Dissertation > Excellent graduate degree dissertation topics show

Research of Power Transformer Fault Diagnosis System Based on Rough Set and Bayesian Network

Author: ZhangZuo
Tutor: LiZhiBin
School: Shanghai University of Electric Power
Course: Proceedings of the
Keywords: transformer fault diagnosis rough set bayesian network MATLAB BNT
CLC: TM407
Type: Master's thesis
Year: 2011
Downloads: 162
Quote: 0
Read: Download Dissertation

Abstract

Power transformer is one of the most important equipment in the power system.Its normal operation is basic guarantees of reliable power supply in the entire power system.However,the power transformer fault is complex and varied,and reasons for fault is complex.So it is difficult to properly determine the type of power transformer.The technology of fault diagnosis for transformer is always taken into account by savant all over the world.The three-ratio method which based on the dissolved gas analysis is widely used in transformer diagnosis. It is easy to use, but also has some defects such as incomplete fault codes and fault divided interval is too absolute.Rough set(RS) theory is an important method in data mining, it is used to extract useful feature and simplify data sets.Bayesian network has a great advantage to solve the fault,which caused by uncertain factors of complex systems. It is considered to be the most effective model in the uncertain field of knowledge representation and reasoning.So in this paper,in order to address the defects of traditional methods,the rough set theory and bayesian network to combine and give full play to the advantages of both methods applied to fault diagnosis of transformer.In this paper,the development and current status of artificial intelligence technology in the transformer fault diagnosis is introduced. Discusses the basic concepts of rough set theory,discretization methods and decision attribute table reduction.Rough set is used to discrete training sample set and reduce the original property.To build the bayesian network,the discretization of rule and minimal set of attributes can be got.It can significantly reduce the number of bayesian network nodes,and beneficial to the structure of network and improvement of learning speed,while keeps the good classification ability of network. Describes the structure and construction method , inference and learning of bayesian network.According to the minimal set attributes by rough set,the bayesian network toolbox of MATLAB(BNT) is used to established transformer fault diagnosis model based on Bayesian network.Encoding rules of The three-ratio method in DL/T722-2000 (Guide to the ananlysis and the diagnosis of gases dissolved in transformer oil) can be used to improve discretization rules.Test results showed that accuracy has been greatly improved. So, in the paper tests were done to demonstrate the proposed method is effective and reasonable.At the end of paper, transformer fault diagnosis system based on bayesian network is designed by using MATLAB and VC++ mixed programming. The designed diagnosis system has a clarity user interface. It not only gives out the type of transformer faults that occurred, but also gives out the reasons which may be cause to fault. It can constantly update, so that the performance of diagnosis system is improved.

Related Dissertations

CLC: > Industrial Technology > Electrotechnical > Transformers, converters and reactors > General issues > Maintenance,repair
© 2012 www.DissertationTopic.Net Mobile