【作者】 严宇；

【导师】 刘天琪；

【作者基本信息】 四川大学 ， 电力系统及其自动化， 2004， 硕士

【摘要】 随着电力系统规模不断扩大，互连部分越来越多，如何保证电力系统的安全运行，以及在安全的情况下更经济的运行显得越来越重要。因此电力系统的安全分析，特别是动态安全分析(dynamic security analysis，DSA)越来越成为研究的重点和热点。 电力系统动态安全分析能够保证实时电力系统的安全运行，保证电力系统的等式和不等式条件的满足，是属于调度自动化的范畴。由于长期以来，安全分析和暂态稳定分别从调度自动化和电力系统的稳定的角度来研究和分析电力系统的等式和不等式条件的满足，两者逐步的综合，使得两者的概念较为混淆。笔者认为，电力系统动态安全分析广义上分为暂态稳定分析(TSA)，电压稳定分析(VSA)，频率稳定分析(FSA)，前者主要指功率的平衡，后两者分别是指电压和频率的稳定。目前较为成熟的是暂态稳定分析(TSA)。因而从某种意义上，现在的动态安全分析的概念含盖更广。本文主要介绍的动态安全分析的方法及相关概念均是针对TSA而言。 本文提出一套基于动态安全域的ANN动态安全分析理论，其中包括基于BP网络和模糊理论的超平面拟合法，用于对运行点类型分类的两类、三类分类器和用于稳定裕度计算的稳定裕度拟合器的ANN动态安全域方法，以及针对系统多故障的自适应ANN方法，特别是提出了一种针对大电力系统注入元素优选的方法。该优选方法将决策表最优特征子集理论和粗糙集元素离散化理论相结合，从大维数的注入元中优选特征元作为ANN注入元素，以降低其注入元素维数，较好的解决了ANN动态安全分析法用于大系统所面临的“维灾难”问题。通过多个系统的仿真和测试，取得很好的效果。本论文主要包括以下内容: 第一章、第二章主要介绍了动态安全分析的现状，特别针对用动态安全域理论进行电力系统动态安全分析的现状，存在问题以及急待解决的问题进行了分析和研究。 第三章针对传统超平面实用动态安全域方法拟合度不高，提出了一种基于BP网络和模糊理论的超平面拟合法，以BP网络代替传统的最小二乘法拟合实用动态安全域。此方法在一定程度上减少了临界点和边界附近的稳定点和非稳定点的误判。 第四章提出了一种较为完善且实用的动态安全分析的方法—ANN动态安全域方法，其中包括用于对运行点类型分类的两类、三类分类器和用于稳定裕度计算的稳定裕度拟合器。此方法在很大程度上弥补了超平面拟合法的不足，例如减少由超平面理论本身局限性所造成的误判，以及无法精确的确定运行点的稳定裕度。 第五章针对ANN动态安全域方法用于大系统所面临的“维灾难”问题，提出了一种注入元素优选的方法。此方法大大提高了ANN动态安全域方法的实用性，特别是针对大电力系统。 第六章提出了一种针对系统多故障的方法—自适应ANN方法，此方法弥补了ANN动态安全域方法仅能对系统一既定故障建构一神经网络的缺陷，从而大大减少了对系统进行动态安全分析的神经网络的个数。 第七章则介绍、分析了动态安全域理论的研究前景。关键词:电力系统;动态安全分析;动态安全域;神经网络;模糊理论;更多还原

【Abstract】 With the expansion of power system and increasing of interrelated parts, it is gradually important to how to confirm secure operation of power system, and economical operation under secure condition. So the security analysis of power system, especially dynamic security Analysis(DSR), becomes the research’s emphasis and focus increasingly.The dynamic security analysis can confirm real-time secure operation of power system, and meet equation and inequality, which belongs to dispatch automation’s area. During the long period of time, the security analysis and transient stability, which are synthesized gradually, aim to equation and inequality from dispatch automation and power system stability respectively for reasearch, so the concepts of two parts are confused. I think, in broad sense, the dynamic security analysis of power system is divided into transient stability analysis (TSA), voltage stability analysis(VSA),frequency stability analysis (FSA). The first one aims to power balance, and the other two to voltage and frequency stability. Nowadays the transient stability analysis(TSA) has been mature relatively. So in some sense, the dynamic security analysis covers more broadly. The methods and theories that are introduced in this paper aim to TSA.The paper proposes a ANN dynamic security analysis theory based on dynamic security regions, which includes hyper-plane imitation based on BP network and fuzzy theory, ANN dynamic security regions methods composed by two-kind orthree-kind classifier for running points classification and stability margins imitator for stability margins computation, adaptive ANN method for multi-fault of power system, and especially a optimization for injections of bulk power system. The optimization based on optimal feature subset of decision tables and element discretization of rough-set is used to select feature elements, which is regarded as ANN injection features, from the lager dimensions injections. The ’dimensions misfortune’ problem caused by application of ANN dynamic security regions method to bulk power system is solved fairly because of injection features reduction. The validity of the optimization is proved by simulation on power systems. The main content of paper is introduced as follows.In chapter one and two, the present conditions on dynamic security analysis are introduced, especially the present conditions, existing and resolved problems of power system dynamic security analysis based on dynamic security regions theory are analyzed and researched.In chapter three, the imitation based on BP network and fuzzy method, which replaces the traditional least square method to imitate the practical dynamic security regions, is presented to the relatively low accuracy of traditional hyper-plane method.In chapter four, the ANN dynamic security regions method is proposed including two-kind or three-kind classifier for running points classification and stability margins imitator for stability margins computation. The method overcomes the deficiency of hyper-plane imitation extensively, such as reduction of misjudgments because of hyper-plane theory itself limitation, and difficulty to decide accurately security margins of running points.In chapter five, the optimization of injection is raised to avoid ’dimensionsmisfortune’ in bulk system. The method enhances the validity of ANN dynamic security analysis largely.In chapter six, the adaptive ANN method is put forwards for multi-fault of power system. The method overcomes the deficiency of ANN dynamic security-regions method, which can only build a network to a fault. So the quantities of ANN are reduced largely.In chapter seven, the research prospect on dynamic security analysis theory is introduced and analyzed.更多还原