Using Artificial Neural Networks to Estimate the Equivalent Resistivity from typical Transmission Line Towers Grounding Arrangement in a Two-Layer Soil

This paper presents a procedure to estimate a homogeneous equivalent resistivity for typical transmission line tower ground arrangements that are immersed in a stratified soil model. This technique is based on artificial neural network training due to a dataset developed in a numerical routine to compute grounding resistance. The results obtained from Levemberg-Marquadt algorithm and Bayesian regularization show good accuracy compared to the applied dataset and to the measurement data from a 230 kV transmission line, a practical case analyzed. We note medians lower than 5% and acceptable values of standard deviation for practical cases, which suggests that the proposed procedure can be used for real-world conditions.