【作者】 周政新；

【导师】 耿兆丰；

【作者基本信息】 东华大学 ， 控制理论与控制工程， 2011， 博士

【摘要】 大型发电机是电站的主要设备之一,其系统运行的稳定性和安全性将直接影响电网的运行质量,而发电机绝缘故障在发电机故障中占有较高的比例,在大电机维护方式由定时离线检修向在线状态维修发展的过程中,采用在线监测和故障诊断技术实现大电机绕组的绝缘性能检测具有十分重要的现实意义。发电机绕组在运行过程中受到电、热、机械、化学等多种因素的联合作用下,其绝缘性能会发生变化,在绝缘性能逐渐老化的同时,电气性能也会发生一系列相应的改变,最终导致发生绝缘事故。因此采用在线监测的方式实现对发电机绝缘状况的长期监测,分析发电机绕组的绝缘性能参数在运行过程中的变化情况,是评估发电机特别是大型发电机系统绝缘性能非常有效的检测方法。在大型发电机绕组绝缘结构和绝缘老化机理的研究方面,分析了大型发电机绕组在运行过程的热老化、机械老化和电老化的发生过程；通过对大型发电机绕组绝缘材料性能数学模型的分析,讨论了单因子电老化、热老化、机械老化和多因子老化的特点；重点分析了老化参数对绝缘性能的影响。在此基础上又对绕组绝缘性能的测试方法进行了比较、分析和研究。在总结国内外大型发电机绝缘性能在线测试方法的基础上,探讨了测量线棒放电量(pC)来检测局部放电性能方法存在的不足和局限性,提出了通过测量中心点电流来检测大型发电机局部放电性能的方案,并讨论了使用不同检测方案的具体要求。在大型发电机局部放电在线远程监测系统的研究方面,在分析原射频监测系统(RFM)的基础上,研究设计了基于智能信息处理的发电机绝缘在线监测与故障诊断系统。完成了系统方案设计、功能设置、性能指标确定、系统预报警和故障报警整定值设定,以及包括高频宽带电流传感器、小信号大动态高频宽带对数放大器和准峰值检波器等关键部件的研制。研制的远程在线监测系统具有大动态范围、高精度、远程通信与监控等功能,该系统经检测,其中放大器滤波矩形系数、系统抗干扰能力等部分性能指标处于先进水平。在模糊故障诊断理论的应用方面,研究基于模糊信息处理的大型发电机绕组绝缘状态判决准则和故障树分析理论,在分析故障特征模糊性的基础上,讨论了故障特征与模糊状态的对应关系,设计了模糊推理机制、模糊知识库以及模糊推理表达式和发电机绕组绝缘故障树结构等,完成的基于模糊理论的绝缘状况专家诊断系统经检验,能有效判断绕组绝缘故障。在神经网络故障诊断理论的应用方面,分析了采集发电机定子绕组中心点信号用于检测转子回路故障的原理。在讨论RBF神经网络故障诊断理论以及权值系数迭代关系的基础上,提出了基于RBF神经网络的发电机转子故障诊断技术和方案,并用实例证实了该方案的有效性。在远程移动智能诊断技术研究方面,讨论了基于Mobile-Agent在发电机绝缘状态智能诊断专家系统方案的构建,在分析Mobile-Agent系统的安全性和稳定性的基础上,提出了MA系统保护方案和安全性措施的设计过程。论文讨论了基于Bayes网络的故障诊断理论用于发电机绝缘故障诊断的方案,在基于信息融合的故障诊断技术研究上做了有益的探索。总结全文,主要的创新点如下：1.研制的大型发电机绕组绝缘状况远程监测系统,具有大动态范围,高精度和远程通信和监控等功能；设计完成了系统关键部件：高频宽带电流传感器、小信号大动态高频宽带对数放大器和准峰值检波器等的研制,系统已经在大型发电机上得到应用。2.提出了基于模糊诊断技术的大型发电机绕组绝缘状态判别方法与准则,结合自主研制的绝缘故障专家诊断系统,实现了对不同容量、不同结构和工况的大型发电机绕组绝缘状况的在线监测与故障诊断。3.提出了基于RBF神经网络诊断技术的大型发电机绝缘监测和故障诊断方案,在分析了发电机定子绕组中心点信号用于检测转子回路故障的原理基础上,用实例证实了该方案的有效性。为基于状态分析的发电机系统剩余寿命预测智能专家系统提供了理论依据。本论文对大型发电机绕组绝缘在线监测与诊断领域的发展,有较大的理论意义和工程应用价值,论文成果创造了较好的经济效益和社会效益。更多还原

【Abstract】 The large power generator is one of the main equipment in the electric station. The security and stability of the system directly affect the operation of the power network, and the insulation fault in a generator failure is in a high proportion with other faults. In the way of the development of regular maintenance of a turbo generator from timing off line to online maintenance, it is a great practical significance using online monitoring and fault diagnosis technology to test the large power generator winding insulation. Under the combined effects of the multiple factors, the generator windings is affected by the electricity, heat, mechanical, chemical and so on, therefore the performance of the insulation will gradually deteriorate. At the same time, a series of electrical properties will change accordingly, and the insulation accident will occur at the end. Hence, using online monitoring of the generator insulation condition for a long-time, and analyzing the changes of the insulation performance parameters is a very effective detection method for evaluating the insulation of a large power generator.On researching the structure and the aging mechanism of the insulation in the generator winding, this paper analyzed the process of thermal aging, mechanical aging and electrical aging of the generator winding during operation and discussed the feature of the single factor electrical aging, thermal aging, mechanical aging and multi-factor aging characteristics; and focused on the analysis of aging parameters on the insulation features. On this basis, the testing methods of the winding insulation were compared, analyzed and studied. This paper expressed the shortcomings and limitations of measuring bar discharge (pC) method to detect the partial discharge performance according to online testing methods at our country and abroad. Meanwhile, this paper proposed the advantages of measuring the current of the neutral point to detect partial discharge of the generator, and pointed out the specific requirements for the different testing programs.As to the research of partial discharge of the remote online monitoring system for a large generator, an online monitoring and diagnosis system for the generator insulation based on intelligent information processing was designed on the basis of the Radio Frequency Monitoring system (RFM). The system design, feature set, performance fix, system pre-alarm and fault alarm value define were completed, especially the high-frequency broad band current sensors, small signal large dynamic high-frequency broad band logarithmic amplifier and the quasi-peak detector were produced. This remote online monitoring system has the functions of large dynamic range, high-precision, remote communication monitoring and so on. The amplifier filter shape factor, the anti-interference ability and some other performance are better than the similar products.Concerning the winding insulation fault diagnosis in a generator, this paper studied the criterion based on the fuzzy state information processing of the generator winding insulation state and fault tree analysis theory; and discussed the corresponding relationship between the fault characteristics and the fuzzy state based on the analysis of the fault characteristics fuzzy; and the fuzzy inference mechanism, the fuzzy reasoning knowledge base, the fuzzy expressions and the fault tree structure of a generator winding insulation were designed. The expert diagnosis system can effectively determine the winding insulation failure based on the fuzzy theory.As to the application of the neural network fault diagnosis theory, this paper analyzed the principle which detecting the rotor fault by use of collecting neutral point signals of generator stator windings. In the discussion of RBF neural network fault diagnosis theory and iterative relationship between the weight coefficients, the rotor fault identification technologies and the solutions were proposed based on RBF neural network, and the solutions were confirmed validity by an example.Concerning the research of the mobile intelligent diagnostics, this paper discussed the construction of the generator intelligent diagnostic expert system based on the mobile agent; and proposed the protection scheme of the Mobile Agent System and the design process of the safety measures based on the analysis of the security and stability of mobile agent. The generator insulation fault diagnosis method was discussed based on the analysis of fault diagnosis theory of Bayes network. A very useful exploration has been done in the fault diagnosis with an information fusion technology.In conclusion, the main innovation points of the paper are as follows: 1. The large generator remote monitoring system applied to the winding insulation condition has a large dynamic range, precision, and remote communication and monitoring functions. The key components of the system such as high-frequency broad band current sensors, small signal high-frequency broad band logarithmic amplifier and the quasi-peak detector and other devices have been designed and completed. It has been applied to the large generators.2. A criteria for identifying state of generator winding insulation based on fuzzy diagnosis technology was proposed, combined with self-made insulation fault diagnosis system. Online monitoring and fault diagnosis for the insulation state of the generator with different capacities and structure was realized.3. The solution of the generator insulation monitoring and fault diagnosis system based on RBF neural network was proposed, the principle which detecting the rotor fault by use of collecting neutral point signals of the generator stator windings was analyzed, and the validity was approved by an ensample. It provided a theoretical basis for expert systems with remaining life prediction intelligent computer system based on state analysis.This thesis has a greater theoretical and practical significance in online monitoring and diagnosing for the generator winding insulation. The greater social and economic benefits have been produced.更多还原