Studi Kasus Prediksi Distribusi Fasies Vertikal Di Sumur Reservoir Batu Pasir Menggunankan Algoritma Random Forest

Ramadhan, Fajar (2022) Studi Kasus Prediksi Distribusi Fasies Vertikal Di Sumur Reservoir Batu Pasir Menggunankan Algoritma Random Forest. Other thesis, Universitas Islam Riau.

Text 163210672.pdf - Submitted Version Download (3MB)

Abstract

Knowledge of facies is important in reservoir characterization activities. Information about facies can be obtained through core data and reading well logs. Core data can provide an overview of rock lithofacies and reading logs shows rock electrofacies at various reservoir depths. In the last 20 years, many studies have been carried out to distribute reservoir facies through the electrofacies classification method. Electrofacies classification methods are increasingly in demand when the use of ML in data classification activities begins to develop. The use of ML in data classification makes it possible to make predictions using parameters or attributes in training and testing activities. This study predicts facies in a sandstone reservoir in an oil field using various types of log data such as GR, SP, LLD, LLS, Density and Neutron obtained from 7 wells. One of the wells will be selected to be a reference well that will serve as a center for facies development and the formation of predictive models. A two-step approach was applied in this study. The first step is to do the number of facies using yahoo K-Means. The next step is the formation of a predictive model through electrofacies classification using the RF algorithm. The first model is formed using four log attributes, namely GR, SP, LLS and Densiras logs. The second model is formed with four different attributes, namely LLS log, LLD, Density and Neutron. The two steps of this research will be carried out using the Python programming language. The results of this study indicate that reservoir facies can be predicted through clustering analysis. The K-means approach predicts the optimal facies classification into three types of facies. The formation of the prediction model using the RF algorithm in both models has a high level of accuracy that is 95% in the first model and 97% in the second model so that the application of this classification becomes more efficient than using conventional methods to characterize reservoirs.

Actions (login required)

View Item