基于Spark的电网扰动影响域识别研究

发布时间：2018-05-31 17:15

本文选题：并行框架(Spark) + 扰动域　；参考：《华北电力大学(北京)》2017年硕士论文

【摘要】：随着电网互联规模及耦合强度的日益扩大,电网运行环境日益复杂,亟需依托大数据技术,提升电网多源大数据的挖掘深度及应用效率。与此同时,局部扰动更易波及较大的区域,有效识别扰动影响域对于抑制扰动传播具有一定的工程应用价值。针对大电网广域时空序列数据的海量存储、高效处理,提出以Spark为核心的电力大数据平台设计框架。论文中指出了Spark在分布式计算中的优势及电力大数据平台建设目标,并对平台各个层次进行详细的论述,阐述了电网时空序列数据处理过程。在搭建的Spark和Hadoop实验环境基础上,对典型聚类算法进行性能对比测试,验证了Spark相对于Hadoop的Map Reduce计算模型数据处理的优势。从电网结构脆弱性和运行状态脆弱性角度,提出势能强度指标衡量电网发生扰动后各个节点受影响程度的强弱。通过变量相关性论述了单一变量识别扰动影响域的约束性,并引入常用结构脆弱性指标电气介数。基于能量函数构造方法,构建节点势能函数方程,将电气介数作为节点势能(运行状态脆弱性)权重提出势能强度指标。通过对IEEE39节点系统不同故障对比仿真分析,得出高电气介数节点更易成为势能传播路径,验证了所提势能强度指标的正确性。以IEEE39节点系统仿真数据为数据源模拟流式数据,基于Spark Streaming组件进行在线扰动影响域计算分析。根据基尼系数理论,提出势能强度基尼系数方法,评估电网故障后网络整体受扰动情况。在线计算势能强度值以及势能强度基尼系数值,以势能强度值作为聚类对象,采用流式K-Means聚类算法进行扰动影响域识别。在未采取任何抑制扰动传播措施情况下,动态地分析了电网发生故障后扰动影响域及势能强度基尼系数的演变情况,综合分析后指出扰动传播至一定程度后,总是趋于引起局部区域稳定性下降。
[Abstract]:With the expansion of interconnection scale and coupling intensity, the operation environment of power grid is becoming more and more complex. It is urgent to improve the mining depth and application efficiency of multi-source big data based on big data technology. At the same time, the local disturbance is more easy to affect the larger region, and it has some engineering application value to effectively identify the disturbance influence region to suppress the disturbance propagation. Aiming at the mass storage and efficient processing of large area space-time sequence data in large power grid, a design framework of electric power big data platform based on Spark is proposed. This paper points out the advantages of Spark in distributed computing and the construction goal of electric power big data platform, and discusses in detail all levels of the platform, and expounds the process of data processing in time and space series of power grid. Based on the experimental environment of Spark and Hadoop, the performance of the typical clustering algorithm is compared and tested, and the advantages of Spark compared with the Map Reduce computing model of Hadoop are verified. From the point of view of structural fragility and operational state vulnerability, a potential energy intensity index is proposed to measure the degree of influence of each node after disturbance. In this paper, the constraint of single variable identification of disturbance influence region is discussed by variable correlation, and the electrical quotient of structural vulnerability index is introduced. Based on the energy function construction method, the node potential energy function equation is constructed, and the electrical medium is taken as the weight of the node potential energy (running state fragility) to put forward the potential energy intensity index. By comparing and analyzing the different faults of the IEEE39 node system, it is concluded that the high electrical intermediate node is more likely to become the potential energy transmission path, and the correctness of the proposed potential energy intensity index is verified. The IEEE39 node system simulation data is used as the data source to simulate the flow data, and the on-line disturbance influence domain is calculated and analyzed based on the Spark Streaming component. According to the Gini coefficient theory, a Gini coefficient method of potential energy intensity is proposed to evaluate the disturbance of the whole network after power network failure. The potential energy intensity and the Gini coefficient of potential energy intensity are calculated on line, and the potential energy intensity is used as the clustering object. Flow K-Means clustering algorithm is used to identify the disturbance influence domain. In the absence of any measures to restrain disturbance propagation, the evolution of disturbance influence region and Gini coefficient of potential energy intensity after power network failure are dynamically analyzed. After comprehensive analysis, it is pointed out that the disturbance propagates to a certain extent. The stability of local region tends to decrease.
【学位授予单位】：华北电力大学(北京)
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TM712;TM732

【参考文献】