SPE Copyright 2002, Society of Petroleum Engineers Inc.

Transcription

1 SPE Integrated Workflow Applied to the Characterization of a Carbonate d Reservoir: Qarn Alam Field A.M. Zellou, SPE, EP Tech; L.J. Hartley, Serco Assurance; E.H. Hoogerduijn-Strating, SPE, PDO; S.H.H. Al Dhahab, Shell; W. Boom, Shell; F. Hadrami, PDO Copyright 2002, Society of Petroleum Engineers Inc. This paper was prepared for presentation at the SPE 13th Middle east Oil Show & Conference tp be held in Bahrain 5-8 April This paper was selected for presentation by an SPE Program Committee following review of information contained in an abstract submitted by the author(s). Contents of the paper, as presented, have not been reviewed by the Society of Petroleum Engineers and are subject to correction by the author(s). The material, as presented, does not necessarily reflect any position of the Society of Petroleum Engineers, its officers, or members. Papers presented at SPE meetings are subject to publication review by Editorial Committees of the Society of Petroleum Engineers. Electronic reproduction, distribution, or storage of any part of this paper for commercial purposes without the written consent of the Society of Petroleum Engineers is prohibited. Permission to reproduce in print is restricted to an abstract of not more than 300 words; illustrations may not be copied. The abstract must contain conspicuous acknowledgment of where and by whom the paper was presented. Write Librarian, SPE, P.O. Box , Richardson, TX , U.S.A., fax Abstract This paper presents an innovative integrated workflow applied to the characterization of a carbonate fractured reservoir in order to generate an effective 3D Matrix Block Size (MBS) distribution based on all available data: geological, geophysical and dynamic production data. The MBS is a key factor determining heating and thereby recovery efficiency during steam flooding in a fractured reservoir. The MBS model allows simulating the complex flow and establishing reservoir management strategies that will optimize oil recovery and facility sizing. The major difficulty in developing a 3D MBS model is the ability to account for the all the information pertinent to natural fractures in the field and develop an understanding of what geological characteristics are linked to the occurrence of fractures. Until recently most fractured reservoir modeling tools were limited to simple discrete statistical models. A new approach in fractured reservoir characterization, using primarily artificial intelligence tools, is presented in this paper. The methodology is based on the assumption that there is a complex relationship between a large number of potential geologic drivers (structure, faults, matrix characteristics etc.) and fractures. The combination of both the continuum fracture modeling (CFM) and discrete fracture network (DFN) modeling provides a quantitative framework for MBS distribution estimation, geological concepts and data integration. The application of this integrated workflow to the Qarn Alam field is presented in this paper. Introduction The Qarn Alam Field of Central Oman contains an estimated STOIIP of 185 million m 3 of heavy oil, which is planned to be developed with a full-field steam injection EOR project. Despite the fact that the oil production from the main reservoir, the Shuaiba, is known to occur via a fracture/karst network, the vast majority of the oil is stored in the matrix. The porosity and permeability network of the matrix is often a function of the distribution of depositional rock facies. If significantly different facies are present in a reservoir, their spatial arrangement has to be understood and modeled adequately, in order to understand hydrocarbon distribution and recovery. Critical for the success of a full field steam injection EOR project is a proper understanding of the subsurface geology, most importantly the distribution of matrix porosity and permeability and the occurrence and distribution of hydraulically conductive fractures. This paper illustrates the combination of CFM and DFN modeling to model the fracture distribution and the MBS distribution. Methodology Methods have been presented to model fracture distribution based solely on fracture density measured in wells 1. We believe that the scarcity of wells where fracturing data are available makes difficult and very uncertain such direct mapping, at the field scale, of these data. Fracturing process can be shown, however, to be related to geological parameters that constrain the mechanical behavior of fractured rocks. These factors include attributes such as lithology, bed thickness, stress and strain states, and large-scale structural features (e.g. faults and folds). These parameters are often better sampled at the field scale (especially when derived from seismic attributes). The relationships between the fracturing variable and such geological parameters (also called drivers) can be exploited to model the spatial distribution of fracturing within a field. Although only one parameter may be needed to characterize a fractured reservoir 2-5, it is often the lack of a methodology to integrate the combined effects of structure, thickness, and lithology that leads geologists to focus only on the most important factor. However, complex reservoirs require a more comprehensive description to allow a reliable fracturing prediction. Based on sparse data, the relationships between fracturing and the controlling geological parameters

2 2 A.M. Zellou, L.J. Hartley, E.H. Hoogerduijn-Strating, S. Al Dhahab, W. Boom, F. Hadrami SPE are uncertain. Therefore deterministic approaches are not appropriate (i.e. there is not a single solution). Only probabilistic modeling can help in better assessing these uncertainties. Such a probabilistic approach has been published 6-8 where fracture indices (e.g. count or length per unit area of core) are assumed to be correlated to some geological parameters of the field (e.g. reservoir thickness, structural curvature) and rock petrophysics (e.g. sand resistivities) by some complex, non-linear relationships. These relationships are established using fuzzy-logic techniques to rank the relative importance of the drivers upon fracturing and are then modeled throughout the entire field. The continuum fracture models are used to constrain the fracture distribution in DFN models using a recently developed robust and systematic approach 9. DFN models are used to calibrate dynamic properties of the fracture system (e.g. connectivity, percentage of open fractures and aperture) against well-test data (e.g. pressure build-up tests and flowmeters). These combined fracture approaches are briefly described in this paper. Continuum Modeling (CFM) Using Fuzzy-Logic and Neural Networks. The complexity and the number of fracture drivers in this study justifies an integrated approach to characterize natural fractures. A thorough description of this integrated approach using artificial intelligence tools is given in Ouenes. 10 In this paper, we will limit ourselves to a short summary of the methodology based on the three following steps: Ranking the fracturing drivers: when considering several possible fracturing drivers, the relative importance of each driver (the input) and the impact on the fracture frequency (the output) is critical. Given the dataset, many different techniques can be used to rank the fracturing drivers. In this methodology, we employ a fuzzy-logic tool that has proven very efficient and fast in ranking geologic parameters. In this specific study, we dealt initially with 16 fracturing drivers. After ranking and eliminating the less influential drivers, the geologist and the engineer have to analyze, understand and validate this ranking and the effect of each geological parameter on the fracturing. Training and testing: the above ranked (and validated) drivers are used to establish the complex, non-linear relationship relating the fracture frequency to these drivers. This is done using a neural-network technique. To that end, the set of available data, where the inputs and output are known, is divided into two subsets: a training set and a testing set. Each regression model is derived by selecting randomly (or according to some rule) the training set. The neural network modeling process consists of adjusting some weights until the actual fracture frequencies match the estimated ones. Once the matching process is done, we can assume that a model is available and can be used for testing and cross-validation. Depending on the ability of the model to predict the fracture frequency at testing locations, the model can be kept for further use or discarded. Simulation process: the stochastic aspect of this approach is related to the uncertainty about the above regression models. Therefore, several realizations or simulations are generated cell by cell over the entire field where the fracture drivers are known to predict the fracture frequency. With a large number of realizations, a model describing the average value and probability of occurrence are estimated over the entire field. Discrete Modeling (DFN). The DFN conceptual model is that a significant component of fluid flow occurs in a network of discrete fracture conduits. Each fracture is modeled as an explicit object defined in terms of its orientation, size/shape and flow properties such as flow aperture (or individual fracture permeability), storativity and porosity. s can vary hugely in terms of size from large faults that can be seen on seismic to sub-seismic fractures that are seen in FMI or core data. Hence, a fracture model may contain a range of fracture sizes that varies from large deterministic fault structures that cross many simulation cells to much smaller fractures where hundreds of fractures occupy a single simulation cell. Since few sub-seismic fractures can be measured directly, a stochastic framework is used to parameterize these fractures. Conditioning and screening procedures are then used to validate the stochastic models. Flux-based upscaling techniques allow equivalent properties, such as fracture permeability and MBS, for continuum models to be derived on a cell-by-cell basis that are consistent with the underlying detailed fracture model. Using our integrated approach, DFN models are not random anymore and they honor all the existing reservoir data taken into account through the CFM model. The key input for the DFN model is as follow: orientation obtained from statistical analysis of dip and dip azimuth log on FMI. spatial distribution obtained as fracture intensity from neural network modeling. Fraction of open fractures estimated by picking fracture types (open, partially open, cemented, etc.) on FMI. However, this is uncertain as not all fractures picked as open on FMI or core are necessarily connected to the network and take flow. This parameter has to be derived from conditioning on dynamic data. length (along strike) distribution is uncertain. It can be measured directly from outcrops. Alternatively, analogs can be used and the sensitivity to fracture length calculated. height (or vertical penetration) is also uncertain. Direct measurements are only possible from outcrop studies. Analogs, or geological concepts such as links to bedding, can be used and sensitivities considered. flow aperture governs the permeability of an individual fracture. It is typically less than the physical aperture seen in FMI due to fracture roughness, channelisation, gauge material, etc. It can be measured directly from high density flow-meters, short interval/short duration tests, or by fitting dynamic data such as pressure build-up or interference tests.

3 SPE Integrated Workflow Applied to the Characterization of a Carbonate d Reservoir: Qarn Alam Field 3 Application to the Qarn Alam Field Background. The Qarn Alam Field is a fractured reservoir with an estimated STOIIP of 185 million m 3 of heavy oil. After its discovery, production peaked in 1976 at 6000 m 3 /d net oil within a year. The peak production originated from the oil in the extensive fracture network, not replenished by matrix oil due to low matrix permeability and high oil viscosity. In 1992, the design of a steam injection pilot plant commenced and in 1996 steam injection started. The Qarn Alam Field is a heavily fractured carbonate reservoir with a strong aquifer drive. The field is 3 by 6 km and has a maximum oil column of 165 m. The top of the reservoir is at 210 mss. The permeability is low at 7 mdarcy, viscosity relatively high (220 cp at reservoir conditions). The oil is heavy, API 16. Solution GOR is 10 m 3 /m 3. Crestal reservoir pressure is just below the bubble point of 33 bar. s are essential for efficient distribution of steam and the collection of the drained fluids. The application of the continuum fracture modeling (CFM) and discrete fracture network (DFN) modeling to the Qarn Alam Field is hence applied. Integrated d Reservoir Workflow. The integrated fractured reservoir workflow is summarized in four main steps (Fig. 1): 1. Conventional geological modeling in order to obtain a 3D model of the matrix properties and a detailed description of the near well-bore fracture intensities (from image logs) leading to a conceptual fracture model. 2. Continuum fracture modeling (CFM). An MBS distribution is obtained. 3. Discrete fracture network model (DFN). A second MBS distribution is obtained. 4. Transfer of the MBS distribution to the reservoir simulator. For data format reasons this is done via the static reservoir model application. The above integrated fractured reservoir workflow is a live model, i.e. whenever new data becomes available (such as an update of step 1) new MBS distributions can be quickly and easily generated. The continuum fracture modeling methodology (step 2) is described in more detail below. It comprises four main steps (Fig. 2): 1. The first step in the described methodology is the ranking of all existing geologic drivers. A fuzzy neural network is used to evaluate the hierarchical effect of each geologic driver on the fractures. As a result, the geologist or reservoir engineer will be able to identify, locally and globally, the key geologic drivers affecting fractures. The ranking is checked against conceptual fracture models. Fig. 3 illustrates the main fracturing drivers for this specific field. 2. The second step of the approach is to create a set of stochastic models using a back propagation neural network that will try to quantify the underlying complex relationship that may exist between key geologic drivers and fracture intensity. The training (Fig. 4) and testing (Fig. 5) of the neural network is accomplished using existing data. 3. The third step of the approach is to perform a statistical analysis by examining the fracture distribution resulting from the stochastic models. Using these three steps, the software will be able to predict the 3D distribution of fractures and their underlying uncertainty at undrilled locations (Fig. 6 and 7). 4. The fourth step in this approach is to convert the resulting fracture models to an estimation of the 3D MBS distribution. Discrete fracture modeling (step 3) provides a methodology (Fig. 8) for integrating additional fracture data (e.g. orientation and length) and for calibrating the models against dynamic data. The distribution of fracture density in the DFN model is contrained to the CFM model. Although the CFM model predicts the density of all fractures, the dynamic response of the reservoir will only depend on the density of open fractures. The percentage of open fractures cannot be determined reliably from FMI, so the percentage of open fractures was varied to calibrate the dynamic response predicted by DFN models to fracture permeabilities derived from well-tests. A range of 18-33% with a mean of 25% open fractures was derived. An estimation of MBS based only on open fractures can then be calculated from the DFN model (Fig. 9). Because the geometry of individual fractures is modelled explicitly in DFN models, more detail such as the shapes of the matrix blocks can also be obtained. Fig. 10 shows a plan view of the full field DFN model. Matrix Block Size Estimation. The key parameter to be calculated as part of this initial fracture study is the size and shape of the inter-fracture matrix blocks as it has a great impact on the viability of steam injection. The MBS is calculated on a cell by cell basis in a similar manner to permeability. The fractures are identified that have any part in a cell, and any fractures that are isolated are removed as they do not contribute to flow. An array of 25 hypothetical parallel cores are used to calculate statistics for the fracture separation within each cell. The MBS is calculated as the total core length divided by the number of fractures intersected. This is repeated in three orthogonal directions to give a directional MBS such that the shape as well as the size is calculated. Since fractures are predominantly sub-vertical here, then the matrix blocks are typically tall and thin. This is significant for steam injection since the minimum MBS direction will be important for heat conduction into the matrix blocks and the vertical MBS will control gravity drainage. Fig. 11 shows the full-field 3D effective matrix block size distribution, with an overall increase from crest to flank and a vertical and horizontal distribution.

4 4 A.M. Zellou, L.J. Hartley, E.H. Hoogerduijn-Strating, S. Al Dhahab, W. Boom, F. Hadrami SPE Conclusion 1. An integrated workflow for the estimation of fracture frequency and matrix block size distribution is presented. 2. This integrated fracture modeling project is performed using fracture frequency from image logs as a fracture indicator. The fracture indicator is based on all identified fractures, i.e. open plus (partially) cemented fractures. 3. Based on the available data is concluded that structural drivers (curvature, distance to fault etc) play a major role in fracturing. Matrix properties are secondary in predicting the fracture frequency distribution. Lithology (facies) does not seem to play a major role. 4. A correlation coefficient higher than 0.75 between the actual and the predicted fracture frequency is achieved. 10. Ouenes, A.: Practical application of fuzzy logic and neural networks to fractured reservoir characterization, Computer and Geosciences, Sahab Mohageg (Ed.), (2000) v. 26, no 7. Acknowledgements The authors would like to thank Petroleum Development Oman [PDO] and the Ministry of Oil and Gas [MOG] for authorizing the publication of this paper. The authors would also like to thank Ian Gollifer and Volker Varhenkamp for their contribution in the geological modeling part of this study. References 1. Guerreiro, L. Costa Silva, A., Alcobia, V. and Soares, A, Integrated reservoir characterisation of a fractured carbonate reservoir, paper SPE 58995, presented at the 2000 SPE Petroleum Computer Conference, Villahermosa, Feb Harris, J., Taylor, G., Walper, J.: Relation of deformational structures in sedimentary rocks to regional and local structure, AAPG Bulletin, (1960) v. 44 p Lisle J. L.: Detection of zones of abnormal strains in structures using Gaussian curvature analysis, AAPG Bulletin, (Dec. 1994), v. 78, no. 12, p McQuillan H.: "Small scale fracture density in Asmari formation southwest Iran and its relation to bed thickness and structural setting," AAPG Bulletin, (1973) v. 57, no. 12, p Murray, G.: Quantitative fracture study- Sanish Pool, McKenzie County, North Dakota, AAPG Bulletin, (Jan. 1968), v. 52, no. 1, p Zellou, A.M, Ouenes, A. and Banik, A.K: "Improved fractured reservoir characterization using neural network, geomechanics and 3-D seismic," paper SPE presented at the 1995 SPE Annual Technical Conference and Exhibition, Dallas, Oct Ouenes, A., Richardson, S., and Weiss,W.W.: d Reservoir Characterization and Performance Forecasting Using Geomechanics and Artificial Intelligence," paper SPE presented at the 1995 SPE Annual Technical Conference and Exhibition, Dallas, Oct Ouenes, A., Zellou, A.M., Basinski, P.M. and Head, C.F.: Pratical use of neural networks in tight gas fractured reservoirs: application to the San Juan Basin, paper SPE presented at the 1998 Rocky Mountain Regional Low Permeability Reservoirs Symposium, Denver, CO, Aug. 9. Ouenes, A. and Hartley, L.J.: Integrated d Reservoir Modeling Using Both Discrete and Continuum Aprroaches, paper SPE presented at the 2000 SPE Annual Technical Conference and Exhibition, Dallas, Oct. 1-4.

5 SPE Integrated Workflow Applied to the Characterization of a Carbonate d Reservoir: Qarn Alam Field 5 Curvatures Slopes Data Gathering/Analysis Conceptual Model Drivers Structure Porosity Seismic Impedance s etc.. 3D Continuum Modeling Ranking of Drivers Indicators Faults DRIVERS Density Estimation 3D Discrete Modeling Fig. 3: Main Fracturing Drivers on Qarn Alam. Matrix Block Size Estimation Fig. 1: Integrated d Reservoir Concept. Data Gathering List of Fracturing Drivers Choice of Fracturing Indicator Training Validation Modeling Ranking of Drivers Frequency Estimation MBS Estimation Fig. 4: Result of the training set. Fig. 2: Continuum Modeling Workflow

6 6 A.M. Zellou, L.J. Hartley, E.H. Hoogerduijn-Strating, S. Al Dhahab, W. Boom, F. Hadrami SPE Fig. 5: Result of the testing set. Fig. 7: Frequency Model: one of the lower layers. 3D Continuum Model Orientation Distribution DFN Model Connectivity Directional MBS Estimation Fig. 6: Frequency Model: one of the upper layers. Fig. 8: DFN Workflow.

7 SPE Integrated Workflow Applied to the Characterization of a Carbonate d Reservoir: Qarn Alam Field 7 Level 1 Connectivity Analysis Level 5 25% open fractures Level 10 Near-Well Permeability from flow simualtion Fig. 10: A plan view of the DFN Model. The 3D model contains over 1 million discrete fracture objects. Notice how the DFN model reproduces the actual geologic reality of the fracture model shown in Fig. 6. Check effective well Permeability against info from dynamic data Fig. 9: model conditioning with well dynamic data.

8 8 A.M. Zellou, L.J. Hartley, E.H. Hoogerduijn-Strating, S. Al Dhahab, W. Boom, F. Hadrami SPE N Base grid: 1*1 km N Block Size < 10 m 20 m 40m > 60 m Base grid: 1*1 km Fig. 11: Full-field 3D effective matrix block size distribution, showing an overall increase from crest to flank with a vertical and horizontal distribution.