Dzakwan, Naufal (2023) Implementasi Machine Learning Regression Analytic untuk Meningkatkan Success Ratio Pekerjaan Well Intervention Chemical Stimulation di Lapangan Steamflood Nd. Other thesis, Universitas Islam Riau.

Text 183210380.pdf - Submitted Version Restricted to Registered users only Download (2MB) | Request a copy

Abstract

According to data in 2021-2022, more than 4000 chemical stimulation works have been carried out in the ND steamflood field, both using acid and solvent, with a success ratio of 56%. To maximize the success ratio value of the chemical stimulation work, this research was conducted using the first algorithm of machine learning, namely regression. It is hoped that this study can create a machine learning model that has an accuracy of up to 90% so that it can predict the success ratio of chemical stimulation work in the ND steamflood field. The research conducted began with collecting historical data from well intervention chemical stimulation work: Oil Before Job, Fluid Before Job, Pump Fillage, Water Cut, and Well Head Temperature as well as making machine learning models and continuing with other tests. From the results of the analysis, the parameters that have the potential to get maximum oil gain are as follows: oil BF in the range of 0 to 5 bopd has the highest value of 29, Fluid BF in the range of 0 to 100 bfpd has the highest value of 31, water cut in the range 91% to 100% has the highest value of 51, WHT in the range of 161°F to 200°F has the highest value of 33, and Pump fillage in the range of 61% to 80% has the highest value of 26. From the results of validation tests that have been carried out ten times, it shows machine learning accuracy with an average of 94%. With this accuracy value, it can predict the value of the success ratio. In the test results, there are 175 wells with the highest score shown by the learing machine which can produce positive oil gain, resulting in a success ratio of 100% with a potential error of 6%.

Actions (login required)

View Item