【作者】 张子蓬；

【导师】 张勇传； 胡毅；

【作者基本信息】 华中科技大学 ， 系统分析与集成， 2007， 博士

【摘要】 绝缘子的工作环境恶劣,影响绝缘子正常运行的因素很多且关系复杂,难以全面监测,对绝缘子的劣化、积污、污情发展及闪络等问题可靠诊断和预测是电力输电领域的重要研究课题。本文在研究绝缘子闪络形成机理、发展过程和现有诊断技术基础上,构建了一个全面反映绝缘子运行状态的专家诊断系统;提出将人工智能理论中专家系统、模糊推理系统、Petri网络、神经网络等理论融合并应用于绝缘子运行状态诊断中的新思想;给出了诊断的理论依据和策略方法,并对基于模糊神经Petri网络的推理系统进行了研究和论证。为了很好地设计绝缘子运行状态专家诊断系统,研究、分析了绝缘子闪络形成机理和闪络发展过程及引起绝缘子闪络的原因。通过分析绝缘子闪络前、后泄漏电流和脉冲数与闪络电压、污秽程度、湿度等之间的复杂关系,总结了绝缘子污闪诊断和预测的有效方法,并利用污闪物理模型和数学模型分析了临界电流、电压和电弧的临界长度的表达,为研究绝缘子运行状态和设计智能专家诊断系统提供了理论基础。针对设计绝缘子运行状态专家诊断系统的知识源和推理规则的制定问题,详尽分析了影响绝缘子工作状态的因素及其相关性,并总结了可用于绝缘子运行状态诊断的特征量,提出根据这些特征量综合分析诊断绝缘子的污秽程度、劣化程度和运行状态,并给出了先进的绝缘子监测及诊断的系统和绝缘子闪络定位系统。针对泄漏电流微弱信号中存在噪声信号,提出采用基于阀值的小波降噪方法对信号进行降噪处理。在分析现有绝缘子运行状态诊断方法和策略的基础上,提出采用基于模糊神经Petri网络的专家诊断系统对绝缘子运行状态进行诊断,将模糊Petri网络和神经网络很好结合,设计具有模糊推理、学习和优化能力的绝缘子运行状态专家诊断系统。在设计专家诊断系统时,采用模块化结构,综合运用知识表达方法,并且对很多量的表达进行了定义和模糊归一化处理。在归纳、总结绝缘子运行诊断关键技术和诊断依据的基础上,设计了模糊神经Petri网络的绝缘子运行状态专家诊断推理系统,对模糊Petri网络从概念和理论方面进行了扩展,使网络能够很好的表达推理系统输入、输出之间的推理关系和多输入之间的相互关系。在网络规则的形成、表达方面做了大量工作。在专家诊断系统的推理过程中,采用连续函数的形式进行推理运算,使推理更精确、可靠。推理过程中将模糊Petri网络和模糊神经网络有机结合,并采用了改进的学习训练方法。诊断结果验证了系统的合理性和准确性。本文的研究具有开拓和创新性,为绝缘子运行状态诊断提供了先进的技术和宝贵的经验。全面、系统地总结了本文的工作和研究成果,并指出了有待改进的地方和需要进一步开展的工作。更多还原

【Abstract】 The working environment of insulator is critical and there are many complicated factors influence its running. It is difficult to monitor those comprehensive variables which influence the running of insulator and which indicate the running state of insulator. Now, it is an important task for the domain of power transmission to research the diagnosis and prediction system of the running state of insulator. After the research of the formation mechanism and the developing process of insulator’s flashover and the intelligent diagnosis technology, a diagnosis expert system is designed for the insulator which can diagnose the operating state of insulator outright. The integrating theory based on expert system, fuzzy reasoning system, Petri net, and neural network is put forward to be used to diagnose the running state of insulator. The diagnosis strategies and theory of diagnosis expert system are presented. The reasoning system based on fuzzy neural Petri net has been thoroughly discussed.In order to design the diagnosis expert system of the running state of insulator commendably, the formation mechanism, the developing process and reason of insulator’s flashover is analyzed and researched. After analyzing the complicated relation of some factors, effective diagnosis and prediction method is summarized. Flashover physical and mathematical models are used to analyze critical current, critical arc voltage and the critical length, which provides a theoretical basis for designing the intelligent expert diagnosis system.In order to resolve the problem that the resource of knowledge and reasoning rules of diagnostic expert system of the insulator’s running state, some effective factors and correlation are detail analysised and some characteristic variables are summed. The idea that the running state of insulator is diagnosed synthetically based on those characteristic variables is presented and the advanced monitoring and diagnosing system for insulator and the positioning system for insulator’s flashover are given. In order to eliminate the noise signal of leakage current, wavelet denoising method based on the threshold value is used to eliminate the noise signal of sampled signal.Based on the analysis of diagnostic methods and strategies for insulator’s running, the expert diagnostic system based on fuzzy neural Petri network is proposed to diagnose the operating state of insulator. Fuzzy Petri net and neural network are integrated to optimize system, which makes expert system have the ability of fuzzy reasoning, learning and optimization. In the design of expert diagnostic system, the modularized structure is designed and synthetical method is used to express knowledge. After summing the key technology and judging basis, the expert diagnostic reasoning system is designed for the diagnosis of insulator. Fuzzy Petri net are also extended in its concept and theory, which makes the net’s express of reasoning system relations between input and output is easier. A lot of works have been done for designing system in formation and expression. In expert diagnostic system, continuous function is made for reasoning, which makes the reasoning process more accurate. The improved learning method is used to optimize reasoning system. The diagnostic results validate the reasonableness and accuracy of reasoning system. This research provides effective diagnostic technologies and valuable experience for expert diagnostic system of insulator running.Finally, the dissertation summarizes all the works and results achieved in this dissertation. The further research works to be developed are also put forward.更多还原