基于关联规则的电力系统暂态稳定评估方法研究

发布时间：2018-11-17 13:36

【摘要】：电力系统暂态稳定评估是保障电力系统安全稳定运行的重要问题之一。传统的时域仿真法与直接法均遇到了各自难以克服的瓶颈问题。随着电力系统网络规模的逐渐扩大,计算量的迅速增加导致时域仿真法的计算速度难以满足在线监测和控制的需求。而直接法难以适用于复杂模型的问题限制了其在电力系统中的应用。近年来随着云计算、大数据等信息技术的快速发展以及在电力行业的广泛应用,基于机器学习技术的电力系统暂态稳定评估方法为在线稳定评估提供了一种新的方向,并逐渐取得了较大的进展。该类方法具有泛化能力强、评估速度快以及能够挖掘关键运行信息等优势,在电力系统在线暂态稳定评估领域中有着广阔的发展前景。本文将机器学习技术中的关联规则算法引入到电力系统暂态稳定评估研究中。在深入总结分析前人工作的基础上,从电网运行数据中挖掘可靠和易于理解的稳定判别规则,为实现大电网运行中的智能决策提供支持。首先,完成稳定信息数据库的建立工作。一方面,开发PSD-BPA软件的自动仿真计算接口程序,对新英格兰10机39节点系统不同运行方式下进行暂态稳定仿真计算,生成海量仿真样本集。另一方面,整理在线安全分析系统中某跨地区电网某月的历史数据,作为另外一份以待研究的数据样本集。其次,提出了一种基于加权随机森林与递归特征消除策略相结合的特征选择方法,找出影响稳定水平的关键特征和因素,去除冗余输入特征,提高关联规则计算的效率和规则的可解释性。然后,由于绝大部分系统状态变量为连续量,而离散化连续特征量是关联规则必不可少的数据预处理步骤之一。本章在总结Chi Merge离散化算法不足之处的基础上对该算法予以改进,并利用改进算法将连续数据离散化到各个离散区间中。最后,在前几章的基础上,将基于FP-Growth算法实现关联规则分析算法应用于电力系统暂态稳定评估中,初步完成暂态稳定评估规则库的构建工作,并分析了部分规则的作用和意义。
[Abstract]:Transient stability assessment of power system is one of the important problems to ensure the safe and stable operation of power system. The traditional time domain simulation method and direct method both meet the bottleneck problem which is difficult to overcome. With the expansion of power system network, the calculation speed of time-domain simulation method is difficult to meet the need of on-line monitoring and control. However, the application of direct method in power system is limited by the problem that it is difficult to be applied to complex model. In recent years, with the rapid development of cloud computing, big data and other information technology, as well as the wide application in the power industry, the power system transient stability assessment method based on machine learning technology provides a new direction for on-line stability assessment. And gradually made great progress. This kind of method has the advantages of strong generalization ability, fast evaluation speed and the ability to mine key operation information, so it has a broad development prospect in the field of on-line transient stability assessment of power systems. In this paper, the association rule algorithm of machine learning technology is introduced into the study of power system transient stability evaluation. On the basis of summing up and analyzing the previous work, the stable discriminant rules which are reliable and easy to understand are mined from the operation data of power grid, which provides support for intelligent decision making in the operation of large power grid. First of all, the establishment of stable information database is completed. On the one hand, the automatic simulation and calculation interface program of PSD-BPA software is developed to simulate transient stability of New England 10-machine 39-bus system under different operation modes, and a mass of simulation samples are generated. On the other hand, the historical data of a cross-region power grid in the online security analysis system are collated as another sample set of data to be studied. Secondly, a feature selection method based on the combination of weighted stochastic forest and recursive feature elimination strategy is proposed to find out the key features and factors that affect the stability level, and to remove redundant input features. Improve the efficiency of computing association rules and the interpretability of rules. Then, because most of the system state variables are continuous, discrete continuous feature is one of the necessary data preprocessing steps for association rules. On the basis of summarizing the shortcomings of Chi Merge discretization algorithm, this chapter improves the algorithm and uses the improved algorithm to discretize the continuous data into each discrete interval. Finally, on the basis of the previous chapters, the association rule analysis algorithm based on FP-Growth algorithm is applied to power system transient stability evaluation, and the construction of transient stability assessment rule base is preliminarily completed. The function and significance of some rules are analyzed.
【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

【参考文献】