基于贝叶斯网的电力系统故障诊断方法研究

发布时间：2018-01-14 08:00

本文关键词：基于贝叶斯网的电力系统故障诊断方法研究　出处：《西南交通大学》2015年硕士论文　论文类型：学位论文

【摘要】：目前信息技术和通信技术得以高度发展,随着数字化变电站以及智能二次设备的更新换代,二次侧信息实现高度共享和集成,电网故障诊断具备充实的数据基础。90年代初至今,我国的专家学者对电力系统故障诊断做了大量的研究。但是目前只有少数基于模式识别、规则挖掘的方法用于电网故障诊断。研究以解决实际工程问题为目标的故障诊断方法具有重要的研究意义。贝叶斯网络理论是公认具有优异容错能力的模式识别方法。贝叶斯模型的构造对故障诊断效果有很大的影响。传统面向元件建模的贝叶斯故障诊断模型结构相对固定,节点之间的连接方式不够合理,贝叶斯网的计算结果过于依赖保护节点。另外获取先验概率的困难阻碍了贝叶斯方法应用到实际故障诊断。本文针对故障区域判断和贝叶斯模型进行了改进研究。提出一种基于故障隔离的贝叶斯故障诊断算法,在连接方式上将保护和断路器节点分开,实现模型对保护和断路器信息的综合利用。模型去除了元件节点的先验概率,只需要获得保护和断路器的条件概率以及节点证据值就可以进行故障诊断。由于保护和断路器的条件概率对应的事件属于小概率事件,不会因为概率的变化而影响诊断结果。因此典型的保护和断路器条件概率即可满足贝叶斯建模的需求。通过对断路器和保护分层处理,简化了贝叶斯模型的构造,并使贝叶斯模型能够兼容含有失灵保护的元件。同时本文利用简单的时序处理,配合断路器分层进一步减少可疑元件的数量,提高了故障诊断的效率。算法模型的有效性通过电网实际算例和典型电网仿真算例进行了验证。对于故障诊断算法的容错能力,大部分的研究只停留在特殊算例的描述上,无法量化算法的容错性能,不同算法之间难以进行横向比较。本文提出了两个指标用来评价故障诊断算法的容错能力。并对传统贝叶斯模型和基于故障隔离的贝叶斯模型进行了仿真对比。结果表明基于故障隔离的贝叶斯模型具有更高的容错性。
[Abstract]:At present, information technology and communication technology is highly developed, with two digital substation equipment and intelligent upgrading, two side information to achieve a high degree of sharing and integration of power grid fault diagnosis have so far full data base.90 in the early 1990s, experts and scholars in China have done a lot of research on Fault Diagnosis of power system. But at present only a method based on pattern recognition, rules for power system fault diagnosis. Have important significance to study the fault diagnosis method to solve practical engineering problems as the goal. Bias network theory, the pattern recognition method is recognized with excellent fault tolerance. Constructing the Bias model has a great influence on the effect of fault diagnosis. Fault diagnosis of Bias model the structure of the traditional component oriented modeling is relatively fixed, connections between nodes is not reasonable, the calculation of Bias The result is too dependent on the protection of nodes. In addition to obtaining the prior probability difficultieshinder the Bias method is applied to the actual fault diagnosis. This paper studies the fault area and determine the improved Bias model was proposed. A fault diagnosis algorithm based on Bias fault isolation, the connection mode of general protection and circuit breaker node separately, realize the comprehensive utilization of the model the information of protections and circuit breakers. The prior probability model of removing element nodes, only need to obtain protection and circuit breaker condition probability and node values could be evidence for fault diagnosis. Because the protection of circuit breakers and the conditional probability of the corresponding event belongs to a small probability event, not because the probability changes affect the diagnosis results. Therefore typical protection and circuit breaker condition probability can meet the demand. The Bias model of circuit breaker and protection layer, simplified The Bias model, and Bias model can be compatible with failure protection components. At the same time with the application of time series processing simple, hierarchical tie breaker further reduce the number of suspicious components, improve the efficiency of fault diagnosis. The effectiveness of the algorithm model through practical examples and typical power grid simulation examples. For fault tolerance and fault diagnosis algorithm, most of the research is only in the special case description, not fault-tolerant performance quantization algorithm, difficult to compare between different algorithms. This paper proposes a fault tolerance of two indexes for evaluating the fault diagnosis algorithm. And the traditional Bias model and Bias model based on fault isolation the simulation results show that the Bias model. Based on the fault isolation has a higher fault tolerance.

【学位授予单位】：西南交通大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TM711

【参考文献】