A Case History on Integrated Fracture Modeling in a Giant Field

In this paper, a case study is used to demonstrate a straightforward methodology of faults, fractures and highpermeability layers integration in a single porosity single permeability (SPSP) reservoir simulation model. The application of this method in the Ghawar Arab-D reservoir indicated an adequate modeling of the water encroachment pattern. The described methodology starts with the identification of structural lineaments from 3D seismic data analysis (radius of curvature, coherency analysis), the availability of image logs being critical to validate this early step. Further engineering data such as mud losses during drilling, flowmeter surveys, transient well test interpretations, geochemical analysis and water encroachment patterns are integrated to further describe sealed or conductive fractured areas among the initially picked set of lineaments. Stratiform very high permeability streaks (Super-K) are also described at this stage, resulting in a finescale geological model conditioned by engineering data. Both conductive fractures and stratiform Super-K features are being assigned adequate transmissibility and relative permeability functions in the upscaled reservoir simulation model. Transmissibility values are subject to further adjustments during the history matching process. Comparison with dual porosity dual permeability (DPDP) modeling is briefly discussed. Cost (CPU time) and benefits of SPSP and DPDP modeling techniques are addressed to conclude that in the present case, the trade-off of implicit fracture/Super-K modeling (SPSP) is compensated by a sufficiently-accurate modeling of the water encroachment pattern and shorter simulation turnover times. Introduction The Ghawar Arab-D reservoir is the largest oil field in the world. Trending on a nearly north-south basement high, the Ghawar field is a carbonate reservoir with a continuous oil column more than 200 km long and up to 40 km wide (Figure1). Since the northern and more prolific area was put on stream in 1951, development progressed southwards and will reach the southern tip of the field before 2010. From north to south, matrix reservoir permeability is globally decreasing with vertical well potential typically dropping four-fold. Due to the combined sheer size of the Ghawar field and a peripheral water injection pattern, water breakthrough did not occur in Ghawar until 1983, more than thirty years after the field start-up. As matrix reservoir permeability is decreasing southwards, effects of permeability heterogeneities such as conductive or sealed fractures, faults or stratiform Super-K are magnified and reflected in the irregular water encroachment pattern (Figure-2). In Saudi Aramco, several efforts have been launched during the last decade to enhance the reservoir modeling of the elusive fractures of the Ghawar field. Indeed, until recently, the reservoir development was predominantly made using vertical wells that have a low probability of sampling reservoir fractures, not to say the fracture distribution and orientation. Note however that horizontal wells, on the other hand, can easily miss stratiform Super-K occurrences. The coupling of technical advances in several domains spanning from seismic, drilling to reservoir characterization and simulation has made possible a better understanding and quantification of the Arab-D carbonate reservoir in Ghawar. Although facies description and prediction in carbonates are necessary to achieve a correct reservoir description – matrix porosity contains 99% of the oil in place– this paper will focuse on the fluid flow modeling in high permeability features such as conductive fractures and stratiform Super-K. Interested readers can report to other authors for details on carbonate facies modeling through the combined use of outcrop description, logs, core data and geostatistics . Short Literature Review Previous papers were recently published on Ghawar reservoir characterization and fluid flow simulation. The previous area of study focus was in the upper part of central SPE 71340 A Case History on Integrated Fracture Modeling in a Giant Field Z.A. Al-Ali, SPE, Saudi Aramco, B.A. Stenger, SPE, Saudi Aramco 2 Z.A. AL-ALI, B.A. STENGER SPE 71340 Ghawar where water encroachment and recovery have reached a mature stage after more than 40 years of plateau production. Successful reservoir modeling starts by acquiring data of a good quality, the basis to derive conceptual models of the reservoir behaviour. For instance, Al-Shahri et al. (5) stressed the need to perform adequate transient well testing in order to detect features such as conductive vertical fractures dynamically connected to stratiform Super-K layers. In the past, Super-K layers were suspected of having a globally negative impact on oil recovery by leaving large amounts of by-passed oil. Consequently, several studies were launched that covered all aspects of reservoir studies such as reservoir description , well behaviour (7) and reservoir simulation . Subsequent reservoir monitoring has shown that Super-K features are generally in good communication with the rest of the reservoir . Super-K features benefited from being well sampled by the vertical well development adopted since the early years of Ghawar production. Fractures on the other hand remained typically unseen by vertical wells although regularly suspected in the past. Lack of material proof remained until the drilling of slanted and horizontal wells during the last five years. Image logs acquired in wells affected by lost circulation during drilling demonstrated the existence of natural fractures. Recent attempts to model conductive fractures communicating with stratiform Super-K using Local Grid Refinements (LGR) in SPSP reservoir simulation models are documented . Efforts to integrate both Super-K and fractures in a single earth model are also reported by Phelps et al. . Case Study Area Our case study area is located in a recently developed region located in southern Ghawar (Figure-3). Although sporadic production occurred since 1970 through a limited number of appraisal wells, plateau production started only by mid 1996. After a few years of production, irregular water encroachment was observed (Figure-4) and modeling efforts were launched to better understand and predict the combined effects of lower matrix permeability and high permeability features, such as conductive fractures and stratiform Super-K. Elements of Reservoir Characterization Before integrating different source of data, it is necessary to study and validate them one by one. We will review some elements of reservoir characterization such as well testing, production logging or fracture identification. Transient Well Test Interpretation from a Super-K Well In addition to near wellbore (skin) and distant effects (constant pressure or no-flow boundaries), transient well test interpretation yields the following ratio: μ eff r H K k × × .............................................Eq-1 For single-phase testing, this ratio simplifies as relative permeability and viscosity values are reasonably well defined. In this case, the only remaining uncertainty relates to the payzone thickess. It is interesting to look at one well test interpretation close to our study area (Figure-5). Although the test quality is debatable due to surface closure during testing, this pressure build-up test shows a clear linear flow regime (Figure-6). Interestingly enough, this well test may be interpreted with an infinite-acting fracture model and with a homogeneous model using a negative skin. The first interpretation provides an estimate of the Super-K extension and matrix permeability thickness (K*H) product while the second one gives an estimation of the K*H of the Super-K itself (Table-1). Note that the interpreted matrix K*H was found in good agreement with the average value known from near-by wells not affected by Super-K features. Flowmeter from a Super-K Well Full Bore Spinner (FBS) surveys are reliable sources for allocating production in the Ghawar producing wells. Indeed, open hole completions do not carry the usual problem of channeling behind poorly cemented casing that render most flowmeter surveys useless in perforated cased producers An example of production profile from a Super-K well shows a huge proportion of the production coming from a few feet with a typical stair-step curve (Figure-7). By comparison, a well not affected by a Super-K feature will present a smoother flowmeter profile (Figure-8). Fracture identification from image logs As said previously, the detection of large mud losses during drilling will lead to investigate its causes using image logging. With the increase of horizontal well drilling in Ghawar, image logs have been critical in quantifying the type, azimuth and distribution of naturally occurring fractures. Both mud losses (Figure-9) and image logs analysis (Figure10) can be compared with the structural lineament map (Figure-11) derived from the 3D seismic interpretation (coherency analysis, radius of curvature). In a latter stage, all information sources pertaining to fracture description are combined to generate fracture maps using a stochastic approach. Earlier works (12) have shown that in Ghawar, fracturing tends to be organized in corridors or ‘swarms’ rather being diffuse. Fine-Scale Reservoir Model Building The question of conditioning geological models using engineering data such as flowmeter, transient well test interpretations and production has been discussed in the literature. Aspects such as non-uniqueness of the solution (13) will not be discussed here although being acknowledged. Noteworthy, Roggero et al. proposed an innovative method for achieving fast convergence of multiple geostatistical realizations constrained by historical pressure and production data . Although promising, this method is still at a research stage and was not used in our current approach. SPE 71340 A CASE HISTORY ON INTEGRATED FRACTURE MODELING IN A GIANT FIELD 3 As mentioned earlier, the treatment of matrix modeling and conditioning will not be dealt in this paper, as this part of the process does not show any noticeable difference with the approach elaborated by Mezghani et al. . To summarize, a number of lithofacies are typically assigned a value of porosity (Φ) and permeability (K), both values being generally initialized using core or open hole data. An optimization process is consequently applied to find optimum values for each lithofacies by conditioning to flowmeter and transient well test K*H values. Super-K occurrences are treated separately from lithofacies through a distinct superposition algorithm. In our study area, this method yielded a very good match between the observed and simulated flowmeter profiles, both for wells with and without Super-K features (Figures-12 and 13). After performing the optimization process, the underlying assumption is that the geostatistical process will generate a reasonably accurate prediction of the inter-well properties. Of course, this point remains highly debatable but as we will see further in this paper, this had a limited impact in our study case. Assuming that matrix and Super-K features have been adequately modeled, fracture swarms are still to be described in terms of porosity and conductivity. Such values are available throughout the literature with porosity values generally less than 1% and fracture conductivities up to 40 D.m. In our approach, median values were globally assigned to the fractures and fine-tuned during the history-matching phase of the study. Upscaling to Reservoir Simulation Model The question of implicit (SPSP) or explicit (DPDP) fracture and Super-K modeling becomes now central. In a SPSP approach, the upscaling process will involve the derivation of pseudo-permeability functions for the fracture/Super-K features while in the DPDP case, a traditional upscaling process will be applied, some refinements being possible through the use of variable fracture block size. Note: fractured block size is involved in the calculation of the shape factor that controls the capillary exchanges between the matrix and fracture media. Far from overlooking this critical part of the methodology, the traditional part of the upscaling process is relatively straightforward in Ghawar as matrix reservoir quality varies generally smoothly areally and vertically. Stratiform Super-K and vertical conductive fractures require using a more elaborate approach, more specifically in the SPSP approach. Waterflooding simulations on finely gridded models are generally necessary to derive pseudoized relative permeability functions. The underlying assumption when using such pseudo relative permeability functions is that simulation conditions in the upscaled coarse simulation model are similar to the ones encountered during the limited simulation on a finely gridded model, generally a crosssection. This assumption is generally not respected and regularly forgotten during the next phases of the work. In our approach, an original method (15) was proposed that consists in directly combining traditional rock relative permeability and linear curves for each simulation grid cell with the assumption that water floods the fracture before entering the matrix medium. End-point scaling is consequently used in combination with a limited number of curves for the sake of simplicity (Figure-14). Note: our study does not involve free gas at reservoir level, limiting the pseudoization exercise to a water/oil case. Boundary Conditions Due to the sheer size of the Ghawar field, a unique highresolution reservoir simulation model is still out of reach in spite of the recent advances in megacell simulations as advocated by Dogru et al. . This implies that sector models in mature areas have to be built with adequate boundary conditions to account for neighbouring producing areas and aquifer influx. At Saudi Aramco Reservoir Simulation Division, a regional Arab-D aquifer model is maintained to provide boundary conditions to sector models. This model is characterized by an inexpensive CPU time of 1 hr on a regular SP2 node and it is history-matched in pressure, providing a head start for the history matching of sector models. Another technique consists in providing each sector models with pseudo-wells to account for neighboring producing areas . This study used both techniques, the latter being applied to the SPSP model with pseudoized relative permeability curves. Discussion Results of the SPSP model were satisfactory in terms of the water breakthrough history match quality (Figure-15 and 16). Particularly, the model was successful in reproducing the irregular water front pattern as depicted earlier. Few global modifications of the fracture conductivity were indeed necessary to converge to an acceptable history match. This demonstrates that SPSP modeling may be successful in representing permeability heterogeneities such as conductive fracture swarms or stratiform Super-K under a certain set of modeling assumptions and reservoir conditions. While the former have been discussed earlier in this paper, enabling reservoir conditions can be identified as follow: • An adequate set of reservoir data describing matrix, stratiform Super-K and fracture entities; • A low variability in lithofacies type and distribution; • A relatively high matrix permeability; • Good vertical and lateral communication between matrix, stratiform Super-K and fractures; • Low pressure gradients in the reservoir due to relatively low depletion rates; • Maximized effect of gravity segregation; • A simple waterflood case with regular oil. In more heterogeneous reservoirs, it can be expected that numerous geostatistical realizations of the lithofacies matrix model as well as the fracture stochastic modeling will have to be performed. The elaborate upscaling process might become too cumbersome and explicit modeling (DPDP) will certainly 4 Z.A. AL-ALI, B.A. STENGER SPE 71340 seem a desirable alternative. As such, existing reservoir simulators allowing mixed formulation , e.g. DPDP regions embedded in an otherwise SPSP reservoir model, are an effective solution although a rarity on the market. The upscaling process remains simple while only the necessary part of the reservoir model is treated as DPDP, thus keeping CPU times to acceptable levels. Note: the issue of fluid reequilibration in fracture cells and its adverse effect on numerical stability remains entire and generally requires setting very low vertical permeability in fracture cells. Finally, direct fracture representation using Local Grid Refinement (LGR) was tried extensively (10) but its interest is probably confined to testing the sensitivity of different fracture configuration and properties. In our study area, three LGR were introduced to model the effect of conductive fractures intersected by horizontal wells (Figure-17). Sensitivities were performed on the fracture conductivity, location and well production rate. In adverse situations such as horizontal wells intersecting conductive fractures connected to the flank aquifer, a rapid water breakthrough and watering out of the well would be witnessed (Figure-18). Pre-emptive measures such as avoidance of heavily fractured areas on flanks, drilling of horizontal sections parallel to the azimuth of conductive fracture swarms are being utilized. Remedial measures such as running liner with External Casing Packer (ECP) to selectively isolate flooded intervals are also being evaluated. Conclusions 1. The integration of fracture and Super-K description is a powerful methodology to obtain reservoir models reproducing the observed production behavior. 2. In our study area, explicit (DPDP, LGR) and implicit (SPSP) ways of modeling permeability heterogeneities lead to successful results. 3. A trade-off exists between explicit modeling (DPDP, LGR), implicit modeling (SPSP), CPU time and ease of model update. 4. Although MPP simulations have made tremendous advances, mixed formulation reservoir simulators seem a desirable tool for reservoirs with discrete permeability heterogeneities such as stratiform Super-K or fracture swarms. Acknowledgments The authors thank Saudi Aramco and the Saudi Arabian Ministry of Petroleum and Mineral Resources for permission to publish this paper. Nomenclature H= reservoir thickness, L, ft K= absolute reservoir permeability, L, md k= relative permeability, fraction, adimensional μ= fluid viscosity, L.M.T, cP Subscript eff=effective r=relative