Enhanced oil recovery by nanoparticles flooding: From numerical modeling improvement to machine learning prediction
Nowadays, enhanced oil recovery using nanoparticles is considered an innovative approach to increase oil production. This paper focuses on predicting nanoparticles transport in porous media using machine learning techniques including random forest, gradient boosting regression, decision tree, and ar...
Guardado en:
| Autores principales: | , |
|---|---|
| Formato: | article |
| Lenguaje: | EN |
| Publicado: |
Yandy Scientific Press
2021
|
| Materias: | |
| Acceso en línea: | https://doaj.org/article/be6dd5985fb34eff82a92eb3a1389c52 |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| Sumario: | Nowadays, enhanced oil recovery using nanoparticles is considered an innovative approach to increase oil production. This paper focuses on predicting nanoparticles transport in porous media using machine learning techniques including random forest, gradient boosting regression, decision tree, and artificial neural networks. Due to the lack of data on nanoparticles transport in porous media, this work generates artificial datasets using a numerical model that are validated against experimental data from the literature. Six experiments with different nanoparticles types with various physical features are selected to validate the numerical model. Therefore, the researchers produce six datasets from the experiments and create an additional dataset by combining all other datasets. Also, data preprocessing, correlation, and features importance methods are investigated using the Scikit-learn library. Moreover, hyperparameters tuning are optimized using the GridSearchCV algorithm. The performance of predictive models is evaluated using the mean absolute error, the R-squared correlation, the mean squared error, and the root mean squared error. The results show that the decision tree model has the best performance and highest accuracy in one of the datasets. On the other hand, the random forest model has the lowest root mean squared error and highest R-squared values in the rest of the datasets, including the combined dataset. |
|---|