基于信息融合技术的油浸式变压器故障诊断方法研究

发布时间：2018-04-17 13:53

本文选题：油浸式变压器 + 故障诊断　；参考：《华北电力大学(北京)》2017年硕士论文

【摘要】：电力传输与人们的生产、生活息息相关,油浸式变压器是输配电系统中关键的枢纽设备,为了不对生产、生活产生影响,及时而准确地检测出油浸式变压器早期潜伏性故障,保证变压器正常、稳定运行,显的尤为重要。本文针对如何提高油浸式变压器故障诊断正确率的问题展开研究,对已有方法进行深入研究,并对其进行改进;在此基础上提出了一种极限学习机与D-S证据理论相结合的故障诊断方法。论文首先针对极限学习机缺乏概率输出的问题,引入了多分类概率极限学习机的方法,此方法先将多分类问题分解为多个二分类问题,再使用Sigmoid函数将二分类极限学习机输出映射为概率输出,然后通过求解一个二次规划问题,使多个二分极限学习机的概率输出融合为多分类概率输出,将多分类概率极限学机应用于变压器故障诊断,为变压器故障诊断提供了概率参考;文中还对D-S证据理论存在冲突证据时,会存在融合结果与实际结果不符的问题,对其进行改进,为此本文引进证据体相似度的概念,计算出证据集的加权平均值,确定证据集主元后,对证据集进行修正,使证据融合在有冲突的情况下,也有较强的处理能力。其次,本文将多分类概率极限学习机与改进的D-S证据理论相结合,建立基于信息融合技术的故障诊断模型,将整个诊断过程分为初步诊断层和融合诊断层。本文将油浸式变压器DGA气体的原始数据构造为气体含量、气体比值、三比值三个特征向量作为初步诊断层的输入,得到三个概率输出,然后使用改进的D-S证据理论进行融合诊断,得到最终的诊断结果。最后,本文在以上研究内容的基础之上,设计并实现了基于信息融合技术的油浸式变压器故障诊断系统。
[Abstract]:Power transmission is closely related to people's production and life. Oil-immersed transformer is the key hub equipment in transmission and distribution system. In order not to affect production and life, early latent faults of oil-immersed transformer can be detected in time and accurately.It is very important to ensure the normal and stable operation of the transformer.In this paper, how to improve the correct rate of fault diagnosis of oil-immersed transformer is studied, and the existing methods are deeply studied and improved.On the basis of this, a fault diagnosis method combining extreme learning machine and D-S evidence theory is proposed.Firstly, aiming at the problem of the lack of probability output, this paper introduces the method of multi-classification probabilistic learning machine, which decomposes the multi-classification problem into multiple two-classification problems.Then the output of binary extreme learning machine is mapped to probabilistic output by using Sigmoid function. Then, by solving a quadratic programming problem, the probabilistic output of multiple binary extreme learning machines is fused into multi-class probabilistic output.The application of multi-class probabilistic limit machine to transformer fault diagnosis provides a probabilistic reference for transformer fault diagnosis, and when there is conflict evidence in D-S evidence theory, there will be a problem that the fusion result does not agree with the actual result.This paper introduces the concept of evidence body similarity, calculates the weighted average of evidence set, determines the principal component of evidence set, modifies the evidence set, and makes evidence fusion in the case of conflict.Also has the strong processing ability.Secondly, a fault diagnosis model based on information fusion technology is established by combining multi-classification probabilistic extreme learning machine with improved D-S evidence theory. The whole diagnosis process is divided into primary diagnosis layer and fusion diagnosis layer.In this paper, the original data of DGA gas of oil-immersed transformer are constructed as gas content, gas ratio and three characteristic vectors as the input of the preliminary diagnostic layer, and three probabilistic outputs are obtained.Then the improved D-S evidence theory is used for fusion diagnosis and the final diagnosis results are obtained.Finally, on the basis of the above research, this paper designs and implements an oil-immersed transformer fault diagnosis system based on information fusion technology.
【学位授予单位】：华北电力大学(北京)
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP202;TM407;TP277

【参考文献】