Evaluación de un modelo predictivo de machine lerning para la determinación de nuevas de drenaje en el campo X de la cuenca X.
The evaluation of a Machine Learning predictive model for the determination of new drainage areas arises from the industry's need to discover new technologies such as data science, aiming to optimize processes that can be performed in the shortest possible time and with low operational cost. Th...
- Autores:
- Tipo de recurso:
- Fecha de publicación:
- 2023
- Institución:
- Universidad de América
- Repositorio:
- Lumieres
- Idioma:
- spa
- OAI Identifier:
- oai:repository.uamerica.edu.co:20.500.11839/9255
- Acceso en línea:
- https://hdl.handle.net/20.500.11839/9255
- Palabra clave:
- Áreas de drenaje
Predicción
Yacimiento
Drainage areas
Prediction
Deposit
Tesis y disertaciones académicas
- Rights
- License
- Atribución – No comercial
| Summary: | The evaluation of a Machine Learning predictive model for the determination of new drainage areas arises from the industry's need to discover new technologies such as data science, aiming to optimize processes that can be performed in the shortest possible time and with low operational cost. Through this research project, a database was developed for 6 wells supplied by CMG from a field whose name is confidential, which underwent exploratory data analysis (EDA). Following this, 3 phases were carried out implementing a supervised machine learning algorithm through CMOST CMG, where input parameters were evaluated, such as Porosity, rock compressibility, the ratio between horizontal and vertical permeability, endpoints of the relative permeability curve for the rock type, oil relative permeability at connate water saturation (KROCW), water relative permeability at irreducible oil saturation (KRWIRO), exponent for water relative permeability calculation (NW), exponent for oil relative permeability calculation (NOW), rock type, and depth of water-oil contact for Region 1. Once the parameters were evaluated, the past reservoir behavior was adjusted to simulate future behavior with an error of 3.36%, thus creating opportunity indexes that identified the new drainage areas. The model was validated by optimizing two new wells, an experimental Prob_1 with a recovery factor of 2.87%, and prob_2, which the model determined with 3.36%. Lastly, model values were compared against actual values, resulting in a model accuracy of 96.5%. |
|---|