Automatization of the Analysis and Diagnosis in Power Transformers using Artificial Intelligence

Abstract

The maintenance of power transformers is normally performed by specialists who interpret technical rules developed by recognized institutions and has included little automation. The high costs of equipment, combined with the immense economic and social importance of its adequate maintenance, point to the importance of efficient control of such equipment. The Itaipu Hydroelectric Plant, among the largest of the world and generates around one fourth of all the electricity consumed in Brazil and more than ninety-five percent of the electricity consumed in Paraguay. This article studies the development of a computational solution to automate the analysis and diagnosis of the operating condition of power transformers. Our approach adopts the technique of Expert Systems - ES, from the field of AI. The ES we propose is based on tables which store information about relations between five gases extracted from the transformer isolating oil through gas chromatography. We conclude that the application of the ES approach constitutes an efficient proposal for the automation of the process. In order to deal with unforeseen cases in the available tables, provided through diverse criteria, we also propose improvements obtained through the application of "Rough Sets". In order to accomplish this, we suggest a prototype containing techniques from Symbolic AI, through ES, together with concepts from Fuzzy Sets.