Implementación de modelos proxy que representen el proceso de inyección de agua como método de recobro mejorado para predecir la producción de crudo y agua de un campo petrolero con base a las variables operacionales de los pozos productores e inyectores

Proxy models are used as an alternative to predict oil and water production in an oil field and to replace the use of reservoir simulation softwares. In this paper, a methodology for the construction and validation of proxy models to predict oil and water production of an oil field was proposed. Thi...

Full description

Autores:
Suárez Malagón, Mario Daniel
Aguilera Ruiz, Valentina
Tipo de recurso:
Trabajo de grado de pregrado
Fecha de publicación:
2021
Institución:
Universidad de los Andes
Repositorio:
Séneca: repositorio Uniandes
Idioma:
spa
OAI Identifier:
oai:repositorio.uniandes.edu.co:1992/51672
Acceso en línea:
http://hdl.handle.net/1992/51672
Palabra clave:
Ingeniería de yacimientos petrolíferos
Inyección de agua en campos petrolíferos
Pozos petroleros-Investigaciones
Industria del petróleo
Ingeniería de petróleos
Campos petrolíferos
Ingeniería
Rights
openAccess
License
http://creativecommons.org/licenses/by-nc-nd/4.0/
Description
Summary:Proxy models are used as an alternative to predict oil and water production in an oil field and to replace the use of reservoir simulation softwares. In this paper, a methodology for the construction and validation of proxy models to predict oil and water production of an oil field was proposed. This methodology is summarized in four main steps: select relevant operational variables and analyze their significance, recollect proxy model training data, proxy model construction, and proxy model validation. This methodology was applied to a particular study case: a section of the Matanegra block located in the oil field Caño Limón. In this case, historical production data was not available, so a section of the Matanegra block was simulated in IMEX in order to recreate historical data that reflect reality. Two different approaches were evaluated in the proxy model construction. In the first approach the operational variables were constant over time and there were two response variables: total oil production and total water production. In the second approach the operational variables change every month and there were two response variables: monthly oil production and monthly water production. This second approach is innovative, as it allows to evaluate how monthly oil production is affected when changing operational variables each month. In each approach, 3 types of proxy models were created for each response variable: polynomial quadratic with interactions, polynomial quadratic without interactions and MARS (Multivariable Adaptive Regression). This results in 6 models per approach for a total of 12 proxy models. For both approaches, the model with the highest predictive power was polynomial quadratic without interactions, followed by MARS and finally, polynomial quadratic with interactions.