【作者】 赵熙临；

【导师】 周建中；

【作者基本信息】 华中科技大学 ， 系统分析与集成， 2009， 博士

【摘要】 电力系统的安全、稳定运行要求诊断系统能够快速、准确地分析故障原因并定位故障点。但随着电力系统规模不断扩大，系统运行机制与结构越来越复杂，仅凭传统方式对故障进行判断与控制难以满足实际需要；同时，复杂系统故障信息不可避免地存在不确定性和不完整性，这也为故障的正确判断增加了难度。因此，研究能够有效处理不确定与不完整故障信息的诊断方法，对电力系统的安全运行具有重要意义。本论文针对复杂电力系统故障诊断中不确定信息处理的难题，从电力系统故障诊断机理分析、系统建模、信息融合等几个方面进行深入研究。主要内容包括：电力系统故障诊断模型设计，不确定故障信息处理，D-S证据理论及基于信息熵的马尔可夫融合模型的选择、设计与应用，含有保护和断路器拒动、误动信息的电力系统不确定性故障诊断策略。本文开展的主要研究工作如下：(1)提出了分层电力系统故障诊断Petri网模型及其建模方法。首先，为解决大规模Petri网推理时可能出现的状态组合爆炸问题，采用反向推理方法建立电力系统元件Petri网子诊断模型，各Petri网子诊断模型可根据电网结构与组成进行组合，构成系统诊断模型的区域分层形式；同时，针对大范围复杂故障的诊断问题，在元件Petri网子诊断模型设计中，将相邻元件的状态引入，形成嵌套的模型结构；进一步，为获取缺失故障信息，以Petri网为基础进行元件状态判断模型的设计，其判断结果将被引入元件Petri网子诊断模型参与故障推理，构成系统诊断模型的递进逻辑分层。(2)提出了概率Petri网模型及其不确定信息推理机制，以解决复杂电力系统故障诊断过程中出现的信息不确定问题。将概率的概念引入Petri网，使之具备不确定状态推理能力，有助于解决诊断过程中出现的“可能性”问题；同时，为计算概率Petri网中各有向弧所对应事件的发生概率值，将对应于节点表示状态的概率Petri网演变为一个节点表示过程的有向无环图，应用贝叶斯网络推理技术对该有向无环图中节点发生后验概率进行计算，获取概率Petri网中有向弧对应事件的发生概率，从而实现Petri网模型不确定信息的推理。(3)提出D-S证据理论与马尔可夫过程相结合的信息融合方法，以提高信息融合的针对性。当不确定信息获取过程中出现多论据支持时，利用信息融合技术对多论据进行融合以获取对不确定信息认识的一致性输出，可以提高不确定信息认知的准确性；同时，鉴于各论据对不确定信息判断差异性的存在，研究了D-S证据理论与基于信息熵的马尔可夫过程两种信息融合方法，根据待融合对象间认知的矛盾程度对信息融合方法进行选择以正确反应各论据在判断上的差异，从而提高信息融合的合理性；进一步，针对D-S证据理论的特点，定义被融合论据与融合结果之间满足应用条件：F(pa，pb)≥max{pa，pb}，据此推导出满足该应用条件下D-S证据理论适用的前提，并首次提出了D-S证据理论选择性判据，作为两种融合方法选择的依据。(4)提出了基于信息融合技术的电力系统不确定性故障诊断策略。通过信息判断模型、故障诊断模型、信息融合模型及融合方法选择等环节内在关系的分析，提出了分布式协同处理的电网故障诊断系统框架；在对分布式子系统结构模型及系统总体模型进行研究的基础上，讨论了系统协同工作策略；同时，对电力系统故障诊断流程进行了梳理，明确了各环节之间的联系，并通过实例分析验证了所研究方法的合理性与有效性。论文最后对上述研究成果进行了总结，提出了进一步研究的方向。更多还原

【Abstract】 The security is very important for power grid running. When the fault is occurred, the fault occurred site and reason need to be detected exactly and rapidly to lessen the influence that caused by the fault. But because of the complexity of the power system, it is not practical for the operators to judge the fault immediately only by experience. At the same time, the detected information will be incompletion and uncertainty inevitably. That will enhance the difficulty of the fault diagnosis. Therefore, it is very important to research the fault diagnostic approach that can solve the problem when the detected information is incompletion and uncertainty.This dissertation researched the fault diagnostic problem of power system with incompleteness information based on the analysis of fault diagnostic mechanism、model design of the system and information fusion after the investigation of literatures. The main research content include: fault diagnostic model design, acquired method of incompleteness information, design and application of fusion model about D-S evidence decision rule and entropy based Markov chains, fault diagnostic methods with incompleteness information.Main contributions of this dissertation are stated as below:(1) Put forward a layered model constructed method for the construction of power system fault diagnostic model. In order to solve the problem of fault illation in power system, Petri nets theory is applied in the research. The Petri nets model of the element in grid is designed by backward illation method. The connection of element models will compose in area diagnostic model. The hierarchical structure of the model can solve the problem that the model state will expend quickly when the system is complexity. At the same time, the acquired model of absent data is designed based on the Petri nets if the detected data is incompleteness. The outputs of the information acquired model are the input of the fault diagnostic model, and form a layered model format in logic. In addition, the nested design method of the model is put forward in the process of the model design. The method can resolve the problem of complexity fault diagnosis.(2) The concept of probability Petri nets and the illation mechanism with incompletion information is defined to solve the problem that the detected information of fault is uncertainty. Combine the Petri nets with probability will find the concept of probability Petri nets. Probability Petri nets can achieve the logic illation with information uncertainty. Depending on the comparability of Petri nets and Bayesian network, a transplant method is put forward in the dissertation. The method can transplant the Petri nets into Bayesian network to calculate the generant probability of the absent data by Bayesian illation technique. Then the obtained probability of the absent data can be used into the diagnostic model.(3) The information fusion technology is used to enhance the precision of diagnosis by the combination of D-S evidence decision rule and entropy based Markov chains. When the estimate of the absent data has several estimations, the information fusion technology will be used to obtain the consistent output to reduce the uncertainty of the judgement. Because the difference that exist in these estimations is inevitably. Different information fusion approaches need to be choiced according to the state of the estimations to enhance the rationality of the information fusion. Therefore, D-S evidence decision rule and entropy based Markov chains information fusion technologies are researched in this dissertation. According to the characteristic of D-S evidence decision rule, the restricted condition is defined as: F(p_a, p_b)≥max{p_a, p_b}. The choiced criterion of D-S evidence decision rule is defined according to the restricted condition. The selected of the information fusion technology can be achieved by the choiced criterion.(4) The strategy of fault diagnosis with uncertainty information is put forward. The diagnostic process is definitized by the analyzed within information judge model, fault diagnostic model and information fusion model. And the research is described and analyzed in this dissertation by the application of some cases.Finally, the research is summarized in the dissertation, and the research of next step is also proposed.更多还原