Page 185 - Machine Learning for Subsurface Characterization

160 Machine learning for subsurface characterization

weighted to accurately reflect their effects on the process. To create an index, it is
importanttoselectrelevantfeatures,identifythenatureofrelationshipbetweenthe
features and the process to be tracked, and finally assign weights to the features.
Features implemented in an index should be based on content validity,
unidimensionality, amount of variance, and degree of specificity in which a
dimension is to be measured.

2 Properties influencing EOR efficiency of light-hydrocarbon
injection
In this chapter, EOR efficiency of light-hydrocarbon injection in shale
formation is assumed to be dependent on the following properties:

a. Movable oil volume
b. Miscible oil volume
c. Water content
d. Kerogen and bitumen content
e. Minimum miscibility pressure (MMP)
f. Volume fraction, dip, aspect ratio, and orientation of fractures
g. Pore size distribution and dominant pore size
h. Compositions of the injected light hydrocarbon and connate hydrocarbon
i. Pore wettability
The oil displacement efficiency of miscible light-hydrocarbon injection
depends on several parameters, such as MMP, pore structure, oil
composition, gas composition, fractures, and formation dip. MMP is the
lowest pressure required for connate oil and injected light hydrocarbon to
achieve miscibility. When reservoir pressure is higher than MMP, the
injected gas can achieve miscibility with reservoir oil resulting in viscosity
reduction, oil swelling, interfacial tension reduction, and single-phase flow
[15]. In this study, we are interested in the volume of miscible oil, defined as
the portion of oil that can achieve miscibility with the injected light
hydrocarbon. On the other hand, movable oil relates to the hydrocarbon
residing in pores that do not trap the hydrocarbon due to high capillary
pressure or due to pore isolation. Movable oil volume and miscible oil
volume are both important parameters governing the EOR efficiency.
EOR efficiencies of gas injection in shale formations using CO 2 , light-
hydrocarbon (C1-C2) mixture, and N 2 have been investigated using
laboratory core flooding and numerical modeling by Alharthy et al. [16].
Recovery factors of light hydrocarbons and CO 2 are comparable, whereas to
achieve high EOR efficiency using N 2 , a higher reservoir pressure is
required to maintain the miscibility because N 2 has a relatively high MMP [17].
Miscible displacement in shales is different from that in conventional
reservoirs because of the nanoscale pore sizes in shales. The injected light
hydrocarbon mixes with the oil in the matrix by molecular diffusion and
advection instead of direct displacement of oil in the matrix [16, 18]. Due to