Summary
This paper presents an innovative approach to integrate fracture, well-test, and production data into the static description of a reservoir model as an input to the flow simulation. The approach has been implemented successfully in a field study of a giant naturally fractured carbonate reservoir in the Middle East. This study was part of a full-field integrated reservoir-characterization and flow-simulation project.
The main input available for this work includes matrix properties and fracture-network, well-test, and production data. Stochastic models of matrix properties were generated using a geostatistical methodology based on well logs, core, seismic data, and geological interpretation. The fracture network was described in the reservoir as lineaments (fracture swarms) showing two major fracture trends. The network and its properties (i.e., fracture porosities and permeabilities) were generated by reconciling seismic, well-log, and dynamic data (Well Test and Production Log Tool, PLT).
The challenge of the study is to integrate all the input in an efficient and practical way to produce a consistent model between static and dynamic data. As a result, it is expected to reduce the history-matching effort. This challenge was solved by an innovative iterative procedure between the static and dynamic models.
The static part consists of the calibration of model permeability to match the well-test permeability. It is done by comparing their flow potentials, kh. In this analysis, the dominant factor in controlling production at each well, either matrix or fracture, was determined. Based on the dominant factor, matrix or fracture permeability was modified accordingly. This way, the changes in permeability are consistent with the geological understanding of the field.
The dynamic part was carried out through a full-field flow simulation to integrate production data. The flow simulation at this stage was used to match production capacity, [i.e., to determine whether the given permeability (matrix and fracture) distribution is enough to produce the fluid at the specified pressure during the producing period of the well]. The iteration is stopped once a reasonable production-capacity match is obtained. In general, a good match was achieved within three to four iterations. The generated reservoir description is expected to substantially reduce the effort required to obtain a good history match.
Introduction
This paper presents the approach, implementation, and results of a fracture-integration process into a reservoir model. The study is part of a fully integrated reservoir-characterization and flow-simulation study of an oilfield in the Middle East. A comprehensive integrated reservoir characterization was conducted by considering all available data, namely well logs and cores, geological interpretation, seismic (structures and inversion-derived porosity), fracture network, and pressure-buildup (PBU) tests. The approach used in the study was a stochastic approach in which multiple reservoir descriptions were generated to quantify the uncertainty in future performance.
Reservoir properties for each realization were generated with a geostatistical technique that produces properties (i.e., porosity, permeability, and water saturation) consistent with the underlying rock-type description. The description was based on core and log data. Additionally, porosity, which affects the permeability description, was also constrained to the seismic-derived porosity. The permeability distribution generated by this method is referred to as the core-derived permeability in this paper. Because core measurement commonly represents the matrix property of the rock, the core-derived permeability mentioned above was also referred to as matrix permeability.
It is commonly observed that the well-test permeability values do not match the thickness-weighted core-permeability averages. This is partly because of the differences in the measurement scales of core samples, which cover a few inches, and well tests, which investigate several hundred feet around the wellbore. In addition, the presence of fractures and/or high-permeability channels will further enhance the difference between the two sources of data. The mismatch between these two permeabilities may be small or as high as three orders of magnitude. Therefore, reservoir descriptions based on core measurements alone cannot honor the well-test results and need to be modified properly.
Fracture Network Numerical Well Test Model Based on the Discrete-fracture Model