A FUZZY INFORMATION APPROACH TO INTEGRATING DIFFERENT TRANSFORMER DIAGNOSTIC METHODS

Methods to identify transformer fault conditions before they deteriora te to a severe state include dissolved gas analysis, liquid chromatogr aphy, acoustic analysis, and transfer function techniques. All of thes e methods require some experience in order to correctly interpret obse rvations. Researchers have applied artificial intelligence concepts in order to encode these diagnostic techniques. These attempts have conc entrated on only a single technique and have failed to fully manage th e inherent uncertainty in the various methods. In this paper, a theore tic fuzzy information model is introduced. An inference scheme which y ields the ''most'' consistent conclusion is proposed. A framework is e stablished that allows various diagnostic methods to be combined in a systematic way. Numerical examples demonstrate the developed system.