Integrated Reservoir Modeling and General Workflow for Determining Minimum Data Set Requirements for CO2 Leakage Detection and Migration Monitoring during Post CO2-EOR Storage

Integrated reservoir modeling and general workflow for determining
minimum data set requirements for CO2 leakage detection and migration
monitoring during post CO2-EOR storage Objective

A generally applicable workflow is developed for credible migration
monitoring and leakage detection of CO2 geological storage during post CO2-EOR
storage with most economical data acquisition investment. A systematic approach
to determine the minimum data set is presented with an integrated reservoir
modeling workflow for post CO2-EOR stage, from the 3D seismic data, geologic
data, well log and core data, and field flowing history data (Figure 1) collected
during CO2-EOR period. The minimum data set considers the minimum well log data
types, as well as maximum vertical and lateral separations of the monitoring wells
and temporal sampling intervals. Reservoir pressure evolution patterns are simulated
with the reservoir modeling with the data so that CO2 migration paths are
predicted.

This project advances the rich data set from Weyburn and
Midale oilfields in south-eastern Saskatchewan, which apply man-made CO2 for
enhanced oil recovery (EOR) for more than 10 years, and become the largest
reservoir on post-EOR anthropogenic CO2 storage with most scrutinized
world-leading research. The structural reservoir model is integrated in Petrel
E&P platform. 3D geostatistical modeling integrating seismic data, well log
data, and core data are analyzed in R. Reservoir simulation during post CO2-EOR
storage is implemented in Compositional & Unconventional Reservoir
Simulator CMG-GEM.   

The minimum parameter set to model the reservoir
hydraulic system, predict storage capacity and monitor CO2 migration that meets
a certain level of risk assessment include:

(1)    The types of
formation petrophysical properties that are most sensitive to the CO2
geological storage prediction and migration monitoring,

(2)    A largest
spatial range of data acquisition that captures the possible CO2 migration
range, which is normally larger than the storage complex,

(3)    The smallest spatial
resolution of sampling measurements, determined from local geologic and
petrophysical heterogeneity and anisotropy,

(4)    The smallest
temporal span of data acquisition for the purpose of reservoir monitoring.

Figure 1 Seismic and well log data in the reservoir

In the exercise of reducing sampling data, the
petrophysical properties of permeability and porosity are simulated in a
stochastic way, and the reduction is accepted only when the pressure evolution
follows the similar results as with the full dataset.  In order to reduce the
dataset, a hierarchical automatic clustering and geostatistic techniques are employed
for hydraulic flowing facies classification with available core and well log
measurements.

(1)    From an
extensive history matching with CMG-GEM reservoir simulation, the field's
effective porosity, anisotropic permeability, and initial fluid composition at
the beginning of (post-EOR) CO2 injection, are calibrated as the baseline for
the minimum dataset derivation process. Permeability field is interpolated by
kriging from core anisotropic permeability and porosity data, and modified to
include local fractures and boundaries from the production history matching.
Pressure distribution evolution is simulated with CO2 injected at various
locations and different injection rates in the field within the rock's mechanic
tolerance. 

(2)    The field of
interest undergoes the primary facies classification, with effective porosity, anisotropic
permeability, fluid composition, and stratigraphic formation indicator (from
seismic amplitude data and interpreted by geologists) as the inputs. The
spatial location of each data point is considered in the process to retain the
spatial continuity of each facies. As a result, the sedimentary hydraulic
flowing facies are automatically categorized honoring geologic information. For
each facies, a secondary typing is implemented from effective porosity,
anisotropic permeability so as to identify subfacies at a lower scale (Figure
2).

(3)    To model the
hydraulic flowing unit spatial heterogeneity of the field, continuous lag
Markov chains based stochastic models are employed to the facies from step (2) to
obtain transition probability matrix.

(4)    Indicator
geostatistical techniques are applied to the facies and subfacies, and
transition rate matrix in 3-dimensional space are estimated.

Figure 2 Result of indicative kriging simulation for facies
distribution

The workflow is applied for 100 realizations for the Weyburn
field, and the monitoring wells are recommended from each of the realization
following by reservoir simulations in CMG-GEM. The workflow shows that seismic
data takes a positive part in the facies recognization in the lateral
directions, but not in the vertical part due to the limited seismic vertical resolution.
The indicative kridging of permeability and porosity field provides more than 70%
reduction of required well data by virtue of the facies indication.