Understanding & Screening of DCW through Application of Data Analysis of Experiments and ML/AI

Abstract

An oil recovery technique, different composition waterflooding (DCW), dependent on the varying injected water composition has been the subject of various research work in the past decades. Research work has been carried out at the lab, well and field scale whereby the introduction of different injection water composition vis-a-vis the connate water is seen to bring about improvements in the oil recovery (improvements in both macroscopic and microscopic recoveries) based on the chemical reactions, while being sustainable from ease of implementation and reduced carbon footprint points of view. Although extensive research has been conducted, the main chemical mechanisms behind the oil recovery are not yet concluded upon. This research work performs a data analysis of the various experiments, identifies gaps in existing experimentation and proposes a comprehensive experimentation measurement reporting at the system, rock, brine and oil levels that leads to enhanced understanding of the underlying recovery mechanisms and their associated parameters. Secondly, a sustainable approach of implementing Machine Learning (ML) and Artificial Intelligence Tools (AIT) is proposed and implemented which aids in improving the screening of the value added from this DCW recovery. Two primary interaction mechanisms are identified as part of this research, gaps in current experimentation are identified with recommendations on what other parameters need to be measured and finally the accuracy of application of ML/AI tools is demonstrated. This work also provides for efficient and fast screening before application of more resource and cost intensive modeling of the subsurface earth system. Improved understanding, knowledge and screening enables making better decisions in implementation of DCW, which is a sustainable recovery option given the current state of affairs with zero carbon and net zero initiatives being on the rise.