Bibliographic citations
Prudencio, G., (2018). Desarrollo de un sistema de selección de un método óptimo de recuperación mejorada de petróleo basado en red neuronal evolutiva [Tesis, Universidad Nacional de Ingeniería]. http://hdl.handle.net/20.500.14076/16437
Prudencio, G., Desarrollo de un sistema de selección de un método óptimo de recuperación mejorada de petróleo basado en red neuronal evolutiva [Tesis]. : Universidad Nacional de Ingeniería; 2018. http://hdl.handle.net/20.500.14076/16437
@mastersthesis{renati/709714,
title = "Desarrollo de un sistema de selección de un método óptimo de recuperación mejorada de petróleo basado en red neuronal evolutiva",
author = "Prudencio Baldeón, Guillermo",
publisher = "Universidad Nacional de Ingeniería",
year = "2018"
}
Enhanced oil recovery (EOR) methods are proven oil recovery processes around the world, which allow mature oil fields to rejuvenate and increase their production to economically profitable quantities. Several countries consider these methods as an attractive option to increase their security of energy supply due to the high potential of increasing their oil reserves. The problem of implementing these EOR methods in practice lies in their complexity, long implementation periods, economic risk and high investments. This causes that the companies of the sector interested in these methods carry out an exhaustive selection analysis before undertaking a project of this type. In the present thesis, an optimal selection system of an EOR method is developed, which allows to identify the EOR method with the highest probability of implementation success for a determined mature oil field. The system is developed using an evolutionary neural network model and using information from successful past EOR projects around the world. More specifically, in this work an easy-to-use reliable system has been developed which allows the identification of a group of eight EOR methods (ASP Injection, Immiscible CO2 Injection, Vapor Injection, Miscible CO2 Injection, Nitrogen Injection, Polymer injection, in-situ combustion, and miscible hydrocarbon injection) an optimal EOR method, with an accuracy of 93.9%. For identification, seven reservoir fluid and rock parameters are used (porosity, initial oil saturation, depth, permeability, API gravity, viscosity and temperature). The developed system differs from other previous systems, in the fact that it contains in its database a greater number of commercial EOR methods, which makes it a system with a greater range of selection. In the construction of the system an evolutionary neural network (RNE) has been used, which has allowed the architecture of the neural network, the basis of the structure of the system, to have a high performance. Additionally, a new methodology for selecting artificial neural network (RNA) architectures from a space of ANNs has been proposed. This methodology allows to select an RNA architecture in a fast and reliable way, which reduces the search time and generates a high performance RNA structure. As part of the application of the tool developed, EOR methods have been identified with a greater chance of success in the mature oil fields of Block 192 located in the Marañon basin, Peru. The results obtained indicate that the EOR methods to be applied are in terms of probability of success, 85% of miscible hydrocarbon, 12% of in-situ combustion and 3% of nitrogen injection.
This item is licensed under a Creative Commons License