【作者】 张凯；

【导师】 赵法起；

【作者基本信息】 山东农业大学 ， 农业电气化及其自动化， 2013， 硕士

【摘要】 电力变压器是电力系统中最重要的电气设备之一，其运行状态直接影响系统的安全性水平。为了保证电力系统的安全稳定运行，必须加强对变压器的故障诊断。近年来，人们对变压器的监测投入了更多的人力和物力，使变压器的监测水平和监测的准确度得到迅速的提高，监测方法也呈现出了多样化。油浸式电力设备油中溶解气体分析故障诊断技术的研究，对发现变压器内部早期存在的潜伏性故障，提高电力系统的安全稳定运行具有十分重要的意义。由于神经网络对外界的输入样本具有并行处理、学习和记忆、非线性映射、自适应能力和鲁棒性等特点，使得人工神经网络应用于变压器故障诊断得以实现。因此，研究以变压器油中溶解气体为特征量的神经网络故障诊断方法，为变压器故障诊断提供了新的途径。本文选择和训练了适用于电力变压器运行状态及其故障在线监测、诊断的BP、RBF神经网络，对网络的结构、优化和算法进行了探讨，通过仿真试验证明，此两种方法相对于传统的变压器故障诊断的方法具有明显的优越性和更高的故障诊断率。实验数据结果分析表明，RBF神经网络在故障诊断率和训练速度上相比BP网络具有一定的优势。将变压器诊断中典型的油中气体分析法、神经网络方法和MATLAB仿真相结合，采用LabVIEW开发出界面友好、性能优秀的变压器故障诊断系统。首先配置选用系统需要的主要硬件设备，分析了这些设备的基本参数、性能、特点。接着设计出系统的软件部分。试验结果表明新型的系统使得变压器故障诊断实现实时在线监控成为可能；而且诊断结果更加准确，精度更高。本论文的最后，总结了所设计的故障诊断系统的优越性能以及它存在的不足，并且分析了未来故障智能诊断系统的前景和发展方向。更多还原

【Abstract】 Power transformer is one of the most important electrical equipments in the electricsystem. In order to insure the stability of power grid, it is indispensable to be reinforced thetransformer fault diagnosis. Therefore，in recent years，monitoring of power transformer hasbeen given more and more human and material resources，so the level and accuracy ofmonitoring raise much.And monitoring methods are diversity. The study of fault diagnosistechnology based on dissolved gas analysis is very important to maintain the reliable runningof elecrtic power system，and also an efficient method to detect the incipient fault intransformer.The neural network has the paralleling proeessing，learing，memorization，nonlinearitymapping，adaptation ability and robustness etc and strong capability to recognize and classifythe input samples.The possibility of the practical application of artificial neural network todiagnose fault of equipment is come true.So， to study the neural network fault diagnosismethod by the dissolved gases in transformer oil as the characteristics has provided the newway for the transformer failure diagnosis.Selecting and training the BP and RBF neural network which suitable to the powertransformer running condition and fault on-line detection，diagnosis and forecasting.We havediscussed the network construction，optimization and algorithmic，and the simulation resultsshow that the two methods compared with traditional transformer fault diagnosis methodshave obvious superiority and higher fault diagnosis rate..The empirical data result analysisindicated that，the RBF neural network is exuding the higher fault diagnosis rate and thetraining speed obviously compared to the BP neural network. Combining artifieial neuralnetworks and MATLAB experiment with Dissolved Gas Analysis is a typical method and thetransformer fault intelligent diangosis system is of friendly interface and excellent capabilityusing LabVIEW. First of all on the system main hardware devices need configurationselection, and analyses the basic parameters, performance and characteristics of these devices.Then design the software part of the system. Test results show that the new system made itpossible to realize real-time online monitoring of transformer fault diagnosis. And the diagnosis result more accurate and higher precision.In the end，the paper summarizes the excellent capability of the design of the faultdiangosis system and its shortcomings，and then analyses the outlook and development trendof fault intelligent diangosis systems in the future.更多还原