Diseño de un algoritmo de decisión con machine learning para la obtención de una respuesta sobre la aplicabilidad de los equipos ESP, basándose en el análisis estadístico del comportamiento de estos bajo condiciones especiales de campos en Colombia.

Oil production through the use of electro-submersible pumping systems is an industrial process that generates the need to collect and analyze a large amount of information, which is stored with the aim of being used as a statistical base for future processes. In this project, a statistical study was carried out with Machine Learning, by programming an algorithm of approximately 2500 lines of code, and using the Caret methodology in the RStudio software. A total of 586 wells and 51 variables were evaluated, of which 80% was used to train the predictive model, and the remaining 20% ​​was used to determine the predictive capacity of the model. The results with 80% allowed us to observe the way in which the probability is distributed in the levels of the variable "Causa_Raíz". With the results of 20%, graphs were programmed that allowed observing the results of the prediction based on the special conditions of the sand, asphaltene and scale field. This allowed establishing that the algorithm has the ability to predict values ​​different from the original ones. The study is strengthened with the judgment and knowledge of the petroleum engineer, since it allows analyzing the final result of the study, in such a way that the best decision on the applicability of the equipment can be made.

• Fecha de publicación: 2021
• Institución: Universidad de América
• Repositorio: Lumieres
• Idioma: spa
• OAI Identifier: oai:repository.uamerica.edu.co:20.500.11839/8774
• Acceso en línea: https://hdl.handle.net/20.500.11839/8774
• Palabra clave: Análisis estadístico del comportamiento; Bombeo electrosumergible; Condiciones especiales; Statistical behavior analysis; Electro-submersible pumping; Special conditions; Tesis y disertaciones académicas
• License: Atribución – No comercial