Resumo:
Substation Equipments that use insulation oil for isolate their internal parts
needs a periodic maintenance program to detect possible fails like cellulose
deterioration of coils insulation, short circuit between their springs, dissolved gas in
oil caused by its deterioration, etc. Actually, preventive maintenance programs have
been used by generation, transmission and distribution companies, as a fundamental
tool to identify incipient faults, trying to avoid that these faults come to take away the
equipment from electrical system, carrying out great financial losses caused by
decreasing invoicing, payment of fines to regulatory agency or by decreasing of
system reliability. Trying to avoid these inconvenient, on-line sensors and intelligent
artificial (IA) techniques has been found application on electrical system engineering.
This dissertation is a study of one of these techniques – gas chromatography
associated with neural networks – looking to support presents and futures fault
diagnosis based on results from chromatography by the analysis of dissolved
insulation oil gases during the useful power transformer life, avoiding this way the
inconvenient related above, making easy the decision of engineers and technicians
about the predictive maintenance of these equipment and also serving as a base for
the on-line sensors actuation diagnosis if installed on these allowing yet an estimated
old age degree and so the useful age of transformer. Techniques like that from this
study may be associated with other IA tools like fuzzy logic, genetic algorithms,
expert system and others, consisting the system called hybrid, attempted to get the
best solution for the problem.