|
Oil-filled electrical equipment widely used in the power system , the typical equipment including transformers , reactors and other oil-filled transformers . Large power transformers as the hub of power transmission equipment, the working conditions directly affect the safe operation of the power system , power system in the event of failure and end-users will have significant economic impacts and hazards. Therefore, the study conditions and transformer insulation transformer fault diagnosis technology, early detection of latent transformer fault , to ensure its safe operation , improve power supply reliability is important. Dissolved gas analysis is an effective method for transformer fault diagnosis , how to improve the method of fault diagnosis accuracy and reliability of the current research focus . This paper analyzes the formation mechanism of the gas in transformer oil , as well as various domestic and international oil dissolved gas analysis based on the traditional method of fault diagnosis and artificial intelligence methods . On this basis , chose to build RBF artificial neural network fault diagnosis system , and for the web presence of slow convergence , not convergence problem, apply the Levenberg-Marquardt algorithm to improve the convergence speed of neural networks , reduce network into local minimum points possibilities. The results showed that a large number of instances of diagnostic methods used in this paper have credibility. In order to express the fuzziness of transformer fault , the use of fuzzy mathematics method characterized gas data preprocessing , established based on fuzzy neural network fault diagnosis method insulating oil , and the basis of the method to achieve a set of insulating oil Intelligent Fault Diagnosis System . In the system, combined with features gas production rate , improved IEC three-ratio method for a comprehensive diagnosis. In this paper build fault diagnosis expert system, using VB language development tools directly in the program design designed by VB calls Matlab neural network function procedures, read the diagnostic data to be analyzed and diagnosed . The system is user-friendly, easy to operate, run fast, reliable and accurate fault diagnosis has been used in conventional production in Hunan Province EHV Authority .
|