A Neural Architecture to Analysis of Dissolved Gases in Insulating Oils used in Transformers

I. Nunes da Silva, M. Massaki Imamura, and A. Nunes de Souza (Brazil)


Transformer oil, neural networks, parameter identification, estimation algorithms, artificial intelligence.


The analysis of insulating oils used in transformers is done through physical-chemical tests, which determine the state of the oil; and chromatography tests, which determine possible faults in the equipment. This work concentrates on determining, from a new methodology, the concentration of gases dissolved in insulation oil, and the existent relationship between the parameters obtained by physical-chemical tests with those provided by chromatography tests. The capacity of neural networks on complex non-linear functions realization becomes them an attractive approach to the analysis of dissolved gases in insulating oil used in transformers. From this methodology, it has become possible to estimate the concentration of gases and to identify the existent relationship between the physical-chemical parameters and the amount of present gases in insulating oil. Simulation results are presented to validate the proposed methodology.

Important Links:

Go Back