Bayesian regularization of neural network to predict leakage current in a salt fog environment

Leakage current (LC) has been monitored extensively to assess the surface conditions of both ceramic and non-ceramic insulators. It has been reported that LC is highly correlated with insulator surface damage and the occurrence of flashover. Hence, it is imperative to predict the LC future value. The objective of this paper is to use Bayesian regularized neural network to predict both the fundamental and third harmonic components of LC under salt fog condition. Three different models of neural network are proposed and each is trained to predict the time series of both the fundamental and third harmonic of LC. The results have shown that there is a high correlation between the fundamental and third harmonic of LC when the nonlinearity of the leakage current increases. Moreover the future value of the LC has been successfully predicted.