Carbonate reservoir rock type classification using comparison of Naïve Bayes and Random Forest method in field "s" East Java

Abstract

Rock type and porosity characterization remain as the main challenges in a carbonate reservoir setting. Rock classification and permeability estimation have normally been done using the rock typing method. Unfortunately, the method utilizes core data, which are often not available and quite expensive. Therefore, there is another alternative using classifier method that cost of wells maintenance or hydrocarbons production may be lower than obtaining the required parameters through the core data process. This study uses two classifier methods, which are Naïve Bayes and the Random Forest method. Those classifier methods make a rock type classification model based on the best rock typing method, which is Lucia method, in the same field. The objective is comparing and choose the best method of those two classifiers. Several assessments are done in the comparing process. Random Forest method has a better result in estimating the error value of the training model, comparing the core permeability with predicted permeability, and the level of accuracy in predicting rock type classification. However, the Naive Bayes method has a separately well distribution of seismic parameters compared to the Random Forest method, which is the most important thing. Therefore, in wells that have no core data, we estimate the permeability value by using the Naive Bayes method.