A Machine Learning Approach to Predicting Wellbore Instability in the Niger Delta

Abstract Wellbore instability is a major drilling challenge in the Niger Delta which, in the worst cases, have been known to result in significant non-productive time (NPT) of over 100 days and up to 10 - 20% of total drilling costs. Some studies state that the total financial losses due to wellbore instability exceed $1 billion annually. Therefore, the imperative of predicting wellbore instability cannot be overemphasized. In this Study, Well-logging data from three wells in the Niger Delta comprising parameters such as Depth, Gamma Ray (GR), Photoelectric Cross-Section (PE), Acoustic Transit Time (DT24QI), Compensated Neutron Porosity (CNCF), Borehole Size/Mud-Weight Corrected Density (ZDNC), Caliper Log, and Bit Size, were analyzed to identify key predictors of wellbore enlargement. Analysis of the data revealed that Depth, PE, GR, CNCF, ZDNC, and DT24QI were the most influential parameters. Two machine learning models, Multiple Linear Regression (MLR) and Artificial Neural Networks (ANN), were developed and trained using these parameters. The ANN model consistently outperformed the MLR model, exhibiting significantly lower Root Mean Square Error (RMSE) values (Well 1: 0.2095, Well 2: 0.2693, Well 3: 0.3813) and Mean Absolute Error (MAE) values (Well 1: 0.1265, Well 2: 0.1697, Well 3: 0.2244) compared to the MLR model. Additionally, the ANN model achieved significantly higher R-squared values (Well 1: 0.9575, Well 2: 0.9183, Well 3: 0.8045), indicating a stronger correlation between predicted and actual wellbore enlargement rates. Analysis of the residual plots revealed that the ANN model exhibited lower variance, reduced bias, and a more symmetric distribution of residuals, suggesting improved accuracy and robustness. In conclusion, the study demonstrated the potential of ANNs to accurately predict wellbore enlargement and improve wellbore stability management in the Niger Delta, enabling proactive risk mitigation strategies, reducing drilling costs, and enhancing overall drilling efficiency and safety.