Automating Prediction of the Formation Tops and Lithology Changes in Real-Time While Drilling Using Artificial Neural Networks

Abstract Determining formation tops in hydrocarbon wells is a critical aspect of drilling operations, impacting decisions related to casing setting depths and drilling fluid design. Traditionally, estimates are made based on geological data during the exploration phase, but these lack precision. Real-time updates to formation tops occur as new wells are drilled, incorporating data from various measurements like Rate of Penetration (ROP), gamma ray, formation cuttings, and mud logging. However, these measurements come with limitations such as high costs, manpower requirements, and time or depth lags. This study introduces an innovative alternative using Artificial Neural Networks (ANNs) to accurately predict formation tops. The ANNs model incorporates drilling mechanical parameters and ROP to address limitations in existing techniques, aiming to predict genuine ROP changes irrespective of other drilling parameter fluctuations. Real-field data from two Middle Eastern vertical oil wells was utilized to validate this approach. The study explored different smoothing techniques and identified the moving average technique with a span of 5 as effective in smoothing data while preserving structure. The ANNs model was trained with 70% of Well-A’s data and tested on the remaining 30%, with validation using Well-B data. The sensitivity analysis revealed that an ANNs model with 25 neurons and one hidden layer, using the trainlm training function, achieved optimal results. For testing data from Well-A and the validation date of Well-B, correlation coefficients (R) were 0.94 and 0.98, respectively. This innovative approach offers a cost-effective solution, overcoming challenges associated with traditional methods. It provides accurate, real-time predictions for formation tops during well drilling, ensuring better decision-making in casing programs and drilling fluid design, ultimately contributing to more efficient and cost-effective well operations.