Prediksi Tingkat Keberhasilan Steamflooding Menggunakan Dengan Metode Artificial Neural Network

Nisa, Azkhiatun (2020) Prediksi Tingkat Keberhasilan Steamflooding Menggunakan Dengan Metode Artificial Neural Network. Other thesis, Universitas Islam Riau.

Preview

Text 153210410.pdf - Submitted Version Download (3MB) | Preview

Abstract

Many approaches attempt to predict the performance of oil production systems, including analytical and numerical methods. However, estimation errors and significant deviations occur between the predicted results and the actual field data. Artificial neural networkis one of the methods in artificial intelligence that can effectively provide predictions that have a maximum error less than other methods, so that they can make better decisions to reduce unproductive time. The application of this artificial neural network method has been widely used in previous research on several fields such as exploration, drilling, production and reservoir. Steamflooding is one of the EOR methods included in this thermal method which is widely used in increasing oil recovery because it can recover 50-60% of OOIP. This is what underlies this predictive research on steamflooding. The method used in this research is the simulation research method using CMG Stars for reservoir simulation modeling and data sensitivity using CMG CMOST with the input parameters of API, oil viscosity, steam quality, porositas, reservoir permeability, rate injection, reservoir temperature and output in the form of recovery factor using artificial neural network with back propagation method so that it can produce accurate predictions. Prediction model using artificial neural network method with backpropoagation algorithm to the recovery factor value with 2187 sample data obtained relatively good results using 20 hidden layer nodes with RMSE for training data 0.090 and testing 0.085. MAPE (mean absolute percentage error) for training data 0.483245 and testing 0.469495. Coefficient correlation (r2) 0.99707 for training and testing 0.99735 so that it can be classified as having high accuracy results because it is close to a value of 1.

Actions (login required)

View Item