Caracterização de eletrofácies utilizando Aprendizado de Máquina: aplicação do algoritmo K-Means em perfis geofísicos

Abstract

The application of machine learning algorithms has emerged as an effective and increasingly employed tool across various industrial sectors, bringing significant advancements in the analysis and interpretation of complex data and performing autonomous tasks in a way that simulates human capacity and reasoning. In the oil industry, these machine learning techniques have played a crucial role in reservoir studies, especially due to the uncertainties involved, allowing for a better understanding of well characteristics and improving exploration and production operations in the fields. In this context, the present work aims to characterize electrofacies in two oil wells located in the onshore portion of the Potiguar Basin, in the Jaçanã Field, through the application of the K-means machine learning algorithm to cluster the well data made available by the REATE platform of ANP. The project's methodology involved selecting a set of well logs, which were subjected to a process of analysis, selection, treatment, and normalization for later use in learning. At this stage, the well logs were pre-processed to correct anomalies and standardize measurements. Then, the K-means algorithm was applied to cluster the data using the data from a "test well" and data from each "target well" in two different cases, identifying distinct patterns that were associated with different electrofacies. The results obtained were promising, showing a high correspondence with the actual conditions observed in the studied wells. The pixel-by-pixel image comparison analysis performed by one of the programming libraries applied to the model, used to validate the method's accuracy, revealed a similarity of 97.22% and 98.13% between the actual and predicted electrofacies in the two examples. These findings demonstrate the effectiveness of the project and the use of K-means in electrofacies characterization, offering a robust tool for geological analysis and contributing to the optimization of exploration operations.