--> Abstract: Improved Estimation of Reservoir Compartmentalization, Tributary Drainage Volume and Connectivity Through Discrete Fracture Network Modeling, by P. R. La Pointe and V. Ivanova; #90928 (1999).
[First Hit]

Datapages, Inc.Print this page

LA POINTE, PAUL R.1, TODD FOXFORD1, and VIOLETA IVANOVA2
1 Golder Associates Inc.
2 Massachusetts Institute of Technology

Abstract: Improved Estimation of Reservoir Compartmentalization, Tributary Drainage Volume and Connectivity Through Discrete Fracture Network Modeling

Introduction

Fractured reservoirs often exhibit a high degree of heterogeneity in their production characteristics. Two wells separated by a kilometer might show interference within hours or days, while wells in between show no interference effects. Likewise, some wells may produce large volumes of oil for many years, while other nearby wells ostensibly in the same geological environment produce much less. This heterogeneity is due to differences in fracture network connectivity and permeability at the scales of interest. Sometimes the natural fractures form clusters that are well connected to themselves, but not to other fracture network clusters. This leads to reservoir compartmentalization. It is important to understand the sizes of these compartments in order to optimize well spacings, Previous HitevaluateNext Hit proposed tertiary recovery processes, or correctly forecast ultimate field production totals, recovery factors or production rates. In an exploration context, understanding fracture connectivity and Previous HithowNext Hit efficiently a proposed well might drain the surrounding reservoir by means of pressure depletion or gravity drainage is important for appraising prospects in fractured reservoirs. More than one highly-touted prospect in a fractured reservoir has turned out to be an economic failure due to limited drainage extent of accessible matrix.

Discrete fracture network (DFN) modeling is a relatively new tool that originated in the nuclear waste repository characterization and simulation field. It is now being successfully applied to problems in fractured reservoir characterization and simulation. This paper describes Previous HithowNext Hit discrete fracture network models can be generated from typical well and seismic data; Previous HithowNext Hit their network connectivity and flow properties are conditioned to well tests; and finally, Previous HithowNext Hit these models can be used to estimate tributary drainage volume for prospect appraisal, optimizing well spacings for conventional and tertiary recovery processes; and estimating potential production gains from hydrofracturing.

Methodology

Discrete fracture network models simplify the complex structural geology of fractured reservoirs. A DFN model explicitly represents each joint or fault that is significant from a fluid flow standpoint at the scale of interest. Figure I shows a DFN representation of a fractured reservoir which was originally deposited as a series of flat lying carbonates interlayered with shales, then experienced compression leading to a horst. The reservoir was later uplifted and tilted, leading to erosion and new strata deposited unconformably on the older reservoir units. At a later time the rock subsided.

Not every fracture is modeled in Figure 1. Only fractures that are large enough to produce non-continuum behavior at the scale of reservoir simulation were explicitly represented. Moreover, the DFN model is a combination of both deterministic and stochastic components. The location of major faults is typically known from seismic, and these faults are placed into the DFN model where they are known to occur. However, there are many sub-seismic faults and larger joints that are not detected through seismic, but do play an important role in reservoir flow processes. These are modeled stochastically, that is, their locations, size, intensity orientation and fluid flow properties are based upon statistical distributions derived from study of fractures identified through FMI logs or core. They are further conditioned to structural features such as folds, faults, geological domains and inferred directions of principal insitu stresses.

Each discrete fracture has fluid flow properties associated with it. These include such parameters as storativity, transmissivity/transmissibility, aperture and roughness. The values for these parameters are assigned to each discrete fracture through the use of non-linear optimization techniques such that the resulting DFN models match well tests of various types. Once the DFN model is conditioned to known geology and matches reservoir connectivity measures derived from well tests, such as flow dimension, the DFN model can be used to study network connectivity, injectivity and productivity under a variety of assumptions and scenarios.

It is straightforward to determine which fractures are connected to each other, and whether the fractures form clusters. These clusters occupy a volume of the reservoir which can be computed using algorithms derived from graph theory. In this way, it is possible to compute the matrix volume accessed by each fracture network, and to compute its cross-section in any orientation. Knowledge of the volume for all of the networks provides an estimate of the total amount of matrix that can be drained from each cluster under pressure depletion or gravity drainage. The statistics developed for all network clusters makes it possible to estimate mean compartment volume, maximum compartment volume, and other relevant statistics that can be used to Previous HitevaluateNext Hit optimal well spacings, orientation and improve prospect appraisal by computing a distribution of possible well recoveries..

Field Application

The Yates Field in west Texas is a fractured carbonate reservoir that is being produced through gravity drainage, and is a testing site for the Thermally-Assisted Gravity Segregation (TAGS) process patented by Marathon Oil. This process combines injection of steam into fracture networks and vertical management of fluid contacts to recover additional oil from this mature reservoir. A cooperative project between Marathon Oil, the Massachusetts Institute of Technology and Golder Associates funded by the Department of Energy has made it possible to Previous HitevaluateNext Hit these new technologies for estimating compartmentalization and drainage for wells in two tracts in the Yates Field. This paper will illustrate Previous HithowNext Hit the DFN model was constructed for the Yates Field, the estimates of compartmentalization and drainage made using the new characterization technology, and Previous HithowNext Hit these estimates Previous HitcompareNext Hit with well test results.

Conclusions

This study shows that DFN models provide a new and powerful method to assess the production characteristics of fractured reservoirs. The data to create a DFN model can include FMS, FMI, core, DST, flow logs, seismic, lineament Previous HitmapsNext Hit and “soft” geological concepts, yet such models can be built from virtually any single one of these data sets. Application of these methods to the Yates Field illustrates that they predict results consistent with single and multiwell experiments in support of the TAGS process.

AAPG Search and Discovery Article #90928©1999 AAPG Annual Convention, San Antonio, Texas