A Neural Network Based Prediction System for Lightning Induced Over-Voltages on Overhead Lines Considering the Soil Electrical Parameters

In this article, we present an artificial neural network-based method to compute the peak lightning-induced overvoltages for the first and the subsequent strokes, respectively. The ground parameters, typically high values of ground resistivity along with the impact of displacement current density, return stroke parameters, stroke location, and height of the overhead line, have been considered to develop the training dataset. The novelty of the proposed approach is that it doesn't require rigorous manipulations in contrast to the numeric techniques such as the finite difference time domain method. However, the prediction model is trained using the dataset from the finite difference time domain simulations for a wide range of the considered parameters. The proposed approach yields reasonable accuracy when validated against the random test dataset. The suggested ANN model predicts the peak lightning-induced overvoltages with a mean squared error of 1.008 using 4168 epochs for the first stroke and 0.074 using 5000 epochs for the subsequent stroke.