Perpustakaan Digital ITB

The process of increasing oil recovery is never separated from the term Enhanced Oil Recovery or commonly abbreviated as EOR. The most commonly used EOR methods are CO2 flooding and N2 flooding. Several studies show that CO2 and N2 can be applied together in the form of injectable mixtures to increase oil recovery. Analysis of petroleum technical and economic aspects is needed as an effort to evaluate the feasibility of the CO2-N2 flooding project that will be applied in the field. This effort can be answered by making a prediction model that can determine reservoir performance in the CO2-N2 flooding project. This study is carried out by focusing on making a predictive model for the 10-year CO2-N2 flooding project. A literature study is conducted to generate a good model for the CO2-N2 flooding project comprehensively. 18 parameters that affect the technical and economic aspects are used to build predictive models and are distributed in three ways, namely discrete, continuous, and formula. The CMG-CMOST software is used to produce experiments using the polynomial regression method through data collection stages using Latin hypercube sampling of 1586 data, training sample data, and then data that has passed quality control, namely 1119 data will be used as training model data and verification data. 80% of the data that has passed the quality control process is used as training model data to create predictive models. The objective of the predictive model is to review the Profitability Index (PI) which represents the feasibility of a CO2-N2 flooding project. The prediction model is made using the polynomial regression method, namely linear, simple quadratic, and quadratic. The validation process is carried out on this model by comparing the results of the initial model using CMG-GEM software and the results of polynomial regression. From this process, the difference in results is not too far away and the R-Square value is still acceptable. This study can be developed and improved by re-selecting parameters, using various scenarios, and adding predictive modeling methods so that better results can be obtained.