Workflow and Analysis Tools for the Characterization of Fractured Reservoirs*

Arnaud G. Lange1, Andre Fourno1, Matthieu Delorme1, Nina Khvoenkova1, and Catherine Ponsot-Jacquin1
Search and Discovery Article #40424 (2009)
Posted June 10, 2009

*Adapted from oral presentation at AAPG International Conference and Exhibition, Cape Town, South Africa, October 26-29, 2008
1

Simulation of flows and transfers in porous media, IFP, Rueil-Malmaison, France (mailto:[email protected])

Abstract
The workflow developed at IFP for characterizing fractured reservoirs is based on: (i) the construction of geologically-realistic models
of the fracture network; (ii) the characterization of fracture properties from available field data; (iii) the up-scaling of the fracture
properties; (iv) the selection of a suitable up-scaled model usable for field-scale simulations of multi-phase production methods. This
fractured reservoir workflow is reviewed, and a software platform is presented, on which methodologies and tools were developed in
order to perform each step of the workflow.
A geologically-realistic model is presented on which constrained modeling of the geological fracture network based on the analysis of
fracture information acquired in wells and derived from seismic data has been performed. Then optimization algorithms and a 3D
discrete fracture network flow simulator are used in order to automatically characterize fracture properties that are consistent with
transmissivities data, flowmeters and/or well tests data. The characterized fracture properties are the mean length, mean conductivity,
orientation dispersion factors, and facies-dependent properties such as the average spacing and the bed-crossing probability. The
effectiveness of the optimization algorithms to characterize physically meaningful and data-consistent fracture properties is discussed.
Finally, full-field upscaling of the fracture properties has been performed such that a single or dual-porosity simulation model can be
used at field scale, taking into account the multi-scale fracture properties.
This consistent workflow allows flow simulation models to remain interpretable in geological terms, therefore facilitating subsequent
model updating. Moreover, specialists in geosciences and reservoir engineers can cooperate in a very effective way to improve the
management of fractured reservoirs.

Copyright AAPG. Serial rights given by author. For all other rights contact author directly.

Notes by Presenter: The methodology developed at IFP in the past years consists in:
- First, integrating all available data about fracture properties - from seismic measurements, well data the geocellular facies model,
- in order to build two models:
- a discrete model of large-scale fractures and faults, mainly derived from the analysis of seismic attributes,
- at a lower scale, a geocellular model of fracture properties that is consistent with the facies geomodel of the reservoir, giving the
fracture density and orientation per fracture set per facies,
These models provide the necessary information for generating local discrete fracture network models anywhere in the reservoir.
These local models are then used for assessing the flow properties of the fracture network via our 3D DFN flow model.
Interference, well tests and PLT can be simulated and calibrated with available dynamic field data in order to:
- validate the geological DFN model geometry.
- characterize the fracture conductivity for each set.
Then the flow properties may be upscaled to an appropriate field-scale flow simulation model to optimise reservoir production.
Therefore this methodology allows the reservoir engineer to convert geological models into representative flow models.

Notes by Presenter: It is generally agreed that a double-porosity, single-permeability model is suitable for a connected fracture
network and a matrix medium that only acts as a source of fluids. Otherwise, as soon as the flow in the matrix has to be taken into
account, the double-porosity, double-permeability model should be used. However, it may be difficult to estimate the connectivity
property of a random fracture network, or whether the matrix/fracture permeability contrast does actually lead to non-negligible
contributions from the matrix medium in terms of flow.
Moreover uncertainties in the geometrical properties of the fracture network, such as the fracture length distribution, may affect the
connectivity of the network, particularly around the percolation threshold. As a result it is not always straightforward to identify
which model should be used in practice.
We illustrate this issue by performing a sensitivity analysis on a realistic case combining both small-scale fractures and major objects
such as seismic and sub-seismic faults.

Notes by Presenter: It is generally agreed that a double-porosity, single-permeability model is suitable for a connected fracture
network and a matrix medium that only acts as a source of fluids. Otherwise, as soon as the flow in the matrix has to be taken into
account, the double-porosity, double-permeability model should be used. However, it may be difficult to estimate the connectivity
property of a random fracture network, or whether the matrix/fracture permeability contrast does actually lead to non-negligible
contributions from the matrix medium in terms of flow.
Moreover uncertainties in the geometrical properties of the fracture network, such as the fracture length distribution, may affect the
connectivity of the network, particularly around the percolation threshold. As a result it is not always straightforward to identify
which model should be used in practice.
We illustrate this issue by performing a sensitivity analysis on a realistic case combining both small-scale fractures and major objects
such as seismic and sub-seismic faults.

Notes by Presenter: A 2-pt scheme in space and the Euler scheme in time are applied to the system of diffusion equations.
K and J are the neighbours indexes of the fracture cell I and its associated matrix block, respectively.

Notes by Presenter : The fracture-to-fracture transmissivity is simply computed from the fracture conductivity, the length of the
borderline between the fracture cells, and the distance between the fracture nodes.
The matrix-to-fracture transmissivity is computed from the matrix permeability, the matrix-to-fracture exchange surface and the
average distance between the fracture cell and its matrix block horizontal area.
Therefore the matrix-to-fracture transmissivity is dependent on the local network geometry, it is not homogenized.

Notes by Presenter : Both single and dual permeability models were developed so that the single-phase flow response of any type of
fractured reservoir, whatever the scale, density and connectivity of fractures can be computed.
In the single-permeability model, the matrix-to-matrix flow is neglected, leading to an explicit relation between matrix and fracture
pressures.
Therefore only one equation needs to be solved for determining the full pressure field.
This model is usually valid for a dense and well connected fracture network only, where the matrix medium only acts as a source of
fluids.

Notes by Presenter: In the dual-permeability model, the matrix-to-matrix flow is considered. Thus matrix-to-matrix transmissivities
have to be computed.
The horizontal transmissivities are computed according to the following formula (explain).
However a simplification occurs if the matrix edge is crossed by a fracture.
In this case, as the fracture-to-fracture transmissivity is usually much larger than the matrix-to-matrix one, the expression is simplified
according to this formula, using the nodes interdistance.

Notes by Presenter: The vertical matrix-to-matrix transmissivity is computed according to a geometrical average of the vertical matrix
permeabilities weighted by the layers thickness and the horizontal exchange surface.
The dual-permeability model is valid for any type of fractured reservoir, however its computational cost is at least twice larger than
the single-permeability model. Indeed both pressures in the matrix and fracture network are unknowns, and the matrix-to-matrix
transmissivities have to be computed on a large number of complex matrix blocks.

Notes by Presenter: A dense Warren & Root model is first considered for validating the dual-porosity, single-permeability model
against existing analytical solutions, and verifying that the dual-permeability model does provide the same solutions.
A regular Cartesian fracture network composed of 100 vertical fractures with a spacing of 1m in each horizontal direction is
constructed. This 2D fractured medium has the following characteristics...

Notes by Presenter: A 25m long horizontal well is defined at the centre of the fracture network. A well test is performed with the
following flow rate history...

Notes by Presenter: Pressure buildup and derivative curves are plotted, together with an analytical solution combining Gringarten's
double porosity method and a horizontal well treatment.
The behaviour of the numerical and analytical solutions are very close, thus validating the single-permeability model, as well as the
dual-permeability model in the case of a dense connected fracture network.

Notes by Presenter: The domain contains one set of faults and 4 sets of systematic joints with the following properties.
All fracture lengths and conductivity distributions satisfy a log-normal law.