Perpustakaan Digital ITB

Facies classification is a crucial component in petrophysical evaluation and reservoir characterization, as it directly relates to variations in rock properties and hydrocarbon distribution. This study aims to interpret conventional well log data and Formation Micro Image (FMI) data to identify facies at each depth and determine the number of facies clusters in two wells, FCB-01 and FCB-02, located in the Jatibarang Field, West Java. The methodology combines supervised learning using the Random Forest algorithm for facies classification and unsupervised learning with the Gaussian Mixture Model (GMM) to objectively determine the number of facies clusters. Conventional well log data from both wells and FMI data (available only in FCB-02) were analyzed using Python in Google Colab. The workflow includes data preprocessing, interpolation, normalization, model training, accuracy evaluation, and visualization of classification and clustering results. The results demonstrate that machine learning approaches can enhance the objectivity of facies interpretation and offer an efficient alternative in the absence of core data. This study recommends the acquisition of FMI data in more wells and the development of the classification model into a broadly applicable system for exploration and field development stages.