DGA: A novel strategy for key gases identification in power transformers

Abstract

There are several proposals for the dissolved gas analysis (DGA) of power transformers through the use of artificial intelligence. All these proposals, based on fuzzy logic, knowledge-based systems and neural networks, among others, are oriented to the diagnosis of the equipment based on the expert knowledge obtained over the years. This paper proposes a new approach in the use of neural networks, not for transformer diagnosis, but rather for the identification of key gases in mineral oil-immersed transformers. The proposal is tested on the dielectric oil of mineral origin traditionally used in transformers since its expected behavior is known. The key gases identified with this proposal coincide with those found in the literature, so the strategy is efficient. However, the potential of the work relies on the application to natural esters, field still under investigation.