Prediksi Tingkat Keberhasilan Desain Hydraulic Fracturing Pada Shale Gas Reservoir Menggunakan Artificial Neural Network

Ralda, Ismi (2022) Prediksi Tingkat Keberhasilan Desain Hydraulic Fracturing Pada Shale Gas Reservoir Menggunakan Artificial Neural Network. Other thesis, Universitas Islam Riau.

Text 173210326.pdf - Submitted Version Download (2MB)

Abstract

Hydraulic fracturing is a key element for petroleum engineering since more than 50 years, and is the most alternative way to increase recovery in shale gas reservoirs with very small permeability between 0.01 to 0.00001 md. This ANN method is a deep learning method from input data and produces output data. By using hundreds of data, it is expected to get prediction results from RF using the ANN method optimally, trial and error will be carried out on the number of hidden layer nodes. The purpose of this study is to predict the success rate of hydraulic fracturing performance using artificial neural network (ANN). The method used in this study is a simulation research method using CMG GEM for reservoir simulation modeling and data sensitivity using CMG CMOST with input rock mechanics parameters, rock mineral composition, fracture half length, fracture spacing, fracture width and formation permeability and output in the form of recovery factor. using Artificial Neural Network with Back Propagation method so that it can produce accurate predictions. By using 157 data with a ratio of 75% of the results of the RF calculation model from the CMG software for training and 25% of the model results for testing. In order to get the prediction results from RF using the ANN method optimally, trial and error will be carried out on the number of hidden layer nodes. The optimal and stable hidden layer nodes are obtained at nodes 10 with RMSE and MAPE values in the training data worth 0.048843; 0.561355 and on testing data 0.084627; 0.963113. Other statistical analysis values such as Coefficient determination (R2) are 0.999088 for taining data and 0.997403 for testing data. It can be concluded in this study that the use of ANN in RF prediction using 10 hidden layer nodes has proven to be very good and successful.

Actions (login required)

View Item