Callista, Peggy (2023) Optimalisasi Asp Flooding Pada Core Model Menggunakan Metode Random Forest. Other thesis, Fakultas Teknik Perminyakan.

Text 183210818.pdf - Submitted Version Restricted to Registered users only Download (6MB) | Request a copy

Abstract

In chemical flooding, Alkali-Surfactant-Polymer (ASP) flooding is one of the most widely applied. ASP flooding has been proven as one of the most effective chemical-EOR methods with a Recovery Factor reaching 98% OOIP in laboratory tests. Random Forest is one of the methods in the Decision Tree which belongs to the enseble learner family so that the Random Forest method is strong against outliers in unbalanced data sets, scalable, and able to handle non-linear trends in data sets and can provide predictive results which is high and stable, has minimal effort in tuning its parameters and can be applied in both classification and regression cases. To achieve optimal conditions in ASP Flooding, it is necessary to optimize the ASP Flooding model with the help of Random Forest which can process a lot of data accurately. In this study, the authors will optimize ASP Flooding in the Core model by considering several parameters, including; adsorption, BHP od production well, surfactant concentration, and salinity. The author first conducted a simulation using a model made with Reservoir Simulation Software (CMG-GEM) and CMOST which functions to make 400 Design of Experiment (DoE). Random Forest is accessed using Anaconda Navigator 3.0 software with the Python programming language. After tuning the hyper-parameters by adding several scenarios to the Random Forest, the results obtained from this study are the optimum model for the parameters of adsorption, salinity, and surfactant concentration, namely at max depth 54 and min samples leaf 14 with results R2 ~ 1; MAE ~ 0; and MSE ~ 0, while the optimum model for production well BHP parameters is at max depth 26 and min samples leaf 1 with R2 ~ 0.982; MAE ~ -0.000747; MSE ~ -1.865666.

Actions (login required)

View Item