This work demonstrated that artificial intelligence methods could be used in predictive maintenance, obtaining valuable results to improve decision-making in scheduled maintenance plans applied to power transformers immersed in mineral oil in power generation plants. The objective of this research is to interpret the Dissolved Gas Analysis (DGA) results in the oil of the power transformers of the generation plants using Artificial Neural Networks (ANN) to obtain a broad panorama for decision-making in the maintenance programs that are planned in the generation plants. The bibliographic research is the primary basis for this study because it consists of the compilation of a set of results of the dissolved gas analysis performed on several power transformers from different parts of the planet. The data were previously validated and determined the state of the transformers after obtaining their respective interpretations. The machine learning method was used by constructing an artificial neural network in Matlab and training it with the data collected and validated.

Predictive maintenance, dissolved gas analysis, power transformers, mineral dielectric oil, artificial neural networks