Prediction of leakage current of composite insulators in salt fog test using neural network

This paper presents a new prediction method for the level of the fundamental component of leakage current in the early aging period. Several silicone rubber (SIR) insulators were tested in salt-fog chamber and the LC was continuously recorded, A neural network has been used to predict the level of LC in the early stage of aging of the SIR insulators. Initial value of LC and its increasing slope in the first hour are used as the input of the network and the value of LC after 10 hours is the output of the network. It was found that a 2-layer feedforward back propagation with a biased output is a suitable network to predict the LC hours based on its initial values with a maximum of 15 % error.