Machine Learning Application for Reservoir Facies Prediction in NAS Field, Niger Delta, Nigeria

Abstract

Recent technological advancements have significantly improved the accuracy of subsurface geophysical interpretations. Facies classification is a critical task in subsurface investigations, as it provides insight into the physical, chemical, and biological conditions experienced by formation units during sedimentation. This study applies machine learning techniques to classify reservoir facies using well log data from the NAS Field, Niger Delta. Data from six wells were combined to train supervised machine learning models, while a seventh well was reserved as a blind test for validation. The workflow included data preparation, exploratory analysis, model development, and performance evaluation. The models evaluated include support vector machine, random forest, decision tree, extra tree, multilayer perceptron, k-nearest neighbour, and logistic regression. The dataset was divided into training, testing, and blind test sets. Model performance on the blind test well was evaluated using the Jaccard index and F1-score. The results show that k-nearest neighbour achieved the highest performance (0.98 and 0.99), followed by support vector machine (0.85 and 0.91) and extra trees (0.84 and 0.91). Other models, including logistic regression (0.79 and 0.87), decision tree (0.73 and 0.84), neural network (0.73 and 0.84), and random forest (0.58 and 0.73), exhibited comparatively lower performance. These results demonstrate the effectiveness of machine learning techniques in improving facies prediction accuracy and reducing computational time. The methodologies applied in this study highlight the potential of machine learning as a reliable tool for subsurface facies classification.