In oil and gas reservoirs, understanding how multiple fluids flow simultaneously is
crucial for optimizing production, and relative permeability is the key parameter
governing this complex process.
Traditional barriers of resource-intensive and costly core flooding experiments
have motivated our study to overcome these challenges, using machine learning
models for a comprehensive understanding of fluid behavior within reservoirs. This
research aims to explore the potential of machine learning to enhance the accuracy
of oil and water relative permeability prediction..
This study utilizes data from core samples representing different lithologies,
specifically sandstone from the Talang Akar Formation and carbonate from the
Baturaja Formation, collected from three oil fields in Indonesia. This initial dataset,
comprised of 519 data points, was obtained after removing data points exhibiting
the impact of fine migration to ensure the accuracy of the machine learning models.
This study introduces the use of hybrid machine learning models to predict oil and
water relative permeability curves without relying on absolute permeability as an
input feature. Altough absolute permeability has been shown to be strongly
correlate ted with relative permeability in previous studies, this research
deliberately excludes it to explore alternative predictive parameters. By utilizing
readily available data, including water saturation, irreducible water saturation,
residual oil saturation, and porosity, our models achieve accurate and rapid
predictions of relative permeability. This novel approach has the potential to make
reservoir characterization easier and reduce the need for time-consuming and
expensive special core analysis.
Furthermore, our study explores the potential of these models to correct core
sample data affected by fine migration, a common challenge in special core
analysis. This finding offers a promising approach for improving the reliability of
relative permeability data, particularly in cases where fine migration is an
important consideration.