Application of Machine Learning for WAG Parameters Optimization in CO₂-EOR and Geological Carbon Sequestration

Abstract

As the concern of global warming, carbon capture, utilization and geological storage (CCUS) is the most promising way to reduce the emissions of the anthropogenic CO₂ into the atmosphere. Injecting the CO₂ into hydrocarbon reservoir not only storing the CO₂ underground to reduce the amount of CO₂ emitted into the atmosphere, but also enhancing oil recovering to cover the cost of CO₂ sequestration in regional to basin scale. Water alternating gas (WAG) is the most used operation for CO₂-EOR, however, finding the best operating parameters of WAG to keep the balance between CO₂ production and CO₂ storage is also a challenge. In this research, we proposed a machine learning based proxy model, which can substitute the numerical simulation because of its strong advances in high accuracy and low time consumption. Moreover, the proxy model can dig out the deep physical mean of reservoirs by introducing the concept of Most Important Area (MIA). As the result indicates, machine learning has good performance to optimizing the WAG process, and more effective than traditional numerical simulation.

AAPG Datapages/Search and Discovery Article #90350 © 2019 AAPG Annual Convention and Exhibition, San Antonio, Texas, May 19-22, 2019