Application of machine learning methods and neural networks in the objectives of lithofacies mapping and reservoir properties assessment: analysis and selection of methods

The paper discusses the application of artificial intelligence (AI) methods to address challenges in lithofacies mapping and the assessment of reservoir properties.The choice of an AI method depends on the nature of the data, objectives of the study (such as classification, regression, clustering, or image segmentation), and requirements on the final interpretation and modeling results.An analysis of various machine learning (ML) algorithms including the support vector machine (SVM), random forest (RF), neural networks, etc. were conducted.Evaluated effectiveness of each method was evaluated on the basis of open-source data and geological datasets.Advantages and disadvantages of these methods were analyzed and factors influencing on the selection of an appropriate AI method were identified.Classification of geological problems and corresponding AI methods, encompassing SVM, RF, linear and polynomial regression, k-means clustering, hierarchical clustering, and convolutional neural networks (CNN) were presented.The article also introduces open-source ML platforms such as TensorFlow, PyTorch, and Keras along with factors influencing on the selection of the optimal AI method for lithofacies analysis and reservoir property assessment.Recommendations to select the most suitable AI methods for specific objectives were provided.The importance of data volume and quality in selection of AI method and prevention of model overfitting was emphasized.