Analisis Faktor Signifikansi Adsorpsi Polimer Pada EOR Dengan Pendekatan Artificial Neural Network

Ramadhany Atmadja, Fouja (2022) Analisis Faktor Signifikansi Adsorpsi Polimer Pada EOR Dengan Pendekatan Artificial Neural Network. Other thesis, Universitas Islam Riau.

Text 173210049.pdf - Submitted Version Restricted to Repository staff only Download (2MB) | Request a copy

Abstract

In the polymer flooding process, the polymer will be adsorbed into reservoir due to the nature of the reservoir, polymer characteristics, rock surface properties, and mineral composition. The presence of large adsorption will cause loss of polymer from solution and loss of mobility control effect. This causes a decrease in the effectiveness of the polymer. Therefore, it is necessary to know the factors that significantly affect the adsorption value. In this study, a study was conducted on the factors that significantly affect the adsorption of the polymer injection so when designing this chemical, we can determine the things that hinder the success factors of the polymer used. This study uses an Artificial Neural Network (ANN) approach because the process is faster, more efficient, and can predict more accurately than conventional methods. Meanwhile, reservoir modeling is done using CMG-STARS. The input parameters in this experiment are reservoir temperature, polymer concentration, molecular weight, injection flow rate, and salt concentration. Of the five parameters mentioned above, this research is focused on determining the parameters that most influence adsorption on polymers using the ANN approach. As a comparison of the accuracy of ANN, this study also used the Random Forest approach to determine the parameters that most influence adsorption on polymers. The results obtained from studies conducted using both the ANN algorithm and the Random Forest algorithm, it is known that the factor or parameter that most influences polymer adsorption is polymer concentration, while reservoir temperature, molecular weight, injection flow rate, and salt concentration have a low effect. For predictive modelling, it is known that the Random Forest algorithm is better than ANN because its R2 value is getting closer to 1, 0.968 in training data and 0.933 in testing data, while R2 on ANN is obtained at 0.858 in training data and 0.819 in testing data.

Actions (login required)

View Item