Selection of station surge arresters based on the evaluation of failure probability using artificial neural networks

Considering the lightning strikes, failure probability of arresters due to the highly nonlinear voltage stresses and current characteristics (v-z) of the arresters is not obvious. It means that the evaluation of failure probability and the energy capacity can be a difficult and long task. This paper presents an artificial neural network (ANN) based approach to estimate directly the failure probability of an arrester and then indirectly select the proper energy capacity to fulfill the adopted failure rate. The application of ANN is applied to a group of arresters and the results of the ANN test coincide with the analytical ones.