Вы находитесь на странице: 1из 68

1

2
It is estimated that more than 60% of the worlds proven hydrocarbon reserves lie
in naturally fractured reservoirs.
It is now well known that natural fractures whether they act as seals or
conduits can strongly affect the tortuosity of hydrocarbon flow
pathways and therefore have a significant economical impact. For
instance during migration from source rock to reservoir rock, knowing
how natural fractures are distributed will provide efficient petroleum
system evaluation. Similarly knowing the distribution of natural
fractures in reservoirs will help optimizing natural reserve recovery
through producing wells.

A large percentage of the worlds oil and natural gas is contained in rocks, such as
carbonates, in which it is often difficult to extract fluids unless the rocks contain natural or
induced fractures.
Although the resources are extensive, the production of fluids from these reservoirs is limited
by our inability to predict the location, orientation, and permeability of the fractures in the
rocks.

3
As a result, the recovery rates, tends to be very low for these reservoirs (5-25%).

If we can detect spatial variations in the fracture distribution and develop methods for estimating the
permeability of these fracture zones using surface geophysical methods, the recovery rates for known
reservoirs could be increased significantly, having a potentially major impact on the worlds oil and gas
supplies.

Since natural fractures are only directly observed along wellbore, the main
question is: How should the natural fractures be distributed between the wells?

3
Naturally Fractured Reservoir (NFR)
Natural fractures normally are generated through diagenesis or
tectonic deformation and affect most reservoirs in some way or
another. They can affect reservoirs in a positive way, such as extra
fluid conduits, or in a negative way, such as barriers to flow or cross-
flow, short-circuiting natural flow paths.
In carbonates, natural fractures typically create secondary
porosity and permeability in low porosity matrix. However,
fractures are the fluid pathways. They also can lead to early water
breakthrough and they can inhibit secondary recovery.
In silisiclastics, they can add some permeability to existing matrix-
dominated production.
Other reservoir types such as basement rocks, volcanic rocks, and
coal-bed methane also are affected by fractures and have become
new potential reservoirs for oil companies in later years.

4
Classification of Naturally Fractured Reservoirs
This classification is based on the contribution of fractures to total reservoir
porosity and permeability. NFRs are reservoirs where naturally occuring
fractures have/are predicted to have a significant effect on reservoir fluid flow
and can change througout the production history:
Increased/decreased permeability and porosity
Increased permeability anisotropy.

Type 1: Fractures provide primary porosity and permeability. These reservoirs


require a large drainage area. They typically are granite or quartize
reservoirs with large fractures. Few wells are needed for
development, but there is a potential for high initial production and
rapid decline (early water breakthrough). Example: Fractured
granites.
Type 2: These reserviors have fairly low matrix porosity and permeability, with
good initial production given by fractures that provide essential
permeability.
Example: Carbonates.
Type 3: These reservoirs already have high matrix porosity, with a fairly good
permeability, but the fractures add an extra permeability component
to productivity.
Example: Sandstones.
Type 4: In compressional settings, folding can occur with differential strain
leading partly to extensional fractures and partly to compressional
ones. In carbonates, stylolites can form and in high porosity
sandstones, deformation bands can occur, all acting as barriers to
flow and some creating compartmentalization. Also, existing open
fractures can be mineralized or filled with clay.

4
Fracture Modeling
Fractures are common in most outcrops. They affect most oil and gas
reservoirs throughout the world.

Natural fractures can introduce a high heterogeneity or a strong


anisotropy in reservoirs and can affect fluid flow or mechanical
stability. An accurate description and characterization of the fracture
network is, therefore, of very high importance for all stages of
reservoir management: drilling, well placement, stimulation,
completion, and production profile design.

There is an interrelationship between fractures, fluid-flow, and effective


stress state in the reservoir. Decisions must be made early to:
identify and characterize fractures (already at the exploration
stage) and evaluate anisotropy at the development stage for better

5
infill drilling campaigns.
assess how fractures (fracture network) affect fluid flow
assess what the recovery factor is in a dual-porosity type reservoir
assess what the effect of change in stress is with production.

The purpose of modeling fractures is to create simulation properties


with the power to predict the reservoir behavior. By modeling the
fractures explicitly, you can honor the spatial relationships between
properties in adjacent cells.

Fracture modeling is a multi-step process that involves several disciplines


within reservoir characterization and simulation. The main idea is to build on
geological concepts and gather data such as interpretation of beds, faults, and
fractures from image log data. For example, you can use field outcrop studies
as analogs for conceptual models and seismic attributes as fracture drivers.

After you gather the data, the next step is to transfer it into a description of
fracture intensity that can be used to populate a 3D geological framework
model. Depending on the analysis of the fracture data, multiple sets of
fractures can be identified. These sets can be the result of different tectonic
events such as over-trusts and extensional faults, conjugate fractures related
to bending or flexure of geological layers, or simply related to differences in
lithology.

After you have identified, analyzed, and categorized the data, you can build
the fracture model itself. From the initial intensity description, you must
populate the fracture intensity in the 3D grid stochastically or deterministically.
If you populate it deterministically, you must have a very good idea of where
the fractures are and how they behave in the 3D grid. To get this information,
you can use high confidence fault patches from the seismic volume attribute

5
process called Ant tracking. If no such data exists, a stochastic method should
be used.

The ultimate goal is to find 3D grid properties that describe permeability and
porosity for fractures as well as the standard permeability and porosity you get
from the matrix. Why do you need this? Because many types of reservoirs are
dual-porosity and possibly dual-permeability reservoirs. They are either
naturally fractured (NFR) or consist of carbonates that are vugular or heavily
fractured due to tectonic processes.

Some of these reservoir rocks are originally dense and have little flow or
storage capacity in the matrix, but when fractured, certain areas can become
high flow zones. To model these reservoir rocks correctly in a simulator is
tricky and, at best, quite inaccurate. To resolve this problem, a Discrete
Fracture Network (DFN) model based on intensity can be built. Upscaling
properties based on a DFN model will generate a second set of properties of
permeability and porosity and a sigma factor that describes the connectivity.
This sigma/connectivity is essential for connecting duplicate cells in a
simulator to describe the matrix and fracture porosities and permeabilities.

Additional Notes
Fractures can increase effective porosity and permeability and introduce
permeability anisotropy, particularly in rocks with low matrix permeability
(Rice, 1983; Nelson, 1985; Fassett, 1991; Teufel and Farrell, 1992). Faults
also can function as fluid migration pathways, barriers, or a combination of
both (for example, Caine et al., 1996).

For modeling and production purposes, it is important to document directions


of preferred fracture and fault orientations within primary hydrocarbon traps
such as anticlines. By understanding controls on fracture and fault orientation

5
and distribution in a given reservoir, the accuracy of flow modeling can be
improved, which can increase primary and secondary hydrocarbon recovery.

Murray (1968) noted that the relationships between bed thickness, structural
curvature and fracture porosity, and permeability can be effective in evaluating
geologic structures as hydrocarbon reservoirs. Rocks, in general, exhibit
increased fracture density with increased deformation (Nelson, 1985). Based
on these assumptions in the literature, it is important to look at fractures.

What data do you have to model fractures?


Well data including picks from image logs and dip and azimuth of
structures/fractures along the wellbore.
Seismic data.
Structural surface data, which can be converted to curvature data.

From FRED Help Documentation


Improving the recovery from fractured reservoirs is an increasingly important
focus for many oil companies. The recovery from reservoirs, where fractures
dominate permeability, is often a fraction of the resource recovered from
conventional reservoirs in which matrix permeability dominates. The lower
recovery and higher risks relate to the difficulty in forecasting how various
completion placements, gel treatments, surfactant injection, and tertiary
recovery processes will actually perform. A reduction in risk and an
improvement in understanding of reservoir behavior will lead to enhanced
profitability from under-exploited fractured fields.

Why are fractured reservoirs so problematic? Fractures do more than simply


increase reservoir permeability. Fractures fundamentally alter reservoir
connectivity and heterogeneity. If this was not the case, fractured reservoirs

5
could be modeled with the same level of confidence and success as matrix-
dominated reservoirs. More successful exploitation of fractured reservoirs has
been hindered by the lack of reservoir management tools that incorporate the
unique flow behavior of reservoir fracture systems.

The primary modeling tool currently used by the industry is based on the dual-
porosity continuum approach. DFN models offer numerous advantages over
conventional dual-porosity models, including:
More realistic representation of geology and fracture network geometry.
Direct simulation of reservoir connectivity, and consistent, comprehensive
integration of geological, geophysical, production, and well test data.

5
Fracture Modeling in Petrel: General Workflow
In Petrel, fracture modeling consists of two main processes:
Create the fracture network.
Scale up fracture network properties.
The heart of the latter process is upscaling of the fracture network.
Schlumberger is cooperating with Golder Associates and
implementing upscaling algorithms developed and used by them.
Petrel is now utilizing algorithms from this industry leading software
suite to provide a complete fracture modeling workflow from seismic to
simulation.

However, other standard processes in Petrel are used prior to and


during fracture modeling. These processes include geometrical
modeling and petrophysical modeling. Processes related to simulation
are used after the fracture model and properties are made.

6
Fracture modeling workflows are varied and often customized for the
conditions and data available in a particular field. The workflow in Petrel is
designed to be flexible, giving you the power to use any data that might be
available. The fracture modeling approach in Petrel is a fully transparent
integration from static data to flow modeling.

Simulation
Petrel also includes integrated handling of ECLIPSE simulation. Matrix and
fracture properties are combined correctly, keywords are written for the
appropriate simulation, and the results are presented.
Single-porosity/permeability: Matrix contribution to both permeability and
storage is negligible.
Dual-porosity: Fractures provide permeability and matrix provides storage.
Dual-porosity/dual-permeability: Both matrix and fractures provide
significant contribution to the permeability of the system.

6
7
8
Over the years many techniques have been developed to detect and
model natural fractures in the subsurface
- Finally the last set of methods are the geomechanical methods
that we will developed in this talk

9
10
As an introduction to geomechanical methods for modeling natural
fractures, let me show you how fracture development can be affected
by local heterogeneous stress field.
Here we want to simulate how stresses is be perturbed around a
slipping fault. This 2D model of a square cut by an oblique
discontinuity is subjected to a vertical load.
When the discontinuity slips to create a fault, the stress is perturbed
showing tensile areas in red and compressive areas in bleu. The
orientation of S1 (the most compressive principal stress) is reoriented
close to the fault.
This stress pattern can be compared to outcrop example such as
reactivated preexisting fracture showing branch crack in tensile areas
and stylolites in compressive areas, both following the orientation of
S1 and S2 respectively, according to the conceptual model of Mutlu & Pollard in
2008
On the right we see how joints are affected by nearby faults in a
carbonate outcrop at Nash Point, UK. There is a clear relationship

11
between strike slip faults (in red) and joints (in black). Joints seems to have
developed within the perturbed stress field caused by slip along faults.

11
12
13
14
As a summary, we can say that natural fracture characteristics, kind
(joints, faults, stylolites, deformation bands, etc.), orientation and
location, at the time of their development, depend on the local state of
stress and rock type.

The goal is therefore to estimate the local state of stress at a given


geological time using 3 key elements:
1. The objects that perturb the stresses. As we have seen earlier
these objects can be faults, folds, salt diapir but also cavities,
sliding layer interfaces etc. Unfortunately only the present day
geometry is available.
2. The tectonic loading: This can be the local or regional paleo-stress
regimes, which is always difficult to estimate and often badly
guessed.
3. The rock type: This includes the rock physical properties and
behavior, which is often well estimated from laboratory testing.

15
Based on the previous description, we propose the following
geomechanical solution that also must respond to the following
constraints:
- Uncertainty: We can reduce uncertainty in estimating the
paleo-tectonic stresses
- Data: We can efficiently use collected hard data. The data
include observed fractures along wells, tiltmeters, seismic
data, GPS, outcrop observation, etc.
- Speed: We can approach near real time computation.
Computation should not last more than 1 minute.
These are achieved with the iBem3D technology.
- Accessibility: We make sure any geologist/engineer can
apply the process without being expert in geomechanics in
an integrated plateform: Petrel

16
The main principle for Paleo stress inversion is as follow:
We run thousands of simulations to cover the range of all
possible tectonic stress configurations.
Then for each simulation we compare the modeled stresses
with the observed fractures.
We then analyze the results to find the best fit.

However, the main problems to be solved are:

1. The dimension of parameter space that


needs to be reduced.
2. The computation that time must be
greatly reduced as thousands of
simulations will take weeks.
3. Finally, result analysis should be
automatized and fast.

17
18
19
20
21
22
23
24
Classify Fractures
The classification is based on the dip angle. There are 6 principal
geological fractures types grouped into 3 mechanical fracture styles
and identified with a code from 0 to 6.

25
Let s see an example of how stress inversion works.

We need fault model , and fracture data, with a strong key factor, the
fracture type
Many technologies and tools exist to perform geomechanical
modeling. Here we concentrate on Poly3D, a geomechanical tool
based on the 3D boundary element method

As opposed to finite element method, only the discontinuities are


modeled here (i.e. faults) as triangulated surfaces ;

the main benefits of such solution are:


- User friendly: Model are built in minutes
- Computation is fast several seconds to few minutes
- Real complex fault geometry can be modeled
- Technology based on fracture mechanics
- it uses Elasticity as first approximation
- Large model comprised of 100 of faults can be run

27
The main principle for Paleo stress inversion is as follow:
We run thousands of simulations to cover the range of all possible tectonic stress
configurations.
Then for each simulation we compare the modeled stresses with the observed fractures.
We then analyze the results to find the best fit.

28
The main principle for Paleo stress inversion is as follow:
We run thousands of simulations to cover the range of all possible tectonic stress
configurations.
Then for each simulation we compare the modeled stresses with the observed fractures.
We then analyze the results to find the best fit.

29
The main principle for Paleo stress inversion is as follow:
We run thousands of simulations to cover the range of all possible tectonic stress
configurations.
Then for each simulation we compare the modeled stresses with the observed fractures.
We then analyze the results to find the best fit.

30
The main principle for Paleo stress inversion is as follow:
We run thousands of simulations to cover the range of all possible tectonic stress
configurations.
Then for each simulation we compare the modeled stresses with the observed fractures.
We then analyze the results to find the best fit.

31
The main principle for Paleo stress inversion is as follow:
We run thousands of simulations to cover the range of all possible tectonic stress
configurations.
Then for each simulation we compare the modeled stresses with the observed fractures.
We then analyze the results to find the best fit.

32
The main principle for Paleo stress inversion is as follow:
We run thousands of simulations to cover the range of all possible tectonic stress
configurations.
Then for each simulation we compare the modeled stresses with the observed fractures.
We then analyze the results to find the best fit.

33
The stress inversion by iterative computation and optimization of the simulation leads to
propose a DFN that considerably reduced the uncertainty of the fracture distribution between
the wells

Now, lets see how these results can be applied to a real study case in the Malay Basin
35
The first case study is an outcrop example from Nash Point, located in
the Bristol Channel in UK.
It is a limestone layer where faults and joints are nicely exposed.

We clearly see that the joints in black are affected by the main strike
slip fault in red. Joint strikes are curved and vary from one fault block
to the other.

The joints were formed during the Late Cretaceous to early Tertiary Alpine
compression N170 according to field observations.

36
For the simulation we have built a 3D model of the 20 strike slip faults
such that they follow the trace seen in outcrop, they are sub vertical
with a height of about 40 m.

Instead of using all measured joints we have decided to simulate 3


wells thus only 3 data points that represent the joint dip angle and dip
azimuth at the well locations.

37
The result shows that:

In 1mn 15s and 50000 runs the paleostress has been recovered and perturbed stress
computed everywhere.
Theta is N161, the stress ratio R is 0.75 for a strike slip fault regime. There is one single
tectonic event responsible for the joint development.

The map on the left shows the computed joint strike that we can directly compare with the
observed joint pattern on the right.
The main trends are nicely reproduced except in specify areas.

39
This second study area is located 15 km North of

Montpellier, France, close to the village called Les Matelles.

The exposure, located in a dry flat-bottomed valley (longitude

348E, latitude 4343N), is well suited for the study of brittle

tectonics in limestones and stress perturbations around meso-

scale faults (Rispoli, 1981; Petit & Mattauer, 1995; Soliva et al.,

2010). Indeed, the exposure displays a large number of

features such as faults, stylolites, joints and veins filled with

40
calcites.

40
This second study area is located 15 km North of

Montpellier, France, close to the village called Les Matelles.

The exposure, located in a dry flat-bottomed valley (longitude

348E, latitude 4343N), is well suited for the study of brittle

tectonics in limestones and stress perturbations around meso-

scale faults (Rispoli, 1981; Petit & Mattauer, 1995; Soliva et al.,

2010). Indeed, the exposure displays a large number of

features such as faults, stylolites, joints and veins filled with

41
calcites.

41
This second study area is located 15 km North of

Montpellier, France, close to the village called Les Matelles.

The exposure, located in a dry flat-bottomed valley (longitude

348E, latitude 4343N), is well suited for the study of brittle

tectonics in limestones and stress perturbations around meso-

scale faults (Rispoli, 1981; Petit & Mattauer, 1995; Soliva et al.,

2010). Indeed, the exposure displays a large number of

features such as faults, stylolites, joints and veins filled with

42
calcites.

42
This second study area is located 15 km North of

Montpellier, France, close to the village called Les Matelles.

The exposure, located in a dry flat-bottomed valley (longitude

348E, latitude 4343N), is well suited for the study of brittle

tectonics in limestones and stress perturbations around meso-

scale faults (Rispoli, 1981; Petit & Mattauer, 1995; Soliva et al.,

2010). Indeed, the exposure displays a large number of

features such as faults, stylolites, joints and veins filled with

43
calcites.

43
The second application of natural fractures prediction can be applied
on geological traps, where two parameteres can be important : top
and fault seal. In this talk , we will also concentrate on naturally
fractured top seal.

44
It is now well known that natural fractures whether they act as seals or
conduits can strongly affect hydrocarbon flow pathways and therefore
have a significant economical impact.

knowing the distribution of natural fractures in top seal rock will help
optimize reservoir recovery through producing wells.

Since natural fractures are only directly observed along wellbore, the
main question is: How should the natural fractures be distributed
between these wells? And which fractures are actually activated ?

Thats the question weve been trying to solve using ibem 3d in order
to use the results as new drivers for DFM

45
We already know that natural fractures are important for top seal
integrity as an important vector of possible leakage. But all fractures
arent necesserly simultaneously open.
(supposing that inactivated fractures are closed by chemical and
confining process) Knowing the active fractures is then a critical input
to evaluate top seal integrity.

The workflow we developed gives a good picture of fractures


reactivation by combining paleo fractures and present day stress.

It consists on using, Well fracturation data (joints, faults breakouts )


to determine a fracture orientation map, then combine it with a Present
day stress map to create a Fractures Reactivation Map.

46
So here is the model we used for our case study.

We can see the principal faults.

An observation grid (which is an horizon)

And 4 wells where we plotted joint data.

These joints have been created during a paleo tectonic phase where all faults where active.

We can see that joint data is not homogeneous from a well to another. This means that the
stress is perturbed by faults. Especially in well 4 who crosses the biggest fault

With this available data and fault geometry, it


is possible to do an inversion
to find the paleo remote stress that generated the
paleofractures.

47
The main principle for Paleo stress inversion is as follow:
We run thousands of simulations to cover the range of all possible
tectonic stress configurations.
Then for each simulation we compare the modeled stresses with
the observed fractures.
We then analyze the results to find the best fit.

The best fit has been found for a Normal regime, with a maximum
horizontal stress oriented at North 44.5 and a stress ratio at 0.11

48
Then, on each point of the observation grid (horizon) we can compute
the perturbed stress tensor and thus the Sigma H orientation which
gives us the joints orientation trends.

We can see the observed well data strike (at the left) in purple color,
compared to the computed paleo joints directions in black. These
results confirm that the minimum cost stress solution have a good
calibration with observed data, and therefore the fracture orientation
trends can be used with good confidence.

We can also compute elastic/plastic attributes such as the maximum


coulomb shear stress illustrated in this picture (purple and blue for
greater shear potential)

49
So here is the same model.

We can see the principal faults, the observation grid (which is an


horizon), and 3 wells where we plotted breakouts data.

These breakouts have been created during a present day tectonic


phase where the active faults are unknown.

We can see that joint data is not homogeneous from a well to another.
This means that the stress is perturbed by faults. Especially in well 4
who crosses the biggest fault

With this available data and fault geometry, it is possible to do an


inversion to find the paleo remote stress that generated the paleofractures.

50
51
The best fit has been found for the fault Cluster 2. with a Normal
regime, with a maximum horizontal stress oriented at North 90
and a stress ratio at 0.11

52
The main principle for Paleo stress inversion is as follow:
We run thousands of simulations to cover the range of all possible
tectonic stress configurations.
Then for each simulation we compare the modeled stresses with
the observed fractures.
We then analyze the results to find the best fit.

The best fit has been found for the fault Cluster 2. with a Normal
regime, with a maximum horizontal stress oriented at North 90
and a stress ratio at 0.11

53
Then, on each point of the observation grid (horizon) we can compute
the presend day perturbed stress tensor and thus the Sigma H
orientation which gives us breakouts orientation trends.

We can see the observed well breakouts data strike (at the left) in
purple color, compared to the computed paleo breakout directions in
black. These results confirm that the minimum cost stress solution
have a good calibration with observed data, and therefore the present
day stress orientation trends can be used with good confidence.

54
55
This map can represent top seal integrity and therefore a prediction for
leakage risk, shown in red for high risk and blue for low leakage risk.

56

Вам также может понравиться