Application of Deep Learning for Well Log Interpretation and Reservoir Modelling: A Case Study of Niger Delta Region

Abstract

Accurate interpretation of well log data is essential for effective reservoir characterisation and hydrocarbon exploration. Traditional methods, while effective, are often time-consuming, subjective, and limited in handling large-scale datasets. This study investigates the application of deep learning, specifically Convolutional Neural Networks (CNNs), for automating well log interpretation and improving the accuracy of reservoir modelling in the Niger Delta region. Due to the unavailability of proprietary Niger Delta datasets, publicly accessible datasets with similar geological characteristics were utilised to develop and validate the model. The research involved comprehensive data preprocessing, model training for regression and classification tasks, and performance evaluation using appropriate metrics. Results demonstrated that the CNN-based regression model achieved high accuracy in predicting porosity, a critical reservoir property, while the classification model effectively delineated lithofacies from well logs. Visual interpretations confirmed the model’s ability to capture subsurface geological patterns. Despite using non-regional data, the models showcased strong generalisation potential, indicating that with localised training data, accuracy could be further improved. The study concludes that deep learning offers a scalable, efficient, and accurate alternative for reservoir characterisation, supporting real-time decision-making and reducing interpreter bias. Recommendations include sourcing Niger Delta-specific datasets, integrating seismic and core data, and developing explainable AI frameworks to enhance model interpretability and field adoption. This research contributes to the growing advocacy for AI-driven solutions in upstream petroleum exploration and reservoir engineering.