Modelling Hydraulic Fracturing Pada Shale Gas Reservoir Menggunakan Extreme Gradient Boosting (XGBoost) Algorithm

Subijakto, Muhammad (2022) Modelling Hydraulic Fracturing Pada Shale Gas Reservoir Menggunakan Extreme Gradient Boosting (XGBoost) Algorithm. Other thesis, Universitas Islam Riau.

Text 173210432.pdf - Submitted Version Restricted to Repository staff only Download (1MB) | Request a copy

Abstract

In an effort to meet energy needs, especially oil and gas, production of oil and gas from conventional reservoirs is not only sufficient, but requires support from unconventional reservoir production. One of the products from unconventional reservoirs is natural gas. The focus of the research is shale gas, shale gas is a type of natural gas produced from a shale gas reservoir. The characteristics of this shale gas reservoir have very low permeability and porosity values, namely 10-6 md and 2-8%. Because it has these unique characteristics, hydraulic fracturing is the right choice for the production of shale gas from shale gas reservoirs. In this study, researchers used machine learning: Extreme Gradient Boosting (XGBoost) Algorithm for modeling hydraulic fracturing in shale gas reservoirs. Beginning with the creation of a basecase model using CMG-GEM and CMG-CMOST is used to create an iteration model to obtain 500 Design of experiments (DoE). After the basecase model, the initialization results, and the iteration model are obtained, it is continued by using the XGBoost algorithm with the Python Programming Language for modeling. The six parameters used as input data (independent variables) are fracture conductivity, fracture spacing, half- length fracture, matrix permeability, matrix porosity, and reservoir thickness, as well as hyper parameter tuning. Of the 500 DOEs that have been prepared, data sharing is done with a ratio of 80% for training data and 20% for testing data. During the simulation carried out with the XGBoost algorithm and Jupyter Notebook media, a predictive model was obtained with an R2 (R-Squared) value of 0.989 (training) and 0.984 (testing). In addition to the R2 value, the Mean Absolute Error (MAE) and Mean Square Error ( MSE) whose value is close to 0. The results of the comparison of the recovery factor values are 63.497% (actual value basecase CMG) and 63.562% (XGBoost algorithm). The results of this research show that the XGBoost algorithm can be relied upon to predict the recovery factor value with an accuracy rate of 98.4% and very high time efficiency.

Actions (login required)

View Item