Sie sind auf Seite 1von 15

UNIVERSITY OF TRINIDAD AND TOBAGO

RESERVOIR SIMULATION
RESR 6008
ASSIGNMENT #2

PREPARED FOR
DR. DAVID ALEXANDER

PREPARED BY
MARVIN FLATTS 54481
MUSTAPHA RAJACK 104000097

October 2, 2019
The objective of this report is to investigate and articulate the concepts of the Sensitivity
Analysis, Laplace Transform and Streamline Simulation and their applications to reservoir
simulation and reservoir studies. What is reservoir simulation? Reservoir simulation is the art
of combining physics, mathematics, reservoir engineering and computer programming to
develop a tool for predicting hydrocarbon reservoir performance under various operating
strategies.

Sensitivity Analysis

Sensitivity analysis is an analytical tool which is used to determine how the manipulation of
an independent variable affects a given dependent variable under specified set of conditions.
Sensitivity analyses usually take an in depth look at the way in which various sources of
uncertainty within a mathematical model will contribute to the model’s overall uncertainty.

In reservoir simulation, sensitivity analyses are extremely useful and crucial to organizing,
manipulating, processing and displaying of the data which is generated by a reservoir
simulation model. The sensitivity analysis has become a mainstay in the determining of
which rock and fluid parameters are relevant to the derivation of a solution for the specified
problem. Sensitivity studies have also proven to be excellent at providing estimates for
various types of data that may be troublesome to acquire. An example of such data can be the
practicality of obtaining measured values of vertical permeability to use in cross-sectional or
3D models. Sensitivity analyses are almost unlimited in its analytical prowess as it has been
used to determine the potential error due to the inaccuracy inherent in the data. For example,
for Kv/Kh ratios greater than 0.1, calculated values are insensitive to the value of the ratio
because of the relatively large area and short distances for vertical flow.

The application of sensitivity analysis to reservoir simulation has significantly improved the
efficiency and it has increased the speed of the process of reducing the randomness inherent
in reservoir simulation. Sensitivity analysis has contributed to the overall improvement and
accuracy of the reservoir modelling process.
Figure 1 – Sensitivity analysis showing the effect of Polymer concentration on fractional flow
within a reservoir
Laplace Transforms

The finite-difference techniques used in reservoir simulation result in many independent


calculations, which are amenable to parallel execution. Parallelism in simulators is limited by
the time-stepping approach of the finite-difference method. The Laplace transform was
applied to the simulation of a one-dimensional oil-water model with the aim of reducing
nonlinearities and increasing the allowable time step size. Since parallel execution could be
performed during each time step, an increase in the time step size would increase the areas
that were suitable for parallel execution.
In an investigation, a modified IMPES equation was discretized with respect to space.
Subsequently, the resulting equation was transformed with respect to time. The pressures and
saturations of the model behaved more linearly in the transformed space than in the real time
space, but the allowable time step size did not show improvement in Laplace space.
However, a favourable result of this study was that, during breakthrough, the transform
method needed fewer iterations than the real time simulation. For example, for the maximum
stable time step, a ratio of 4.55 iterations in real time space to one iteration in Laplace space
was obtained. The decrease in the number of iterations indicated that the sequential part of
the simulator has been reduced. As a preliminary study, the results of the transform method
were considered sufficiently successful to merit further investigations.
An attempt was made to reduce the sequential part of a reservoir simulator by applying the
Laplace transformation. If a reservoir equation is less nonlinear in the transformed space than
in real time space, then the interval spanned by the simplified linear equation may translate
into a bigger time interval in real time space than the finite-difference approximation method.
The underlying reason is that the finite difference method approximates the nonlinear real
time equation by a straight line.
The objective of the investigation was to see if the nonlinear effects can be reduced by
Laplace transformation. If nonlinearity were to be reduce, larger time steps, faster
convergence and greater potential for parallelization should be obtainable.
The Laplace transform method was developed based on approximating the transform of an
equation by a simplified linear equation. The linear simplified transformed equation is fitted
to a portion of the transform of the nonlinear equation. Therefore, a series of simplified
transforms represents the transform of the nonlinear equation. In this way, the procedure
approximates the discrete spline method that fits polynomials to a function
The choice of the transformed space is dependent on the type of equations. In this
investigation, the Laplace space was chosen because reservoir behaviour is described by
equations that involve first order derivatives with respect to time. Although both linear and
nonlinear equations may be solved analytically in a transformed space a numerical approach
was used since numerical methods are more generally applicable . In addition, numerical
transformation and inversion methods for some transforms, such as Laplace and Fourier, are
available.
Simulation involves the repeated calculation of many variables. Hence, errors may be
compounded with each iteration. In addition to the truncation and differencing errors that are
present in red time simulation, Laplace space simulation includes the numerical errors
involved in inverting transforms by the Stehfest algorithm.
For the purpose of comparison, a real time simulator and a transformed simulator were
programmed for two cases, an incompressible oil-water model and a slightly compressible
oil-water model. Equations for the real time simulators were derived from material balance
considerations. The general equations that describe a two-phase black oil model are:

where is the transmissivity of phase 1,n is the non-wetting phase and u,

is
the wetting phase.
The following assumptions apply to both models:
1. One-dimensional flow in the x direction only,
2. The flow is horizontal. Therefore, no gravitational effects are present and z = 0,
and
3. No capillary effects are considered. Therefore, po = pw = p.

The final equation for the incompressible case is:

are the total transmissivities. Similarly, the slightly compressible case


can be derived from Equations 1 and 2. The equations that result from applying the previous
assumptions are:

A single equation in terms of the pressure and water saturation is obtained by combining
Equations 6 - 8 to give an IMPES formulation.

Derivation of the transform method

The algorithm was initially tested on a time independent incompressible system. The
transform method was derived by taking the Laplace transform of Equation 5 with respect to
time. The pressures are assumed to be continuous for the time domain t = 0 to t = oo.
Therefore, the transformation can be written as:

Assuming constant flow rate q and replacing the transformed pressures in Equation 11,
the equation that generates the Jacobian matrix is as follows:
In the case of the slightly compressible model, the transform method had to be modified to
achieve convergence. At a given time, the pressures are approximated by functions that are
nonzero between t = tl to t = tZ, and zero everywhere else. Combing equations 6 and 7
gives:

The step change approach is applied to pressure pi and the derivative dp/dt to give:

On transforming Equation 14, the final equation is obtained:


Results of Simulation

In the transform method, the Laplace transforms of the block pressures were solved
implicitly, and the Stehfest algorithm was used to invert the transforms. The saturations were
solved explicitly in real time space. In the case of the slightly compressible model, an attempt
was made to solve the Laplace transforms of the saturations explicitly. However, this method
did not converge. The model used to test the transform method was one-dimensional with a
constant water injection of 0.1 cc/sec at one end, and a production well at a constant pressure
of ten atmospheres at the other. The block pressure and water saturation of the injection block
and the water saturation and flow rate of the production block from the Laplace space
simulation were compared to the values that were obtained from a conventional real time
simulation of the same problem. Zn addition, comparisons were made between the
transforms of these variables in order to study the relationship between Laplace space
simulation and real time space simulation for the case of the slightly compressible system.
The effects of time step size, convergence criteria and rate of convergence were considered.

Effects of time step size

Both real time simulation and Laplace space simulation approaches were tested for the effects
of time step size. The size of the time step was kept constant during a test. Different sizes that
ranged from five seconds to one hundred seconds were tested. Comparisons between the
results of the two simulation methods showed good agreement between the values from the
simulations in Laplace space and real time space when the time step size was smaller than 30
seconds. The pressure of the injection block showed a significant deviation in the first time
step for the time step size of 50 seconds and larger time step sizes. The maximum deviation
was 30 percent when the size of the time step was 50 seconds.
Conclusions

1. In general, for a slightly compressible oil-water system, a good fit between real time
and Laplace space simulation results was obtained. However, the time step size was
not improved. The overall number of iterations was reduced by this method, and the
number of iterations at breakthrough was greatly reduced. The overall number of
iterations may be improved if the early time transformation can be approximated
better.

2. The pressures, saturations and flowrates from the real time and transformed
approaches agreed closely. Therefore,' for the systems studied, the convergence and
stability of the transform method were expected to be affected by the same factors as
in real time space simulation.

3. The application of the transform method to an incompressible model also yielded


good fit for the reservoir pressures. However, the number of terms in the Stehfest
algorithm had to be no more than six for the saturations to give a reasonable fit.

4. The Romboutsos algorithm was useful to see the behavior of a function in the Laplace
space. This allowed screening for systems that could benefit from simulation in the
Laplace space.

Streamline Simulation

Streamline based simulation at its core is unique in its approach as this method breaks up the
reservoir into 1D systems or tubes which is different from the conventional finite difference
simulator which moves the fluid from cell to cell. After the 1D systems are developed, the
transport equations are solved down the path of the 1D space defined by the streamlines using
the concept of time of flight (TOF). This manoeuvre enables the fluids to be transported far
more efficiently as the transport problem is separated from the original 3D geological model.
The benefits of decoupling the transport problem from the 3D geological model are bigger
time steps can be used, the problem of numerical diffusion is reduced, and the CPU time
varies almost linearly with the model size.

Streamline simulation has its foundation on five key principles which are:
1. Tracing streamlines in a velocity field.
2. Writing the mass conservation equations in terms of TOF (concept of time of flight).
3. Numerical solution of conservation equations along streamlines.
4. Periodic updating of the streamlines.
5. Operator splitting to account for gravity.

Because of increased speed and accuracy, 3D streamline simulation can model a wide range
of reservoirs. Large reservoirs with hundreds or thousands of wells, several hundred thousand
grid blocks, and an extensive production history are a challenge for finite-difference
simulation. The size and complexity of these reservoirs generally have limited simulation to
sections or well patterns. The streamline technique enables simulation of these reservoirs by
reducing the 3D domain to a series of 1D streamlines, along which the fluid-flow
computations are performed, offering computational benefits orders of magnitude greater.
Maintaining the sharp flood fronts from the displacement processes and reducing grid-
orientation effects increases accuracy. The streamline-simulation results have additional
value as a reservoir management tool when used with traditional reservoir-engineering
techniques, such as standard finite difference simulators. Streamline and stream tube methods
solve the complex fluid-flow models numerically for multiphase flow in porous media with a
reasonable balance between the computational efficiency and the physics modelled. As
computers became more powerful, attention turned toward developing simulators that use
finite-difference methods, including more physical effects. However, computer models of
reservoirs have grown in complexity and geological models with tens of millions of grid cells
can be constructed.
Conventional finite-difference methods have two faults: numerical smearing and low
computational efficiency for models with a large number of grid cells. For models with a
large number of wells, the number of cells required to achieve acceptable accuracy between
wells can be prohibitive. Also, accurate modelling of geological heterogeneities can require a
very large number of cells.
A finite-difference method based on an implicit-pressure/explicit-saturation (IMPES)
approach suffers from the time-step-length-limiting Courant- Friedrichs-Lewy condition. As
the number of cells grows, the maximum time-step length gets shorter for a given model. For
a very large number of cells, the shortness of the time-step can render the required CPU time
impractical for a simulation. A fully implicit approach provides stable solutions. The time-
step length is limited only by nonlinearities; however, these can be strong and, in practice,
keep the time-step length relatively short. A disadvantage of the fully implicit approach is the
tendency to smear the solution (numerical dispersion) even more than the IMPES approach.
Streamline Methods
Numerical approximations of the fluid-phase conservation equations are based on either a
finite-difference IMPES or fully implicit formulation. The IMPES formulation uses operator
splitting to separate one equation for the pressure such that the pressure solution can be found
separately from the equations that model fluid movement. This approach recognizes the
different nature of these physical processes and the potential advantage of using different
numerical approaches for simulation. The standard finite-difference approach is to explicitly
solve the saturation equations. However, the streamline method solves these equations along
streamlines.
Figure 1 shows the basic algorithm for streamline simulation. A 1D front-tracking method
was used to model the convection flow and gravity segregation. The numerical
approximation behind this method is a fractional-flow function, which is approximated by a
continuous piecewise linear function and the initial data by a step function.

Figure 1 – Flow diagram for


Streamline Simulation N =
No. of loops per time-steps

Reservoir Management
The streamline method
detailed in the full-length
paper has potentially
dramatic computational
efficiency advantages.
Also, front tracking can
enable very accurate
tracking of fluid fronts.
This feature can reduce the
number of cells used to obtain
accurate flow approximation
between wells in reservoirs with a large number of wells. This approach was demonstrated on
a history-matching project for the Prudhoe Bay field in Alaska when a comprehensive
reservoir analysis was required to quantify and predict flood fluid movement. The streamline
simulator provided good history matches on a well-by-well basis for more than 500 wells and
was accomplished in less than 6 months by one engineer. Streamline models assist daily
reservoir- management decisions, such as dynamic injection-pattern allocation, well
workovers, and well planning. The same technique can be used to track the movement of
small slugs through the reservoir. Use of finite-difference methods for this type of problem
tends to smear the slug quickly. The proposed method, combined with advanced tracer logic,
was used successfully as a management-decision tool for a large reservoir. Geological-
modelling tools have become more sophisticated and can produce reservoir models
containing a very large number of grid cells. The need for a fast screening tool is apparent.
Because of the ability to run relatively large models at realistic CPU times, streamline
methods are used for geological-ranking exercises. Scale-up of the very fine geological
model might be appropriate, even with the use of streamline methods. The goal is not to
simulate the greatest number of cells, but to simulate at the most appropriate scale as fast as
possible. This goal could mean simulating at the very fine scale because it saves the effort of
scale-up. However, handling very large data sets may not be feasible and a systematic
geological approach is required because local edits may be inconsistent. After an appropriate
scale-up method is selected, running the new model through the streamline simulator can
validate it. Once a satisfactory scale-up is achieved, the finite-difference model can be used to
optimize reservoir performance with confidence that the geological features have been
captured effectively. Use of a gradient method for optimization might be instructive and
reduce the number of simulation runs required to obtain parameter sensitivities.
Fig. 2 is a schematic of a possible geoengineering process.
The geophysical/geological model contains both coarse and very fine scale information from
seismic, well logs, geostatistical-variation, pressure, and production-analysis data. The
workflow of an integrated team can deliver many models covering all known uncertainty and
risk away from the wellbore. The initial models can be run through a streamline simulator to
increase the knowledge of how the geological model behaves dynamically. The initial runs
will not result in a perfect history match, so an iterative process is required until a range of
models that cover the range of uncertainty is produced. Again, no one model will contain the
perfect solution. However, selected models should be capable of describing the likely range
of possible solutions. The ability to switch among tools like a streamline simulator, a finite-
difference simulator, and an assisted history-matching tool brings streamline simulation to
mainstream reservoir simulation.
In conclusion, Streamline simulation can be incorporated at all stages of reservoir
characterization workflow. It can be used with conventional finite-difference techniques to
assist history matching and to validate scale-up methods. Streamlines can provide valuable
insight into the fluid-flow dynamics in the reservoir with the streamline displays, well-
drainage regions, and dynamic injection-pattern-allocation factors. Case studies highlighted
successful use of streamlines in a variety of reservoirs. It is instructive to consider streamline
methods routinely in the reservoir-modelling workflow to check whether the potential speed
and accuracy benefits can be realized.

REFERENCES

1. Ablowitz, M., Kaup, D. J., Newell, A. C., and Segur, H.: "The Inverse Scattering
ransform-Fourier Analysis for Nonlinear Problems," Studies in Applied Mathematics
(1974).

2. Aziz, K. and Settari, A.: Petroleum Reservoir Simulation, Applied Science


Publishers (1979).

3. Barua, J.: A study on Newton Related NonLinear Methods in Well Test Analysis,
Production schedule Optimization and Reservoir Simulation, PhD dissertation,
Stanford University (1989).

4. Pacheco, T.B., Carvalho da Silva, A. F., Maliska, C.R., 2016. Comparison of


IMPES, Sequential and Fully Implicit Formulations for Two Phase Flow In
Porous Media with The Element Based Finite Volume Method,
http://www.sinmec.ufsc.br/site/arquivos/n-nfgthfammj_24353-72586-1-sm.pdf

Das könnte Ihnen auch gefallen