【作者】 蒋强；

【导师】 肖建；

【作者基本信息】 西南交通大学 ， 电力电子与电力传动， 2011， 博士

【摘要】 1998年,Watts和Strogatz在((Nature》上发表文章,发现了复杂网络背后的小世界特性,1999年Barabasi和Albert在《Science))发表研究成果,提出节点度分布具有无标度特性。这掀起了研究复杂网络理论的热潮,吸引了物理、数学、工程、社会学、生物学等各领域的学者去探究复杂世界背后隐藏的本质规律,因此成为众多领域研究的热点问题。复杂网络在自然界普遍存在,比如：因特网,交通网,电力网,人际关系网,科研引用网,病毒传播网,航空运输网,国际贸易网,人体生理网,新陈代谢网等等。近年来,复杂网络带来的问题日益突出：SARS病毒在极短的时间流行全球,金融危机波及各国,气候问题越发严重,电力事故不断出现等。为此,人们非常需要去认知复杂网络,了解其背后隐藏的本质规律,便于扬长避短地应用之。复杂网络系统从不稳定到稳定是一个动态的过程,在该过程中存在临界状态,对于处于临界状态的复杂网络系统又该呈现何种特征,如何设计出更加优良的同步稳定控制器,趋利避害的对待复杂网络问题,是当前该理论研究新的问题。随着社会的快速发展,能源紧缺问题日益突出,以石油、煤炭为核心的能源格局必须打破。同时,也带来严重的环境污染问题。发展新能源成为时代主题,也带来新的电网安全问题。2003年,美国电力科学研究院(EPRI)将未来电网定义为intelligentgrid；2005年,欧洲首先提出了"Smart Grid"----智能电网概念,意在解决新形势下人类所面临的能源、环境和电网安全问题。如何真正的从根本上解决电力系统安全问题,被推到了时代浪尖,迫使我们重新梳理和反思。复杂网络理论的研究还存在待完善之处,其应用研究才刚起步,还存在诸多亟待解决的问题。为此,本文结合当前复杂网络理论研究现状和电网建设的特点,采用上述理论,就目前存在问题展开研究工作,主要包括以下几个方面：1.进行复杂系统临界同步机理研究。复杂网络系统由于节点特性和边耦合特性的多样性而更加复杂。对于这种复杂网络系统有两个截然相反的整体形态(稳定和非稳定),对其临界特性的研究尤为重要。本文对其临界同步特性进行了研究,提出了判断一类复杂网络临界态的充分条件和必要条件,通过实验证明了这种临界态的存在及判断判据的正确性。对进一步研究复杂网络系统的临界特性有重要意义。2.复杂网络同步牵制控制理论的研究。牵制控制是实现复杂网络同步最有效的方法,通过对少量节点的控制驱动网络到期望的状态或轨道。本文通过引入节点重要度指数,在复杂网络和实际节点对网络贡献力上架起了桥梁。提出了基于节点重要度指数的牵制策略。同时,将节点重要度指数引入到了控制器的设计中,得到了牵制控制器设计更合理有效的方法,解决了牵制控制器设计中参数确定缺少理论依据的问题。3.基于复杂网络理论的交直流输电系统研究。利用复杂网络理论,深入分析超高压直流输电对电网网络特性的影响。提出了交直流混合输电系统的网络特性评价方法,算例证明直流输电的引入将使电网向无标度特性转移。同时对高压直流输电换流站选址进行研究,算例证明根据度大的节点选择换流站能增加网络的抗毁性。4.提出基于复杂网络牵制控制理论实现电网AGC牵制控制。将复杂网络牵制控制理论用到AGC控制中,将电力输送的潮流约束和传输能力转换成复杂网络的边权,并将边权视为电网节点间的耦合强度,考虑节点的实际容量,结合电网的物理连接得到节点的点权。同时,根据发电节点的点权大小,运用复杂网络牵制控制策略对部分发电节点进行牵制自适应控制,实现了复杂电网中实际有功平衡和各发电机转速同步。5.建立基于模糊支持向量机学习理论的电力系统负荷在线预测模型。机器学习是继专家系统之后人工智能应用的又一重要研究领域,也是人工智能的核心研究课题之一,其显著的特点就是小样本学习,且在学习过程中不存在过学习和局部极小的缺点,有较好的推广性,得到广泛的应用。本文运用模糊集理论对负荷影响因子进行模糊化,并采用最小二乘支持向量机(LS-SVM)学习方法,结合嵌入维思想进行建模和预测,利用粒子群(PSO)算法进行模型参数寻优。同时,为降低运算量,提出小批量更新方法,实现在线学习和准确预测。更多还原

【Abstract】 In1998, Watts and Strogatz in the" Nature" published the article, revealed the small world property behind a complex network. Barabasi and Albert in1999, in "Science", published research results and proposed scale-free properties of distribution of node degree. This set out a wave of research in complex network theory and has attracted scholars of physics, mathematics, engineering, sociology, biology and other fields to explore the nature law of the complex world. So it has become a hot research issue in many areas.Prevalence of complex networks in nature, such as Internet, transportation networks, electricity networks, personal relationships, research and cited networks, virus transmission networks, air transportation networks, international trade networks, physiological networks, metabolic networks and so on. In recent years, problems caused by complex networks have become increasingly prominent, like, SARS virus epidemic in a very short time, the global financial crisis spread to many countries, the more severe climate, and electricity accidents have occurred. Hence, there is a great need for understanding complex networks and if we understand the nature law, it is easy to avoid weaknesses in applications. Complex network system from instability to stability is a dynamic process. In this process there is a critical state, which features should be shown in a critical state for the complex network system, how to design a more excellent and stable synchronization controller, while avoiding disadvantages of the treatment of complex network problems. This is a new hot topic of theoretical research in the present time.With the rapid development of society, the energy shortage problem is becoming increasingly prominent, the oil and coal as the core energy resourse will become short soon. At the same time, it also brings serious environmental pollution problems. Development of new energy has become theme of the times, and it also brings new power grid security issues. In2003, the U.S. Electric Power Research Institute (EPRI) defined the future grid as intelligent grid; in2005, the European first proposed "Smart Grid "----smart grid concept, intended to solve the new situation facing human energy, environment and power grid security. How to really solve the power security issues from the grid physical structure has motivated us to the era of this new research field.Therefore, in this thesis, we have considered the current status of complex network theory study and power grid construction characteristics and the use of this theory to study existing issues, including mainly the following aspects:1. Study critical complex synchronization mechanism. Complex network system will be more complicated for the node characteristics and the diversity of coupling characteristics of edges. For such a complex network system has the overall shape of two diametrically opposite (stable and unstable). In order to better control such a system and use it, it is necessary to study the critical state of synchronization. Experimental results demonstrate the existence of this critical. After extensive research find the sufficient and necessary conditions of a class of complex networks to determine the critical state.2. Study the pinning control theory of synchronization of complex networks. Pinning control is the most effective way to realize complex network synchronization, controlled by a small number of nodes to drive the network to the desired state or path. In here, a bridge is proposed by introducing importance index between the complex networks and the actual contribution of power of nodes. Pinning control strategy has beed proposed based on the node importance index. Meanwhile, by introducing the node importance index into the design of controller, a more reasonable and effective way of designing controller is obtained, to resolve to the issue of the lack of theoretical basis during the design of pinning control parameters.3. Study AC and DC transmission system based on complex network theory. Study properties of complex networks of china AC and DC hybrid transmission system, depth analysis the effect of DC transmission to characteristics of the grid. AC-DC hybrid transmission system is proposed using vulnerability assessment methods. Examples demonstrate the introduction of DC transmission will transfer grid characteristics to the scale-free feature. At the same time HVDC converter station location is also studied. Eamples demonstrate the node of converter station selected according to degree can increase the network survivability.4. Study the Grid AGC pinning control based on Complex Network Theory. Under the requirements of the smart grid, it is more flexible and efficient electric power, in order to real-time tracking changes in network load, and can well predict the situation on the power grid. Transform the transmission constraints and capacity into complex network edge weight, then consider node’s actual capacity, combined with the physical network connection to get the node point weight. Basen on complex networks pinning control strategy add pinning adaptive control on the part of nodes, achieve synchronization in the each generator speed.5. Established the model of power system load online forecasting model based on fuzzy support vector machine learning theory. Machine learning is an important area of research and the core research of artificial intelligence; its notable feature is the small sample study. It is a trade-off between learning ability and the model complexity based on a limited sample of the information and to get the best generalization ability, and does not exist in the learning process over learning and local minimum, be widely used. In this thesis, use of least squares support vector machine (LS-SVM) which is a learning methods for small sample, combined with thinking of embedding dimension, and the use of particle swarm optimization (PSO) algorithm for model parameter optimization, research online prediction method to arrive to real-time online learning and accurate forecasting, which is the basis of power adjusted correctly.更多还原