Based Transformer Oil Dissolved Gas Analysis Intelligent Fault Diagnosis

During the seventh party generation meeting of Yunnan Province, it was established to take the electric power with the water electricity for the lord as one of the pillar industry of economic society development in Yunnan. Since then, the electric power of Yunnan has already got a booming development. The electric power transformer is one of the core equipments of the electric power net. And along with the quick development of transmitting and changing electricity technology in extra high voltage and super high pressure , the enlargement of the capacity and the cover of electricity net , the normal circulate of the electric power transformer will produce more and more great influence to the electric power system safe and social economic performance , so to find the fault early and in time has a important meaning.This topic is the research and the realization of the transformer faults diagnosis system. The purpose of the system establishment is to make use of artificial intelligence to analysis and process the data, which was acquired from transformer state detection, then control the transformer's state—the movement or the fault.This thesis combines the concrete fulfillment of the topic, carrying on to the key technique of fault diagnosis system thorough of theories and practice research. Main is:(1)It deeply analyzes the principle that taking the DGA as character to diagnose the internal fault of transformer, and makes research and application to three-ratio diagnosis and grey relation diagnosis.(2) To collect as far as possible many fault sample data (main is the DGA data) through check a great deal of related document and data, and construct the training sample of the nerve network.(3)It carries on a thorough research to the basic theories of the nerve network, and an improvement to the BP network for its slow characteristics of train speed; discusses how to set up the mold method of the nerve network, makes sure the reasonable network structure and training parameter with empirical formula and repeating experiments.(4) It carries on a deeply research to the principle of expert system, the gain of knowledge , the structure of the knowledge base, the design of reason logically machine , database and so on.(5) Put together the expert system and the nerve network knot, analyze and design a mix expert system, of which it applies an information fuse (many experts to diagnosis strategy ) in the reason logically mechanism. And it uses the example to explain the application of many expert voting and adding the power strategy. In practice, this kind method has certain applied value. This is an innovation.During the period of development, this thesis carries on instructive quest to the bottleneck problem, the trained samples collections of the nerve network, and put forward correspond of resolution. Also does the instructive elucidation and research for the newly arisen technique of the information fusion in the meantime. This topic, which takes the practical application as the purpose, the system intelligence as characteristics, raises the capability of transformer fault diagnosis in real time and open .Currently the system is basically established, the system tests enunciation: the system can satisfy a concrete applied request and have a better result.The design method of the system have referenced worth to the other similar system.