eprintid: 19689
rev_number: 2
eprint_status: archive
userid: 1
dir: disk0/00/01/96/89
datestamp: 2024-06-04 14:19:25
lastmod: 2024-06-04 14:19:25
status_changed: 2024-06-04 14:15:37
type: article
metadata_visibility: show
creators_name: Alatefi, S.
creators_name: Agwu, O.E.
creators_name: Azim, R.A.
creators_name: Alkouh, A.
creators_name: Dzulkarnain, I.
title: Development of multiple explicit data-driven models for accurate prediction of CO2 minimum miscibility pressure
ispublished: pub
keywords: Computational efficiency; Forecasting; Genetic algorithms; Genetic programming; Multiple linear regression; Neural networks; Sensitivity analysis; Solubility, Accurate prediction; CO2 utilization; Data-driven model; Explicit models; Gas flooding; Minimum miscibility pressure; Multi-gene genetic programming; Multiple linear regressions; Multivariate adaptive regression splines; Support vector regressions, Carbon dioxide
note: cited By 1
abstract: This study presents utilization of multiple data-driven models for predicting CO2 minimum miscibility pressure (MMP). The aim is to address the issue of existing models lacking explicit presentation. With a database of 155 data points, five models were developed using artificial neural network (ANN), multigene genetic programming (MGGP), support vector regression (SVR), multivariate adaptive regression splines (MARS), and multiple linear regression (MLR). Comparative analysis was conducted using statistical metrics (R2, MSE, MAE, RMSE), and sensitivity analysis was performed on input variables. The results showed that ANN and SVR had comparable predictive performance (ANN: R2 = 0.982, MSE = 0.00676, MAE = 0.9765, RMSE = 0.082), SVR (R2 = 0.935, MSE = 0.0041, MAE = 0.72, RMSE = 0.064) followed by MARS, MLR, and MGGP. Sensitivity analysis revealed that reservoir temperature was the most influential parameter across all models, except for the MLR algorithm where injected CO2 amount was crucial. These models can be used for a wide range of CO2 MMP ranging from 940 psi to 5830 psi, thus rendering them useful for any reservoir globally. These models offer improved accuracy and computational efficiency compared to existing ones, potentially reducing costs associated with laboratory experiments and providing rapid and precise CO2 MMP predictions. © 2024 Institution of Chemical Engineers
date: 2024
official_url: https://www.scopus.com/inward/record.uri?eid=2-s2.0-85190944350&doi=10.1016%2fj.cherd.2024.04.033&partnerID=40&md5=54d5ef254a677894d17e7bf9d5f81013
id_number: 10.1016/j.cherd.2024.04.033
full_text_status: none
publication: Chemical Engineering Research and Design
volume: 205
pagerange: 672-694
refereed: TRUE
citation:   Alatefi, S. and Agwu, O.E. and Azim, R.A. and Alkouh, A. and Dzulkarnain, I.  (2024) Development of multiple explicit data-driven models for accurate prediction of CO2 minimum miscibility pressure.  Chemical Engineering Research and Design, 205.  pp. 672-694.