Prediction of leakage current peak value

Abstract

The characteristics of leakage current (LC) have been widely used to study the relation between the contamination level, the peak value of LC and flashover in outdoor insulator. In this paper, a feed forward nonlinear autoregressive artificial neural network (ANN) model has been trained to predict the LC peak value from the past data history of LC peak time series. The applied voltage, the fog conductivity and the length of each insulator were varied during data collection phase to monitor and record the change in LC peak developing on the insulator surface over different environmental conditions. The ANN was then trained and tested using the collected data under variable conditions. The results have shown that as the average peak of LC changes with time intervals, the contamination accumulation on the insulator surface followed a similar trend and hence the instantaneous peak of the LC is a good measure of contamination accumulation on the insulator surface. The proposed ANN model can be used to predict the future peak value of LC for one full hour and hence can be used to alert the utility's engineer of any critical condition before it happens.