【作者】 王劲松；

【导师】 沈继红；

【作者基本信息】 哈尔滨工程大学 ， 系统工程， 2010， 博士

【摘要】 近20年来,电力工业通过解除管制、引入竞争,实现提高发电、输电、配电经济效益的目的。随着我国经济建设的深入发展,我国电力工业市场化改革在稳步推进的同时,保证解除管制的电力系统安全可靠的运行至关重要。解除管制的电力系统及电力市场是一个复杂系统,在运行中会受到各种各样的风险的影响,根据复杂系统脆性理论,除了复杂系统所具有的一般特性之外,还具有脆性。本文以解除管制的电力系统作为研究对象,采用复杂系统脆性理论和方法,针对解除管制的电力系统及电力市场的风险进行研究。首先,以复杂系统作为研究对象,研究复杂系统的复杂性及其度量。在所提出的脆性是复杂系统的一个基本特性的理论基础之上,综合现有的复杂系统脆性研究成果,深入研究脆性的定义、特点和模型。其次,对解除管制的电力系统风险及其脆性的研究。在解除管制的电力系统环境下,独立发电商、输电系统运行机构、独立电能零售商可以被认为是一个脆性基元内的三个子系统。并且,按照复杂系统脆性理论,他们之间是非合作博弈的关系。为了保证系统正常运行,他们都需要负熵的注入。例如:风险会引起电力系统熵增,而信息是一种负熵,信息的不完全,将导致他们负熵的缺乏。在系统无外界干扰的情况下,他们彼此之间通过信息交换,可以保证熵增在一定范围内,而保持稳定状态。在一个脆性基元内,当外界干扰作用于一个子系统时,导致此系统的熵迅速增加,若无负熵补充,此系统会因此崩溃,根据复杂系统脆性理论,其他的两个子系统也会崩溃。同时,研究电力信息系统的脆性。再次,基于核独立分量分析方法,分析风险对电力市场影响的估计的研究。在不完全系统参数,仅仅了解部分风险对系统的综合影响的情况下,完成了独立风险对电力市场影响的估计的研究。核独立分量分析KICA (Kernel Independent component analysis) ,即白化的核主分量分析加上独立分量分析算法,提供了一条途径将线性PCA扩展到观察数据的非线性子空间。KICA先通过一个函数将原观察数据空间,或称输入空间映射到一个高维的线性空间,称为特征空间,然后PCA在这个高维的特征空间进行。经过处理过的数据,在白化的空间使数据尽可能呈现非高斯性。最后,通过的仿真实验可以看出,在不知道系统结构的情况下,KICA算法可以通过部分数据估计出独立的电力市场风险影响曲线。更多还原

【Abstract】 In the past twenty years, the electric power industry has realized its aims at improving the economic performance of generating, transmission and distribution of the electricity by means of deregulation and competition. Along with the development of economy construction and the gradual reform of power market in China, it is very important to guarantee the safety and stable operation of the deregulated electric power system.Under the influence of different kinds of risk, deregulated electric power system and power market can be regarded as a complex system, which should has the brittleness besides the other characteristics owned by complex system based on complex system brittle theory. In this dissertation, the deregulated electric power system is considered as the object, and the researches have been carried out according to the risk of the deregulated electric power system applying complex system brittle methods and theories.Firstly, complex system is considered as object, which complexity and its measurement have been studied. Further researches about the definition, characteristics and models of brittleness have been emphasized depending on the integration of latest researches about it on the ground that it is one of basic characteristic of complex system.Secondly, study on the risk and brittleness of deregulated electric power system and its power market.Under the deregulated environment of electric power system, the independent power producer, operation utility of electric transmission system and independent electric retail can be considered as three subsystems in one brittle element. In addition, their relationship between each other should be non cooperative game according to the complex system brittle theory. They all need the injection of negative entropy in order to be in good operation state. However, for example,risk can induce the entropy increase of system, while information is one kind of negative entropy, which incomplete information will result in the deficiency of them. Because they can keep the increase of entropy in the bounded range by means of information exchange among them, they are in stable state. When one subsystem in the brittle element is collapsed owing to the inference outside and its entropy increase, the other two subsystems will face the same collapsed result because of the deficiency of negative entropy. At the same time brittleness of information system of electric power system has been studied.Thirdly, on the basis of the kernel independent component analysis algorithm, the research for estimating the influences of power market brought from risk has been studied. The estimation on the influences of power market brought from single risk has been finished by means of only part of knowledge of system and integrated influences of power market induced by some kinds of risks. Kernel principal component analysis (KPCA), i.e. whitened kernel primary component and independent component analysis algorithm, provides one way to use linear principal component analysis in the expanded nonlinear subspace of the observed data. That is, KPCA maps the original data space to a high dimension space called feature space through a kernel function, which PCA is carried out in this feature space. The distribution of the projected data can be made as non-Gaussian as possible.Finally, results of simulation experiments show that KICA can be used for estimating the individual risk influence profile of power market by the means of part of data of system without the knowledge of structure of system.更多还原