Application of Gene Expression Programming (GEP) in Modeling Hydrocarbon Recovery in WAG Injection Process
Water alternating gas (WAG) injection has been successfully applied as a tertiary recovery technique. Forecasting WAG flooding performance using fast and robust models is of great importance to attain a better understanding of the process, optimize the operational conditions, and avoid high-cost bli...
Guardado en:
| Autores principales: | , , |
|---|---|
| Formato: | article |
| Lenguaje: | EN |
| Publicado: |
MDPI AG
2021
|
| Materias: | |
| Acceso en línea: | https://doaj.org/article/23a7a9b3a56d462f809ebc487c38e3cd |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| Sumario: | Water alternating gas (WAG) injection has been successfully applied as a tertiary recovery technique. Forecasting WAG flooding performance using fast and robust models is of great importance to attain a better understanding of the process, optimize the operational conditions, and avoid high-cost blind tests in laboratory or pilot scales. In this study, we introduce a novel correlation to determine the performance of the near-miscible WAG flooding in strongly water-wet sandstones. We conduct dimensional analysis with Buckingham’s <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mi>π</mi></semantics></math></inline-formula> theorem technique to generate dimensionless numbers using eight key parameters. Seven dimensionless numbers are employed as the input variables of the desired correlation for predicting the recovery factor of a near-miscible WAG injection. A verified mathematical model is used to generate the required training and testing data for the development of the correlation using a gene expression programming (GEP) algorithm. The provided data points are then separated into two subsets: training (67%) to develop the model and testing (33%) to assess the models’ capability. Conducting error analysis, statistical measures and graphical illustrations are provided to assess the effectiveness of the introduced model. The statistical analysis shows that the developed GEP-based correlation can generate target data with high precision such that the training phase leads to <i>R</i><sup>2</sup> = 92.85% and <i>MSE</i> = 1.38 × 10<sup>−3</sup> and <i>R</i><sup>2</sup> = 91.93% and <i>MSE</i> = 4.30 × 10<sup>−3</sup> are attained for the testing phase. The relative importance of the input dimensionless groups is also determined. According to the sensitivity analysis, decreasing the oil–water capillary number results in a significant reduction in <i>RF</i> in all cycles. Increasing the magnitudes of oil to gas viscosity ratio and oil to water viscosity ratio lowers the <i>RF</i> of each cycle. It is found that oil to gas viscosity ratio has a higher impact on <i>RF</i> value compared to oil to water viscosity ratio due to a higher viscosity gap between the gas and oil phases. It is expected that the GEP, as a fast and reliable tool, will be useful to find vital variables including relative permeability in complex transport phenomena such as three-phase flow in porous media. |
|---|