Deep learning approach for predicting the lithofacies using standards and measurements taken during the drilling process

Abstract

Drilling is crucial for oil and gas exploration but is costly and time-consuming. Enhancing productivity, reducing hazards, and cutting costs are ongoing challenges. Deep learning has emerged as a promising solution for improving decision-making in drilling operations, requiring less human intervention. Key parameters like surface pressure (SPP), rotary speed (RPM), and rate of penetration (ROP) are essential for creating models that predict subsurface conditions. Accurate identification of oil reservoirs is vital due to the risk of destruction from mud density. Engineers often rotate the drill bit at the same location, which is inefficient and costly. Real-time identification of reservoir access using drilling data can significantly reduce costs and improve productivity. Machine learning, especially deep learning, is effective in predicting geological layers and lithofacies, enhancing drilling safety and efficiency.