Sei sulla pagina 1di 11

MODELLING Construction of 3D structures of the test sites The main work task so far in the WP02 has been

the construction of 3D models for the different test sites, describing the structural and geological setting. The models are based on subsurface data from boreholes and seismic investigations. The geological setting as well as amount of information varies greatly between the different sites. A big challenge has been to construct comparable models from respective sites. The main task has been to visualize the different sites and to generate exportable depth data (X, Y, Z) representing bounding surfaces for different layers (aquifer, caprock, faults etc.). An example of mapped surfaces at the South Scania site is presented in figure WP 02-3. A large part of the executed work has involved subsurface mapping of identified reservoirs, caprocks and faults. The results have then been handled in a GIS software as to generate exportable data files and grid files which can be handled by the numerical modelling tools. D2 - Modelling In support to the natural analogues studies, numerical investigation is underway describing the processes leading to CO2 escape from caprocks. The objectives have been to: Model two-phase flow replacement in heterogeneous caprocks, ground truth to experimental work. Include coupling to TMC processes. Investigate significant combinations of processes which could lead to failure of containment system. (MUSTANG_18mos_Core_of_the_Report) The modeling work will extend most of the duration of the MUSTANG project and the procedure and modeling scenarios outlined will be continuously updated and developed as the work proceeds, as more data and methods become available. The overall objective of the site models is to evaluate; (i) The general behavior (spreading and trapping) of CO2 during injection and storage; (ii) How much CO2 can be safely injected into the aquifer and how this can be optimized by means of different injection strategies. These results will provide insight into the issue of how the true (technical) capacity is in relation to the theoretical volume-porosity based capacity? (iii)At what rate and in which manner and which phase composition the injection should be carried out, (iv) What are the possible risks, such as possible leakage through the faults and fractures, negative effects of pressure buildup and CO2 during injection, consequences of the large scale pressure plume, etc. (36_mos_core_of_the_report_1208029) The modeling will address the characteristics of these sites in terms of their suitability for CO2 geological storage. The overall objective is two-fold: Firstly, to get an understanding of the behaviour of these particular sites in terms of CO2 geological storage and secondly, to get an overview of what type of data characteristics can be encountered in potential CO2 storage sites in general. In order to achieve these goals, it is important that a common approach is used in organizing the data and in interpreting it into model parameters. Therefore, a generic framework of data compilation has been established, which comprehensively describes the essential information needed for the modelling to be carried out within the project. MUSTANG_D021_Templates Test Sites Data 2.4. Demonstration example (Hontomin)

We simulate CO2 injection in a deep saline aquifer accounting for chemical reactions and compare the results with a simulation that does not consider chemistry. We use a simple axisymmetric geometry of a 100 m thick horizontal aquifer, of which the top is situated at 1500 m depth, which corresponds to the depth of the reservoir at the Hontomin test site. The model expands laterally 5 km. An injection well with a radius of 0.15 m is placed in the center of the domain. The properties of the aquifer correspond to those of a permeable limestone with homogeneous grain size and are detailed in Table 3. Therefore, the entry pressure is low and the shape parameter of the van Genuchten (1980) retention curve is high. Relative permeabilities follow a linear law of saturation for both phases. The initial conditions are hydrostatic pressure and constant gas pressure of 0.1 MPa. As a first step, a steady-state calculation is carried out to ensure equilibrium for the pressure field. The hydraulic boundary conditions are a prescribed CO2 mass flow rate at the injection well (2.5 Mt/yr) during 1 year, a constant pressure on the outer boundary and the upper and lower boundaries are no flow boundaries. However, the nature of the outer hydraulic boundary condition does not affect the results. This is because the radius of the pressure perturbation cone is smaller than the radius of the domain for the injection time scales presented here. Therefore, the model behaves as an infinitely acting aquifer. The temperature of the aquifer is 60 C, which corresponds to a surface temperature of 10 C and a geothermal gradient of 33 C/km. The mesh is made of unstructured quadrilateral elements. The size of the elements is of 5 m close to the injection well and increases up to a size of 30 m close to the outer boundary. Overall, the mesh has 6600 elements. Table 3. Material properties of the Aquifer aquifer. Property Permeability, k (m2) 10-13 Gas entry pressure, (MPa), eq. (8) 0p 0.02 van Genuchten m, eq. (8) 0.8 Porosity 0.1 Youngs modulus, E (GPa) 10 Poisson ratio, 0.3 Figure 7. Gas pressure evolution at a point placed 75 m away from the injection well in the middle of the aquifer with and without considering chemistry. Figure 8. Liquid saturation degree evolution at a point placed 75 m away from the injection well in the middle of the aquifer with and without considering chemistry. Figure 7 and Figure 8 show the temporal evolution of gas pressure and liquid saturation degree, respectively, at a point located 75 m away from the injection well in the middle of the aquifer with and without considering chemistry. Both simulations give very similar results. Hence, flow dynamics is slightly affected by chemical reactions, at least for the time scale of these simulations, i.e. 1 year. However, CO2 dissolution into the brine reduces the pH, dissolving calcite, which increases the porosity of the aquifer (Figure 9). The calculation of the porosity show spurious instabilities, which have not been resolved yet for these first results. Figure 9. Porosity after 1 year of injecting 1 Mt of CO2. As a reference of scale, the thickness of the aquifer is 100 m. Figure 10 displays the CO2 bubble after 1 year of injecting 1 Mt/yr of CO2. The shape of the bubble is identical in both simulations. So only that of the simulation considering chemical reactions is shown. Chemical reactions have been calculated as a function of CO2 pressure. Figure 11 shows CO2 pressure after 1 year of injecting 1 Mt/yr of CO2 for the simulation

considering chemical reactions. Note that the region containing CO2 is larger than the CO2 bubble, especially in the lower part of the aquifer below the CO2 bubble close to the injection well. This can be explained by a vertical downward flux driven by density difference (Figure 12) in the aqueous phase below the CO2 bubble (Figure 13). Since dissolved CO2 increases brine density, an unstable situation occurs in which denser brine is placed on top of lighter brine. Therefore, the CO2 saturated denser brine moves downwards, bringing CO2 to the bottom of the aquifer. This CO2 saturated brine also dissolves calcite, as can be observed in Figure 9. Figure 10. Liquid saturation degree after 1 year of injecting 1 Mt of CO2. As a reference of scale, the thickness of the aquifer is 100 m. Figure 11. Gas pressure after 1 year of injecting 1 Mt of CO2. As a reference of scale, the thickness of the aquifer is 100 m. Figure 12. Liquid density after 1 year of injecting 1 Mt of CO2. Dissolved CO2 increases brine density, causing an unstable situation where denser brine is placed on top of lighter brine. As a reference of scale, the thickness of the aquifer is 100 m. Figure 13. Water flux vectors after 100 days of injecting 1 Mt of CO2. As a reference of scale, the thickness of the aquifer is 100 m. (MUSTANG_D055 ah puedo ver las tablas) 5.3.3. Numerical solution a) Elastic models The hydromechanical characterization test is simulated using the fully coupled finite element code CODE_BRIGHT (Olivella et al., 1994, 1996). An ideal homogeneous horizontal aquifercaprock system with the geometrical distribution of the Hontomin pilot test is considered (Figure 5.3). The aquifer has a thickness of 100 m, which we define as the characteristic length of the problem. The top of the aquifer is located at a depth of 1500 m. The aquifer is overlaid by a low-permeability caprock. Several thicknesses of the caprock have been considered: from 50 to 1500 m. The caprock is covered by a low shear strength media, which do not need to be included in the model. The system is axisymmetric and extends laterally up to 20 km. An injection well with a radius of 0.15 m is placed at the centre of the domain and the observation well is placed 50 m away. A structured mesh of quadrilateral elements has been used. The element size grows progressively from the injection well to the outer boundary. As a first step, a steady-state calculation is carried out to ensure equilibrium for the pressure and stress fields prior to injection.

The energy consumption for a CO2 mass flow rate of 1.0 Mt/yr, assuming the surface storage pressure and temperature conditions of the pilot test site of Hontomn (2.0 MPa and -20 C), is 7910 kW for SC CO2 injection, while it is just of 675 kW for liquid CO2 injection. However, a fairer comparison for such a high mass flow rate should consider pressure and temperature values resulting from transport through a long CO2 pipeline. CO2 transport and injection scenarios simulated by Nimtz et al. (2010) showed that CO2 can arrive at the injection site at 8.5 MPa and 12 C. Thus, to get the injection conditions shown in Table 6.3, SC CO2 injection would require a combination of heating and throttling, while cold CO2 injection would require cooling and expansion (see Figure 6.8).

Interestingly, if CO2 is injected at the wellhead conditions proposed by Nimtz et al. (2010) in their application, i.e. 8.5 MPa and 12 C at the end of the pipeline and a mass flow rate of 117.3 kg/s distributed in 60 injection wells (1.95 kg/s in each well), CO2 would reach the aquifer at 17.5 MPa and 35 C. Since the bottom hole pressure is similar to that of the reservoir simulation (see Section 6.4), CO2 could be injected directly from the pipeline without any conditioning operation. Therefore, both at pilot and industrial scales injecting CO2 in liquid phase conditions leads to a much lower energy demand. The results of the simulations presented in this section show that it is possible to inject CO2 in dense liquid-phase by controlling the operational variables, which could lead to a reduction of the reservoir overpressure and, theoretically, of the operational energetic costs. The nature of the outer hydraulic boundary condition does not affect the results because the radius of the pressure perturbation cone is smaller than the radius of the domain for the injection time scales presented here. Therefore, the model behaves as an infinitely acting aquifer. The top of the aquifer is located at a depth of 1500 m, which corresponds to the depth of the reservoir at the Hontomn test site. The seals that overlay and underlie the aquifer have a thickness of 200 m. The seal that is placed on top of the aquifer, i.e. the caprock, is covered by a 1300 m thick low shear strength media that do not need to be included in the model. An injection well with a radius of 0.15 m is placed in the center of the domain. (Thesis_Vilarrasa)

PETREL MANUAL

Simulation
Reservoir simulation is the study of how fluids flow in a hydrocarbon reservoir when put under production conditions. The purpose is usually to predict the behavior of a reservoir to different production scenarios, or to increase the understanding of its geological properties by comparing known behavior to a simulation using different geological representations. See physical principles for further information. Petrel Reservoir Engineering supports the Schlumberger ECLIPSE family of simulators:

ECLIPSE 100 - 3-phase finite difference fully implicit black oil simulator. ECLIPSE 300 - 3-phase finite difference fully implicit black oil and compositional and thermal simulator. FrontSim - 3-phase black oil streamline simulator.

All the simulators read their input from ASCII keyword and binary files, and write binary files as output. Petrel generates these input files, manages different cases, and presents the results in graphical form. See also: ECLIPSE Keywords, ECLIPSE Technical Description and ECLIPSE File Formats FrontSim Keywords and FrontSim Technical Description

Overview of how to set up a simulation case

1. Build a grid and populate it with properties. 2. (Optionally) scale up the structure and properties onto a coarser grid. See Upscaling. 3. Define or import well paths. See Well Design. 4. Define or import well completion events. See Well Completion Design. 5. (Optionally) define unique (<=8 character) simulation name aliases. See Well Attributes and Exporting simulation keywords. 6. (Optionally) import historical production rates. See Import Data and Well production and well event formats. 7. (Optionally) define a well segmentation set. See Well Segmentation. 8. Define a fluid model, describing the properties of the reservoir fluids at varying pressures, volumes and temperatures (PVT) and the initial conditions (pressures and contacts) in the reservoir. See Making a fluid model.

9. Define a saturation function, describing the relative permeability and capillary pressure of the fluids as a function of saturation. See Making rock physics functions. 10. Define a rock compaction function, describing how the rock expands and compresses with changing pressure. See the Make rock physics functions process. 11. Define aquifers, describing the type, size and connections of the acting aquifer(s), using the Make aquifer process. 12. Define a development strategy to control how the wells will produce and inject. See Development Strategies. 13. Define a simulation case, putting all the above data objects together. See Defining a simulation case. 14. Analyze and view the results using the function window and the Results pane. See Simulation Results.

Physical principles
To simulate fluid flow in a reservoir, we first need a geological model of the reservoir. This model will have to be described in such a way that the data can be used for numerical computations. This is done by "gridding" the reservoir, which divides the reservoir into a finite set of homogeneous grid cells. Each of these cells contains data for each of the geological parameters in that cell. Next we need a description of the fluids in the reservoir. This is done by describing the so-called relative permeability functions and the PVT data. Relative permeability functions are a description of how the fluids flow relative to each other. The PVT data represent how pressure and volume of the fluids relate. We also have to describe where the fluid phases are present in the reservoir - this is done by describing where their contacts are sitting. We need to apply a force to the fluids and this is done by setting locations of wells and their production/injection rates. Lastly, we must specify the time period for the simulation and how many time steps we would like to use. The flow of fluids is governed by three basic physical principles:

Mass is neither created nor destroyed, that is Conservation of Mass. The rate of change of momentum of a portion of the fluid equals the force applied to it, or Conservation of Momentum. Energy is neither created nor destroyed, that is Conservation of Energy.

Conservation of Mass
The first principle is modeled by a so called partial differential equation that says that if there is a difference between flow into and out of a tiny volume of space, this will either cause a build-up or a draw down of mass in this tiny volume.

Conservation of Momentum
The second principle is approximated by an experimental law, called Darcy's law, that relate the pressure difference (force) across a porous rock containing a fluid and the resulting velocity (momentum) of this fluid. This experimental law is also represented by a partial differential equation that is combined with the equation based on the first principle to form a set of partial differential equations that is the mathematical description of the flow of fluids through a porous media.

Conservation of Energy
The third principle is approximated by a relationship between Pressure, Volume and Temperature (PVT) for the rock and the fluids. The PVT data is represented by a set of tables for each reservoir.

General information on streamline simulation


FrontSim is a streamline simulator. This numerical method first solves the pressure, then computes the streamlines and thereafter computes the changes in saturation along the streamlines. The equations for saturations are solved as several one dimensional problems along streamlines, as opposed to solving one large three dimensional problem. Consequently, FrontSim can be very fast and hence capable of handling grids with millions of grid cells. For reservoirs where the movement of the fluids is mostly driven by the potential field induced by producing wells, FrontSim is particularly efficient. Examples include:

Water flooding. Highly heterogeneous reservoirs.

Since FrontSim can be run on large models it is also a good tool for:

Screening of large geological models. Validation of up-scaled models.

Streamline Calculation
Given the fluid pressure in each of the cells, we can find the streamlines by selecting a number of so called starting points in the reservoir. Beginning at these starting points, we find the lowest pressure (or alternatively the highest) in close vicinity of the starting point, such as the steepest downhill we can find, and follow this downhill to the next point, etc. This process will eventually lead us to a producing well. When we have gone through this process for all the starting points, the reservoir should be covered by streamlines.

Streamline representation
Since the streamlines describe the direction of flow at the moment in time that they are computed, they can be used as a computational tool to move the fluids along. The streamlines can be seen as a set of tubes representing the total reservoir volume and through which the fluids are moving. The tubes have exactly the same geological properties as the underlying geological model has. This approach divides the fluid flow into a set of one-dimensional models and therefore reduces the problem to a computational simplicity that can have significant advantages. Since each streamline carries about the same volume, the density of the streamlines is an indication of the velocity of the fluid.

Streamline simulation advantages


The sequential process described above (solve pressure - move fluids along streamlines) constitutes one time step. An advantage with this approach is that the time steps can be very long and therefore the computational effort needed can be reduced. The obvious advantage of a streamline method is the possibility it offers to do relatively quick simulations on large geologically and architecturally complex models. This is the main reason that the streamline method is a popular tool for validating upscaling methods, measuring uncertainty, screening and ranking different model scenarios. Another advantage of the streamline method is the visual information it gives about the fluid flow pattern and connectivity in the reservoir. This can be used to study the effects of different well patterns on the flow, and to monitor which volumes, wells and boundaries that support a specific well at any given time.

Streamline Time dependency


Even though a streamline simulation can be significantly faster than standard simulations, it can still be a time consuming affair if we want to model in detail all physical properties available in FrontSim. Some of the most important time dependencies are discussed below.
Number of Cells

The size of the model represented by the number of cells, will impact the time it will take to compute the pressure and move the fluids along. For example, for a model with 1 million grid cells it might take up to several minutes to compute the pressure and move the fluids one time step, depending somewhat on the complexity of the fluid description and the computational speed of the processor of your PC. Also, the memory requirement might be in the order of one gigabyte.
Fluid description

The fluid description will impact the CPU time. The fastest model will be a low compressible two-phase model (oil/water). The low compressibility will cause the pressure solution to be easier to solve and the requirement for time steps for accuracy is

low. In addition, if the mobility and densities of the fluids are close, one time step might still result in a reasonably accurate (also from engineering perspective, that is, standard simulation) simulation result. For a one million-grid cell model, one time step might require 3 minutes of computer time with a state-of-the-art PC.
Time steps

The number of time steps and the length of each time step will influence both the result and the time it will take to achieve a result. As we introduce more detail in the fluid description, we also potentially introduce a need to use more and shorter time steps. Such a factor is gravity segregation. To be accurate, more time steps are required even though FrontSim will include the effect in a one time step solution.
Compressibility

Another factor is the compressibility introduced when using a three-phase oil/water/gas model. To achieve close to engineering accuracy, we will have to enable the gravity segregation and use many time steps. In addition, the non-linearity of the model will cause more computational effort per time step. Note that FrontSim's streamline concept will allow the user to use only one long time step (possibly years), even for this type of model, but the engineering accuracy defined by standard simulation will be degraded. The resulting set of partial differential equations cannot be solved by any analytical means due to its typical complexity in geometry, rock property and fluid description. Instead, a so called numerical approximation is used. Many types of numerical methods are available to solve these equations. Most use some form of a finite difference/volume method that divides the geometry into many small subsections called cells containing rock and fluid properties. These cells cover the whole domain and the fluid flow is represented as relationships between these cells. These relationships result in an equation system with a number of unknowns on the order of number of cells or higher. This equation system will have to be solved for every time step during the simulation. These methods might require very many small time steps, and because of this, a reservoir simulation process can be a very time consuming affair depending on the geological detail represented and fluid property modeled.

4. PetraSim Basics
Work Flow in a Typical Analysis
Many problems will be run in two stages: (1) an analysis that establishes a steady state initial condition (2) an analysis that loads the steady state results as an initial condition and then proceeds with a transient disturbance, such as a spill or production from a reservoir. (3) steps of an analysis. These include: (4) (5) Defining the problem boundaries, creating a conceptual model, and creating a mesh. (6) Selecting the global options to be used in the analysis. (7)

(8)

Defining the default initial conditions for the model, either directly or by loading the results of a previous analysis. (9) Defining cell-specific data, such as material, sources, sinks, and initial conditions. (10) Setting the solution and output options. (11) roblem. (12) Post-processing of results using contour and time history plots.

Potrebbero piacerti anche