Development of an evolutionary artificial neural network-based tool for selecting suitable enhanced oil recovery methods

Abstract

Enhanced oil recovery (EOR) methods are proven oil recovery processes that allow mature oil fields rejuvenating and increasing their production to economically profitable quantities. The main issues associated with the implementation in practice of these methods relate to their complexity, long implementation periods, economic risks, and high investments. As a result, companies interested in using these EOR methods carry out exhaustive selection processes before embarking on a project of this category. Accordingly, in this work, for selecting the most suitable EOR method for a given mature oil field, a reliable and user-friendly computational tool is developed. This tool allows identifying, based on seven reservoir fluid and rock parameters, the EOR method with the highest probability of implementation success. In the development of the referred EOR methods selection tool, an evolutionary artificial neural network-based approach is utilized, which allows the associated neural network architecture having a relatively high performance. Indeed, the referred tool allows identifying, with an accuracy of up to 94.0%, the EOR method with the highest probability of success. Compared to previous tools, the one developed in this work features a database involving a larger number (eight) of commercial EOR methods. The application of the developed tool to a specific mature field shows its usefulness for selecting the most suitable EOR method for the referred oil field.