In this paper, the ambient temperature values and load signals are applied in an architecture of artificial neural network with the objective of estimating the internal temperature of oil-immersed distribution transformers. The architecture of neural network used in this application is a multilayer perceptron. The training of the network was carried-out using the ''Resilient Propagation'' algorithm and it was based on design details and experimental data relative to the oil-immersed distribution transformers. Simulation results of the proposed approach indicate that this methodology can be efficiently used in the protection processes of transformers, increasing the selectivity, reliability and the management of the electric energy distribution system.
Transformer oil; artificial neural networks; parameter identification; artificial intelligence; estimation algorithms
