【作者】 胡文平；

【导师】 尹项根； 张哲；

【作者基本信息】 华中科技大学 ， 电力系统及其自动化， 2005， 博士

【摘要】 随着电力系统规模的不断扩大,大型变电站的数量逐步增加,对变电站电气设备的可靠性及技术水平要求也日益提高,现阶段国家电力公司正大力推广的电力设备状态监测与可靠性维修正如火如荼的在各地展开。电力设备状态监测和故障诊断主要是对当前设备前期的、潜伏性故障通过各种技术手段找出它的故障规律,对这类故障的诊断是目前电力系统研究的热点之一。这与瞬时性的、保护动作之后的故障判断是一个问题的两个方面,继电保护并不能解决隐藏的、潜伏性的前期故障,所以研究电力设备状态监测与故障诊断具有重要意义。电力设备状态监测与故障诊断不仅是设备状态检修模式的基础,也符合变电站综合自动化正在实施的电气运行模式的需要。无论是常规变电站还是无人值守变电站,在其监控系统中,都需要增加一个在线监测和故障诊断专家系统用以作为辅助决策手段,进而提高监控能力。要想实现真正的无人值守,需要加入电气设备在线监测和故障诊断的内容,这样变电站综合自动化才更加完善和更有效。所以在测量、控制、信号、保护和远动等综合自动化的基础上,如能融合电力设备在线监测与故障诊断系统,必将推动变电站综合自动化向前发展,这对提高我国变电站综合自动化水平具有重要意义。本文阐述了智能化技术在电力设备故障诊断中的应用,介绍电力设备故障诊断的发展历史、研究现状以及未来在线监测发展的趋势,提出了对这些问题的研究思路和解决方法,明确了本文研究的主要内容。为了解决电力设备故障诊断中所遇到的主要技术难题,突破常规方法进行故障诊断的局限,论文将不同类型神经网络用于变压器绝缘故障诊断的数学模型和实现原理进行了比较分析,研究了基于BP 神经网络的多种改进学习算法,详细比较分析了不同学习算法对绝缘故障诊断收敛性的影响。针对BP 隐含层结构不确定的缺点,文中对不同的隐含层数目都进行了仿真分析。论文还提出了基于径向基网络、概率神经网络和LVQ 模式分类神经网络用于故障诊断的模型和方法,通过比较分析各种不同类型神经网络的性能和故障诊断的准确率,确定了适用于变压器绝缘故障诊断的神经网络模型。为了解决电力设备故障征兆、故障原因和故障机理之间的复杂关系,提出了一种更多还原

【Abstract】 Along with the scale of electric power system extends continuously and the amount of large transformer substation increases gradually, the electric equipment’s request of reliability and technical level in the large transformer substation also rises increasingly. At present the national power company just generalizes energetically the electric power equipments’condition monitoring and reliability centered maintenance everywhere. The electric power equipments condition monitoring and fault diagnosis is to find fault rule by all kinds of technical method aimed at equipment’s prophase and latency fault. The diagnosis of this kind of fault is one of investigative hot spot in electric power system. This problem and the fault judgment after instantaneous protect action are two aspects of one problem. The relay protection can’t solve hidden and latency prophase fault. So it is important to study the electric power equipments’condition monitoring and fault diagnosis. The electrical equipments’condition monitoring and fault diagnosis is not only the base of condition maintenance pattern but also according with the demand that electric running pattern of unattended operation in transformer substation. It is needful to adding an on-line monitoring and diagnosis expert system as assistant decision-making means no matter what general transformer substation or unattended operation transformer substation. This can advance monitoring and controlling capacity. It is needful to add the electrical equipments’on-line monitoring and fault diagnosis in order to realize unattended operation indeed. In this way transformer substation synthesized automatization will be much more perfect and availability. So it will impel the transformer substation’s synthesized automatization if we can fuse the electrical equipments’on-line monitoring and fault diagnosis based on synthesized automatization such as measure, control, signal, protection and so on. It is important to promote transformer substation’s synthesized automatization level in China. The intelligentized technology applying in electrical equipments’fault diagnosis is expounded in this paper. The paper introduces the development history and present study condition of electrical equipments’fault diagnosis systematically. This paper also introduces the development trend of on-line monitoring in the future. Then studied and solved method has put forward in this paper. The main content to study is definite. The paper has broken the localization of normal fault diagnosis’method in order to solve the main technical problem in the electrical equipments’fault diagnosis. The different type math model and realized theory of transformer’s insulated fault diagnosis applying neural network method are compared. All kinds of study arithmetic based on BP neural network have studied. The influence on network’s constringency of different study arithmetic has compared detailed. Aimed at the shortcoming of BP network-hidden layer structure’s uncertainty, the different hidden layer amounts have simulated in this paper. This paper has put forward fault diagnosis model and method based on radial basis function network, probabilistic neural network and learning vector quantization neural network. By compared the capability and nicety of different type neural network, the appropriate neural network model applying in transformer’s fault diagnosis make sure. The paper has put forward a kind of fuzzy mathematical diagnosis method based on transformer insulated fault diagnosis in order to solve complicated relation of electrical equipments’fault omen and fault reason and fault mechanism. The fuzzy phenomenon subset and subjection degree function corresponding with different fault type are established. Thereby fuzzy synthesized judge of fault type carry on. Author has combined advantage and fuzzy reasoning method. It has overcome the problem of difficult to determine fuzzy regulation in transformer fault diagnosis. Making use of from the orientation nerve network of from study function, pass the study of the nerve network made sure the misty rule to belong to with faintness degree. ANFIS model of transformer fault diagnosis has established. This model has realized the electric power equipment fault diagnosis. It reflects actual running state of transformer. Author has established information fusion model applied in on-line monitoring and fault diagnosis according to information fusion principle. This paper has put forward the method of electric power equipment fault diagnosis applying D-S proof theories for the first time. D-S fusion model and method has fused neural network and fuzzy reasoning diagnosis result again. It makes diagnosed information more definite. It has improved diagnosis accuracy. Author has put forward the fault diagnosis method synthesized many artificial intelligence models such as the nerve network, fuzzy mathematics, adaptive neural fuzzy inference system and information fusion. Then many intelligence information fusion judgments are come into being to diagnose electric power equipment fault. The author has developed the fault diagnosis expert system in transformer substation leading the expert system into the electric power equipments’fault diagnosis. This paperhas described detailed the fault diagnosis expert system’repository and database. Then diagnosed example of breaker and transformer is given in the end. This paper has summarized actuality and development trend. Then the thought of many parameters integer design taken the transformer substation as on-line monitoring object has been put forward in this paper based on the request and object of electrical equipments’on-line monitoring in transformer substation. The concentrative on-line monitoring scheme aimed at large-scale and medium-sized transformer substation is designed. Moreover the structure of synthesized on-line monitoring system that can synthesize many types of electrical equipment such as transformer, breaker, mutual inductance and capacitance device and so on has expounded in this paper. This paper has finished the design of collectivity scheme and hardware circuit. The testing and debugging of hardware circuit has come through. The above research results are summarized finally. The further investigative direction is put forward in the end.更多还原