【作者】 邓慧琼；

【导师】 艾欣；

【作者基本信息】 华北电力大学（北京） ， 电力系统及其自动化， 2007， 博士

【摘要】 电网连锁故障问题是近年来比较引人关注的一个问题。由于大停电事故的发生通常伴随着连锁故障现象,随着最近几年来世界各地大停电事故的频繁发生,针对连锁故障问题的研究其现实的应用背景愈发明晰,围绕如何对电网连锁故障进行建模分析的研究也正在日益增多。由于现代电网的复杂性,这项研究工作显得比较富于挑战性。这主要表现在连锁故障通常是一系列错综复杂事件综合作用的结果,构造完整的描述其过程的解析方程组往往比较困难。论文主要针对电网连锁故障的建模、评价、面向实际应用的预测、基于连锁故障预测的面向实际应用的电网脆弱线路评估等问题进行了研究,并进一步对统计学习理论在连锁故障评估中的应用问题进行了研究。论文首先通过对历史上已经发生的大停电事故的分析,总结了电网连锁故障发生的一般表现形式,讨论了当前解释电网连锁故障发生机理的自组织临界理论,大停电发生的幂率行为及其1/ f噪声特性。然后从预测分析的需要和特点出发建立了电网连锁故障的概率分析模型。采用概率分析模型可以充分考虑预测当中存在的不确定因素,并对其可能性大小进行量化,进而可方便地利用量化的结果对不同连锁故障路径的发生可能性进行比较分析。论文首次提出了反映电网连锁故障发生可能性、蔓延速度的评价指标,进而相应提出了利用评价指标对不同的连锁故障路径进行综合评价和聚类划分的方法。所提出的评价指标体系和合理的连锁故障模拟终止原则相结合,有利于在节省计算资源的前提下,反映不同连锁故障路径的发生可能性和蔓延速度等主要特征,并有利于在预测时及早发现那些引发连锁故障可能性高、蔓延速度快的关键扰动事件及电网的脆弱位置。论文提出了一种基于概率优先排序并结合电网状态监测的电网连锁故障路径搜索策略。由于电网中各种随机因素的组合数目巨大,对所有的连锁故障路径进行搜索其往往不切实际。采用概率优先排序的搜索策略并从现实的可能性出发,可尽量保证搜索得到的连锁故障路径是可能发生的,进而可为实际运行提供参考。在此基础上论文提出了基于连锁故障分析的电网脆弱线路评估方法和预警策略。最后,论文提出了一种基于统计学习理论的电力系统连锁故障评估方法,该方法的出发点是利用实际当中的数据资源并结合统计学习理论中具有良好推广能力的支持向量机技术,为快速评估以及模拟仿真提供一种有益的补充工具,具有一定的实际应用价值。更多还原

【Abstract】 Large-scale blackouts of electrical power systems have been concerned around the world in recent years. Most of the blackouts are caused by cascading outage, the study of which has been a popular issue nowadays. Due to increasing complexity of modern electric network, the work appears to be more challenging. This is reflected by the facts that the outages are usually linked to a complex combination of events, constructing analytical equations of the process and is often difficult. This paper mainly study on the model building, the evaluation, the forecasting for practical application and the vulnerability assessment based on the cascading outages forecasting. Further, the author studies on the statistical learning theory applied in the cascading outages assessment.The paper discusses some important issues about the self-organized criticality of the large-scale blackouts in power system in detail, including the characteristics of power laws distribution, the fractal and 1/f noise. And then the author makes the probability analysis model considering the needs and features of the predictive analysis. As from the perspective of the forecast, the factors, such as the location of the initial failure, the reclosing etc., are not certainty. To analysis in the probability way, firstly, we can quantify the probability of the above factors to quantify the probability of the chain fault path searched in the forecasting; secondly the result of the quantification is used to compare the probability of different chain fault paths.In this paper, the evaluation index of the probability, the possibility and the speed of the cascading outage is presented and the method based on fuzzy clustering and integration evaluation on the cascading outage is proposed. The evaluation index will be used in the forecast to find the cascading outage path which is most likely and the most quickly overspread, and then we can find the vulnerability of the network earlier.This paper proposes a search strategy of cascading outage path, based on probability prioritization and by combining network monitor. As the number of random factors in network is tremendous , it is hard to search the entire cascading outage path. Using probability prioritization strategy and starting from reality and possibility, the search can ensure that the pathes are some chains that may happen, and then the searched pathes may be useful for the actual operation .On this basis, this paper brings forward a way to analyze the vulnerability lines in network and warning strategy.Finally, the paper presents a power system cascading outage assessment methods based on statistical learning theory, the starting point of this method is the actual use ofdata resources and combine statistical learning theory with the Support Vector Machine (SVM) technology which can be easier to promote, it also provides a useful additional tool for rapid assessment and simulation and can be used in practical.更多还原