Real-Time Sonic Log Prediction Using Artificial Intelligence Tools Case Study of Hassi Terfa Oilfield southern Algeria

Abstract

In practice, sonic velocity in rocks is measured using well-logging tools or through laboratory-based techniques providing reservoirs geometrics and petrophysics parameters. However, these methods can be expensive and time-consuming, making it necessary to explore alternative approaches that provide accurate results more efficiently and it may affected by signal loss or alteration caused by high mud density. This study applies various artificial intelligence models to predict sonic velocity in Hassi Terfa oilfield in real time. Machine learning (ML) techniques, including Gradient Boosting (GB), Random Forest (RF), and Extreme Gradient Boosting (XGB), were employed using drilling data as input parameters. A vertical well with 295 data points was analysed, focusing on rock formations composed of sand and claystone. The results demonstrated that all three models effectively predicted sonic velocity with high accuracy. The correlation determination (R²) values were 0.812 for the RF model and 0.789 for the XGB model. The highest accuracy was achieved using the GB model, which had an R² reach 0.835, making it the most effective model for predicting sonic velocity. These models provide a cost-effective and time-efficient solution for estimating sonic velocity in subsurface formations.