Study On The Application Research Of High Voltage Reactor Oil Dissolved Gases Monitoring And Fault Analysis Technology In Gezhouba Power Plant

High voltage reactor is a kind of important devices used in electrical power system, and is closely related to safety of whole system. Dissolved Gases Analysis (DGA) can be used as one of the most important measures to diagnosis internal insulation faults of operating oil-filled electric equipment. How to combine traditional DGA method with online monitoring and intelligent fault diagnosis technology is an important and difficult subject in the field of electric equipment condition monitoring and fault diagnosis, which has significant value in both academia and engineering. Therefore, the research of this paper focuses on the traditional DGA, online monitoring technology and artificial intelligence fault diagnosis method based on bayesian network to meet the need of 500kV reactor insulation monitoring and the appliance of fault diagnosis technology in Gezhouba hydro power plant (GHPP). The research achievements are shown mainly as follows:1) The development about present and future researches of DGA is introduced and the main works and targets of this paper are proposed, according to the requirement of application of dissolved gases on-line monitoring and fault diagnosis in GHPP.2) The principle of occurrence of dissolved gases in oil is introduced and the relationship between ordinary faults inside transformer and gases dissolved in oil is pointed out. The traditional ratio fault diagnosis methods for DGA and then the factors influencing diagnosis veracity are analyzed.3) According to the practical application of oil online monitoring system in GHPP, as well as the analysis to field-measured data, the possibility that the online monitoring devices substitute for the traditional measures is studied and the present shortage in engineering practices is presented.4) Based on the analysis to practical faults cases of 9 500kV reactors in GHPP, their historic oil chromatogram data and the fault diagnosis results, the practicability and veracity of the traditional three ratio methods of fault diagnosis are studied, and some useful suggestions for the safely operating administration of high-voltage reactor in GHPP are proposed.5) Based on the researches of principle Bayesian network diagnosis method, a basic and multiple tree-structured bayesian network classifiers are studied for DGA,with consideration of DGA characteristics. The results of two classifiers used to diagnose a large quantity of oil chromatogram data of transformer are discussed. And at last, the multiple tree-structured bayesian network classifier is testified to be suitable for DGA of 500kV reactors in GHPP.