|
Software
OxFilet
OxFilet is a FracSys/DOS module to derive the conductive fracture intensity
the fracture transmissivity distribution, based on transmissivity measurements
in packer and well tests. A Windows 95/NT version of OxFilet is currently under
development.

OxFilet uses forward modeling of the
well testing process to produce simulated well tests which can be compared against
measured well test results. Since the simulated test results are based primarily
on assumptions of fracture intensity and transmissivity distribution, OxFilet
provides a consistent approach for derivation of these parameters.
The primary advantage of the OxFilet approach is that it recognizes that well
tests in fractured rock generally depend on the properties of the fractures
which happen to be intersected by the tested interval. These well test results
do not reflect the "averaged" properties of a "representative
elementary volume." Rather, they represent the specific properties of specific
properties. As a result, the well tests results are best used to derive a statistical
description of the tested fractures, rather than effective continuum properties.
OxFilet requires the following assumptions to derive fracture properties from
well tests:
- Well interval transmissivity is the sum of "at borehole" fracture
transmissivity
- "At borehole" fracture transmissivity can be related to the effective
transmissivity of the fracture at larger scales as a function, incorporating
effects of uncertainty and intersecting fractures
- The spatial pattern and statistical properties of fractures are similar
among tested intervals
- Each fracture is represented only once in well test results
OxFilet derives the conductive fracture transmissivity and intensity by comparing
simulated and well test results, and adjusts the assumed intensity and transmissivity
distribution to match the measured well test transmissivity distribution.
OxFilet Features
| Feature |
Description |
| Platform |
DOS (FracSys Version 2.601)
Windows 95/NT (under development) |
| Approach |
Simulation of well test results within DFN model
realizations |
| Transmissivity Distribution Assumptions |
Constant
Uniform
Exponential
Lognormal
Normal
Power Law (Pareto) |
| Intensity Assumption |
Baecher (Poisson model)
Fractal model (under development) |
| Well Test Assumptions |
Well Test Transmissivity is Sum of the Transmissivity
of Intersected Fractures
Well Test Transmissivity is a Function of Intersected
Fracture Networks
Well Test Transmissivity is a Function of the Aperuture
Field on Intersected Fractures at the Well |
| Goodness of Fit Criteria |
Kolmogorov-Smirnov
Chi-Squeared
Moments |
| Search Algorithms |
Conjugate Gradient
Simulated Annealing
Interactive Visual Optimization |
| Statistical Reporting |
Well Test Transmissivity Mean, Std Dev. Skewness,
Kurtosis
Percent of Non-conductive intervals |
| Graphical Reporting |
Histogram (PDF)
Cumulative Distributions (CDF) |
| |
|
Algorithm Summary
The input for the OxFilet algorithm consists of:
- The geometry of the wells from which the data was collected, and
- The measured packer interval transmissivities.
Based on the packer interval transmissivity information, OxFilet calculates
the following packer interval transmissivity statistics:
- Mean, standard deviation, skewness, and kurtosis of interval transmissivity
- Percentage of non-conductive intervals
OxFilet obtains an initial estimate of the conductive fracture intensity P10
(conductive fractures per meter) from the percentage of non-conductive packer
intervals P0 as (Snow, 1965),
P10 = -ln(P0)/L
where L is the mean packer interval length.
The user then specifies the assumed conductive fracture intensity P10, fracture
orientation distribution (Fisher, Bingham, bootstrap, etc) and transmissivity
distribution (log normal, power law, exponential, etc.).
The following parameters are used to control the FracSize simulation:
- The transmissivity threshold corresponding to "non conductive"
intervals
- The search space maximum, minimum, and step size for fracture transmissivity
mean and standard deviation
- The number of realizations to be combined to calculate the simulated packer
test transmissivity statistics for a given fracture transmissivity and intensity
assumption
- The type of search (single point, grid search, conjugate gradient search,
or simulated annealing search)
- The search criterea (Kolmogorov-Smirnov or Chi-Squared Statistic)
- The maximum number of iterations, parameter temperature, and jumping scale
(for a simulated annealing search), or the damping factor (for a conjugate
gradient search)
Based on the parameters provided, OxFilet generates stochastic fracture realizations,
and calculates the packer interval transmissivity statistics and percentage
non-conductive intervals.
OxFilet compares simulated packer interval transmissivity distributions against
in situ measured test results using either Kolmogorov-Smirnov or Chi-Squared
statistics. OxFilet can search for the fracture transmissivity distribution
and conductive intensity with the best goodness of fit measures using either
a conjugate gradient or simulated annealing search.
|