A novel prediction model of grounding resistance based on long short-term memory

This study aims to investigate the use of Long Short-Term Memory (LSTM) models for predicting temporal variations in grounding resistance using time series data. This analysis is the first to apply LSTM models to grounding resistance prediction, utilizing experimental data, including soil resistivity and rainfall. The LSTM model is trained, validated, and tested with various parameters, enabling a comparative assessment of its accuracy in capturing grounding resistance variations. Furthermore, the study benchmarks the LSTM model’s performance against traditional Artificial Neural Networks, confirming the LSTM’s superior predictive accuracy regarding time-dependent changes in grounding resistance. The results of the prediction show that LSTM significantly surpasses traditional methods in terms of mean absolute percentage error, with an improvement of 72.73% across various metrics.