Application of artificial neural networks for electric arc extinction modeling in high voltage circuit breakers

The analytical models of electric arcs quenching in high voltage (HV) circuit breakers remains very difficult to formulate and requires hypotheses that are simplified exaggeratedly with regard to the reality. On the technical point of view, the application of neuron networks would be a no negligible supply for the simulation of electric arc quenching, enabling thus to be closer with the real properties of the breaker. The aim of this article is to introduce in a first time the neural networks in the mathematical modeling of the arc quenching in high voltage breakers, and then to present a comparative survey between the different training algorithms in order to enable to select the feed-forward propagation neural network and the retro propagation algorithm the most adapted to simulation. This survey has been applied for a line breaker 245kV/50kA/50Hz, for which a default current of 90% of the breaking capacity has been applied.