On Re ning Equipment Condition Monitoring using Fuzzy Sets and Arti cial Neural Nets

Utilities invest signi cant time and nances into substation maintenance in order to anticipate failures or accelerated aging in power equipment. Current practices include regular insulation condition tests for switching devices, reactors, power transformers, and so on. Many of the indicators of equipment condition are imprecise, unreliable and costly to obtain on a regular basis. To achieve useful results, engineers must gain extensive experience with particular tests. This work suggests several methods for extracting information from test data and through these methods compares intelligent system approaches. More speci cally, this paper investigates tuning performance based on training examples.