The
production of sand (also known sanding) is an universal phenomenon that occurs during
the hydrocarbon production in the weakly consolidated sandstone reservoirs. Sand
production may cause many undesirable problems including, but not limited to, filling
and blocking of wellbore, wellbore collapse, plugging of perforations or
flowlines, erosion failure of downhole and surface facilities, and additional
cost of remedial and cleanup operations (Rahmati et al. 2013). Nowadays,
it has been well accepted that certain amount of sand production can be helpful for improving near-wellbore
formation conductivity and productivity. But poor management or uncontrol of sand
production can lead to problems of sand overproduction with severe consequences
aforementioned (Garolera et al. 2019). For
decades, the problem of sand production has been approached with the use of
sand avoidance techniques, called sand control methods. These methods involve mechanical
method to physically block sand grains with screen or gravel packing, and
chemical method to reconsolidate the formation and prevent sand from flowing
with resin or other chemical agents (Dong et al. 2009, Parlar et al. 2015, Garolera et al. 2019).
As
the foundation of sand control design, sand production prediction is extremely
important work to clarify the sanding mechanism and law, which can offer strong
support for sand control optimization (Dong et al. 2015,2017). Over the decades, due to its importance in hydrocarbon
production industry, systematic efforts have been made to develop the numerical
methods for sand production prediction. The early studies on sand prediction approach
focused on the macroscopic sand production law (Weingarten et al. 1995; Nouri et al. 2002; Papamichos et al.
2018), which mainly depended on the solid failure mechanics
of compression, tension and shear effect (Risnes et al. 1982; Van Den Hoek et al. 2000). These methods concern the use of rock
strength, formation stress distribution and failure criterion of rock to calculate
the critical drawdown pressure for onset of sand production (Wan et al. 2004; Rutqvist et al. 2012). This type of sand prediction models can
only answer under what production condition (production
rate or bottomhole flowing pressure) the wells begin to produce sand, but is
unable to predict the sand production rate and volume quantitatively (Crook et al. 2003; Dong et al. 2017). In order to make volumetric
predictions of sanding, series of numerical models for sanding simulation and prediction
were developed continuously. The early sand production and erosion models for
predicting sand production in unconsolidated formation was presented by
Vardoulakis et al (1996). This mathematical model was based on mass balance of
the produced solids and flowing fluid in porous media, and can be used to
predict the sanding rate with production time. This model was developed
furtherly by Papamichos et al (1998) by coupling the poro-mechanical
behavior of the solid-fluid system with the erosion behavior of the solid
matrix due to fluid flow. Then, further improvements were made by Papamichos et
al (2010,2019) and
Gravanis (2016). This group of models
provide mainly an access to predict sand production rate based on flow-coupling
and stress analysis. In recent years, the new developed sanding simulation
methods, in general, are categorized as continuum and discontinuum approaches. In the continuum approach, cemented granular materials of
sandstone formation are treated as continuous matters in deriving the governing
differential equations (Morita et al. 1989; Nouri
et al. 2006, 2007; Papamichos et al. 2010; Hussein et al. 2018).
This type of sanding prediction models depends on the presupposition of
continuous and homogeneous medium of formation. Garolera
et al (2019) presented a micromechanical
approach based on zero-thickness interface elements for modelling advanced
localization and cracking states of cemented granular materials. The increase
in computational resources due to discretization of the grain area is
compensated by modelling the remote area with continuum finite elements. The
model is capable of modelling localization of deformation, disintegration and
cracking of cemented rock formation. However, the heterogeneity is not fully
considered in his model. The typical method of discontinuum approach is discrete element method (DEM)(O’Connor et al. 1997; Wan et al. 2003),
which is an useful access to simulate sand production, especially to understand
the mechanism of sanding (Rahmati et al. 2013;
Ghassemi et al. 2015; Han et al. 2016). However, it’s difficult to describe the heterogeneity of rock properties and its
application is limited in large scale simulation (Climent et al. 2017). To date, similar studies also have been done to
sand production prediction for gas hydrate bearing sediments with particular
consideration of hydrate decomposition (Cheng et
al. 2010; Li et al. 2013; Pinkert et al. 2015; Uchida et al. 2016; Kajiyama et
al. 2017; Dong, et al. 2019). Generally speaking, the prediction results
of above-mentioned continuum and discontinuum methods just show the macroscopic
sanding law with oil or gas production, but are not capable to describe the
sanding pattern and cavity shape of near-wellbore formation and their
heterogeneity.
In an actual well in weakly
consolidated sandstone reservoir, after a long-term production with sanding,
the continuous removal of solid from formation tends to cause the obvious
change of microscopic structure and porosity-permeability properties (Wan et al. 2003; Dong et al. 2017,2020). From
the engineering point of view, one of the interesting issues the petroleum
engineers care is what the sanding pattern and cavity shape would be after a
given period of sanding production, the recognition of which can essentially
help to optimize sand control design. Unfortunately, although major
improvements have been achieved in the past decade, it’s still a difficult and
challenging issue to describe the sanding pattern and cavity shape of an actual
sanding well before sand control design and implementation. What we just know
is that the sanding cavity may be very complex and non-homogeneous depending on
the characteristic of formation properties (Dong et
al. 2020).
In this work, we construct a new particle-scale microstructure model of weakly consolidated formation core and the subsequent numerical methodology to simulate the microscopic sanding process during production. The proposed microstructure model describes the heterogeneity of formation using random function, considering the particle size distribution (PSD) curve, well logging data, formation stress distribution and dynamic rock strength. This integrated methodology provides a new access to predict and describe the sanding cavity shape of an oil well after long-term sanding production, which is essentially meaningful for sand control design optimization.
Sand
production in weakly consolidated sandstone reservoir is a complex phenomenon.
The sanding performance depends on various factors involving sand grain size,
grain sphericity and inclination angle, intergranular cementation strength,
stress distribution, fluid properties, fluid flowing field, etc. We present a microstructure model in this work as an effective
access to reconstruct the structure of formation near wellbore, which can
reproduce the complex behavior of intergranular aggregation and offer the
foundation for subsequent sanding simulation. The proposed microstructure model
not only fully utilizes the limited information such as PSD curve, well logging
data and principal stress, but also specifically take the heterogeneity and randomness
of reservoir properties in account.
Based on the microstructure model, the proposed sanding simulation approach is capable of simulating the microscopic sand production process, depicting sanding cavity characteristic and predicting sanding law quantitatively. In particular, the simulation method and programs can be implemented to visually present the whole process of sand production and sanding cavity propagation. From the simulation results of series of case analysis, the sanding performance shows some interesting and special characteristics corresponding to the heterogeneity and randomness of reservoir properties characterized by the microstructure model. Sensitivity analysis results show that sanding frontier and ESC radius decrease with increasing average cohesion strength and increase with increasing average liquid production rate. For weakly consolidated sandstone reservoir with different properties, it is found that three typical sanding patterns involves PLS, PWS and CCS patterns. In an actual well, the sanding pattern may be the combination of two or three of the typical patterns. The present simulation method and program can be used to predict the sanding cavity shape along the well axis and identify the production profile with severe sanding. This will help to determine the focus target for sand control implementation and improve its effectiveness considerably.