, an improved two-layer optimization model for network reconfiguration is developed. On the basis of this, the upper model aims at maximizing the power generation capacity of the system, and the time of obtaining the starting power from the non black startup unit is the optimization variable; the lower layer model is used to restore the path. In the lower layer model, the proportion of the line capacitance and operation time in the weight of the line is changed in the lower layer model in order to avoid the problem that the recovery path is too long and the recovery unit can not recover as soon as possible, which is based on the semi supervised spectral aggregation. The black start partition strategy of class. First, using the electrical distance among nodes to assign the weight of the line to the weight of the line, the undirected weighted graph of the power system is set up on this basis, and then the black start partition strategy is set up according to the proportional cut set criterion. The strategy is composed of two steps: the first step is to establish the group model of the unit to be restored, which is to maximize the system. In order to overcome the disadvantages of the traditional K-means algorithm for the initial clustering, the semi supervised spectrum clustering algorithm is used to overcome the shortcomings of the traditional K-means algorithm for the initial clustering. In the last step of clustering the eigenvectors, the k-means++ algorithm is used to cluster the feature vectors.3, and the current tracing method for network simplification in the black start partition is proposed. First, the undirected weighted graph of the power system is constructed for the characteristics of the black start partition problem. The mathematical model of the black start partition problem is established by the standard cut set criterion. After that, the results of the black start partition are solved by the three step method: 1) the group model of the unit to be restored is set up to maximize the system power generation, and the genetic algorithm is used to get the information of the unit group, and 2) based on the information of the unit grouping obtained. The current tracking method is used to calculate all the power components in the branch current, and the power network graph is simplified on this basis. 3) the black start partition model for the simplified electric network graph is solved by spectral clustering algorithm. Finally, the final black start partition result is obtained by reducing the joint node in the simplified process. Finally, the result of the black start partition is obtained. This paper summarizes the research work in this paper, and points out the following aspects that can be further studied and improved.

【学位授予单位】：浙江大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TM73;TM711

【相似文献】