Shima Batubara, Qunazatul (2022) Prediksi Tingkat Keberhasilan Waterflooding Menggunakan Feed Forward Algorithm. Other thesis, Fakultas Teknik Perminyakan.

Text 183210913.pdf - Submitted Version Restricted to Registered users only Download (1MB) | Request a copy

Abstract

Waterflooding is one of the most frequently used EOR methods to increase oil recovery because it can increase 30% -60% of total production. It is necessary to apply a production system performance prediction approach to minimize uncertainty in increasing production figures, such as analytical methods and numerical methods. Artificial Intelligence in the world of oil and gas is not a new thing, but it has often been used in various fields such as exploration, drilling, production and reservoirs. So this is the basis for the prediction of the success of waterflooding research carried out. The purpose of this research was to predict the success rate of waterflooding using an Artificial Neural Network (ANN). The method used in this study is the simulation research method using CMG Imex for reservoir simulation modeling, running CMG CMOST for 500 sensitivity data with input of seven parameters of compressibility, horizontal permeability, vertical permeability, pressure injection, injection rate, thickness, oil saturation and the output is recovery factor using Artificial Neural Network (ANN) with a ratio of 70% of the RF calculation model results for training and 30% model results for testing. In order to get optimal prediction results, trial and error was carried out on the number of hidden layer nodes, so that optimal and stable hidden layer nodes were obtained at node 10 with RMSE values of 0.339035 for training and 0.442663 for testing and MAPE for training 1.15% and 1.62% for testing. The statistical analysis value is 0.906139 for training and 0.899525 for testing data. It can be concluded in this study that the use of ANN in predictions using 10 hidden layer nodes proved to be very good and successful and predictions in this study were classified as High Accurate Prediction.

Actions (login required)

View Item