It is well known that the permeability of porous media represents a first order control on fluid flow in hydrocarbon reservoirs and that the magnitude of the permeability often depends on the rock volume under consideration. Core plug permeability does not necessarily equal whole core permeability and permeability from core may not necessarily reflect well test permeability. Each of these disparate data sources measures permeability at a particular scale. For the purposes of reservoir modelling, the permeability systems characterizing a giant offshore oil field have been broadly categorized as either matrix permeability or excess permeability. Matrix permeability further subdivides into two categories based on the abundance and type of microporosity. Excess permeability subdivides into three sub-categories depending whether it is the result of depositional processes emplacing anomalously high permeability storm beds (HKS or high permeability streaks), diagenetic processes creating dissolution enhancement of permeability, or fractures. Although not all these permeability systems are active in any reservoir interval, each reservoir interval possesses at least two if not three of these systems. The multi-scale nature of permeability arises because of 1) differences in the spatial extent of these permeability systems and 2) permeability contrasts between the systems.
Several techniques have been developed and will be explored in this paper that attempt to account for the influence of multi-scale permeability systems on reservoir performance behaviour. In what might be the simplest case, the mixture of permeability systems includes only matrix permeability without significant microporosity and excess permeability resulting from HKS. In this case, matrix permeability was modelled independently of excess permeability creating significant short-range permeability contrasts that better predicted reservoir pressures and water movement in the reservoir during history-matching. In another case with the same type of matrix permeability, the excess permeability represents the contributions from a mixture of fractures and HKS. In this case, matrix permeability was also modelled independently of excess permeability. Estimates of the relative contributions of HKS and fractures to excess permeability were tested as a history-matching parameter. Ultimately, this approach to characterizing permeability attempts to capture some of the rudimentary aspects of a dual permeability model without incurring the associated computational expense.
Several giant carbonate reservoirs have undergone decades of waterflooding, and are now transitioning to EOR recovery processes. Simulation models that were calibrated via history matching while undergoing a waterflood (i.e.two phase flow
performance) are utilized now to predict three phase flow performance encountered with EOR processes. How reliable are these predictions? Are they accurate enough to be used for business decisions?
In this work, validity and reliability of simulation models, that has been history matched by two-phase flow processes of water flooding, to predict the performance of three-phase flow of WAG processes was assessed.
To accomplish study objective, fine grid of two 5-spot sectors model was built and then upscaled. Upscaled model was then history matched to the results obtained from the fine model using water flooding data and utilizing pseudo functions data. The
resulted cases as well as the fine model were then taken to prediction to estimate the performance of three-phase flow of gas and WAG processes. Results of fine and coarse models were then analyzed and compared to draw conclusions on the
reliability of the coarse models to match the predicted results of the reference model of the fine simulation model
Oil reservoirs are very complex systems with flow properties varying from the pore to reservoir scale. To simulate fluid flow in the reservoir, there are many uncertainties ranging from the spatial distribution of basic rock properties to the quantification
and impact of rock/fluid SCAL models of wettability and associated hysteresis. The process of history matching attempts to reasonably reproduce the past field performance by fine tuning some or all of these uncertain parameters while reasonably
maintaining their physical nature in the simulation model. It is well known that well history matched simulation models are more trustworthy in predicting future reservoir performance. Errors in these future performance predictions may be introduced,
however, when future development plan include processes that were not experienced historically.
The main purpose for using pseudo functions is to reduce the number of grid cells of reservoir models, trying to reproduce the behaviour of the fine scale system with coarser models. More specific, pseudo functions have been used historically to history
match the fine grid models into one representative upscaled model. This model (the upscaled one) should be used for further field development which employs the prediction of reservoir performance under different depletion and operating scenarios.
Most of the found literature was focused on simplifying multi-dimensional (2D) or (3D) systems to 1-dimensional (1D) models 1,2,3,4. Very limited number of grid blocks where allowed by the available computers at that time. The use of pseudo
relative permeabilities and capillary pressures was one way of decreasing grid dimensions into a more tractable level with minimal loss of simulation accuracy.
Another reason was to account for numerical dispersion that occurs through upscaling process. The upscaled relative permeabilities (pseudos) can compensate for the increase in numerical dispersion as the grid is coarsened 2,5,6.
Sustaining hydrocarbon production to meet growing energy demand is an industry-wide challenge, especially for Middle East projects. Consequently the need for sound, robust models is increasingly important to accurately represent the subsurface ahead of field development/production. A common hurdle to overcome involves scaling core-based geologic characterization to models, such as representing thin high perm streaks or populating facies mosaics. This paper summarizes the approach used to characterize and model a thick carbonate reservoir from a giant offshore Abu Dhabi oil field.
Core descriptions from 36 key wells across the field establish the setting as a structurally modified ramp characterized by patchy facies distributions that was frequently reworked by storms. Core:log calibration was investigated but deemed not applicable for this reservoir, thus core was the primary control on the rock type model. High permeability grainstones and algal/rudist floatstone interpreted as storm deposits are thin (<1 foot) and not correlative between cored wells. The wells were rock typed and mapped to reveal trends between rock types, thickness and reservoir quality.
These trends were integrated with previous learnings and modern analogs to help bridge between the core and model scales, such as for geobody dimensions and spatial variability within the model. Deterministically-guided stochastic distribution of rock types using Truncated Gaussian Simulation was the preferred method to build the rock type model based on data coverage and the vertical stack of facies observed in core. Fine-scale layering in certain zones of the geomodel was necessary to capture the thin high perm streaks. This necessity led to coarser model layers, where applicable, to maintain a manageable geomodel size. Distribution and connectivity of high perm geobodies were guided by multiple feedback iterations from history matched results. These iterations were key to achieve a sound, robust model.
Geologic Setting and Field Overview
The carbonate reservoir of interest, termed Res2, is situated within a giant oil field in offshore Abu Dhabi. It is one of several reservoirs in the field that was discovered in the early 1960s. Res2 is part of the Early Cretaceous Kharaib Formation that is productive throughout the UAE area. Presently the field has more than 500 wellbores in Res2 and has produced for more than 30 years, yet the production life is still in its very early stage. The average thickness of Res2 is 140 feet and is primarily composed of shallow water carbonate facies that are partitioned into 6 zones (termed Zones 1-6) by five thin stylolite intervals (termed Stylolite A-E). Res2 averages 20-25% porosity over much of the field and matrix permeabilities generally range from 1's to 100's mD. Diagenetic enhancement of the reservoir is limited to a thin dolomitized interval in one of the lower subzones. Calcite cementation is minimal in the main part of the field but increases volumetrically away from the structural crest. Fractures are very minor in Res2 and are limited to the very top and base of the reservoir.
A new generation geologic model for a giant Middle East carbonate reservoir was constructed and history matched with the objectives of creating a model suitable for full field prediction and sector level drill well planning. Several key performance drivers were recognized as important factors in the history match; 1) unique carbonate fluid displacement; 2) data validation and horizontal well trajectory issues; and 3) distribution of high permeability streaks. Ultimately a full field history match consisting of more than 1000 well strings and several decades of history was achieved using detailed distribution of the high permeability streaks, while honoring measured core poro-perm relationships, lab-validated displacement curves, and well test data.
This paper discusses the role of the geometry and the vertical distribution of the high-permeability streaks in the history matching of a giant offshore carbonate reservoir. Specifically, the modeling of the high-permeability streaks - which consist of thin rudist and algal rudstone, floatstone, and peloidal grainstone, with abundant, well-connected inter-particle porosity - became possible after extensive revamping of the reservoir rock type model, updating well descriptions, and a detailed zonal mapping of the high permeability streaks and dolomitic zones. The areal and vertical model resolution was doubled over the previous models to accommodate the internal sub-layering of the upper four reservoir zones in order to capture the thin (~1.4 ft) high-permeability streaks.
During the history match, local modifications of the high-permeability streaks were the integral part of the feedback loop between the simulation engineers and geoscientists that kept the common-scale simulation model and geologic model synchronized. The final history match was validated by extensive analysis of the models' vertical conformance as compared to production logs. This approach made it possible to construct a more heterogeneous model than previous models; while honoring both field KH and matrix poro-permeability without local permeability multipliers. The combination of these features provides a higher confidence model of long term well injectivity/productivity.
The subject reservoir is a giant offshore carbonate reservoir deposited in an extensive, low to moderate energy, low-angle ramp setting that stack into an overall shallowing-upward carbonate depositional sequence. It is overall mud-dominated from the base of the reservoir to the middle zone and becomes grain-dominated from the upper portion to the top of the reservoir. Major reservoir rocks include: mud-dominated (mudstone and wackestone), packstone, grainstone, algal-dominated floatstone, rudist-dominated floatstone, dolomite, and a thin generic "high-permeability streaks?? in ascending reservoir quality.
The reservoir was operated under primary depletion for over a decade before being converted to a pattern waterflood and then eventually to a line drive as its current depletion plan. Recent field development activity necessitated constructing and history matching a new generation model using the lessons from the previous model and emphasis on understanding and capturing the injected water movement including vertical conformance. It includes the latest seismic interpretation, a revamped reservoir rock typing model (Al Ameri 2011), updated well descriptions, and a detailed mapping of the high permeability streaks and dolomitic zones (cf. Yamamoto 2011).
Yamamoto, Kazuyuki (Zakum Development Co.) | Kompanik, Gary (ZADCO) | Brantferger, Ken (ZADCO) | Al Zinati, Osama (ZADCO) | Ottinger, Gary (ZADCO) | Al Ali, Abdulla (Zakum Development Co.) | Dodge, Scott (Zakum Development Co.) | Edwards, Henry Ewart
This paper presents a method to condition the permeability modeling of a thin, heterogeneous high-K dolomitized unit. The interval is an important drilling target for field development, so precise permeability modeling is required to optimize well placement and completion designs in order to maximize oil recovery and minimize early water breakthrough.
Detailed core observations from 85 wells classify the unit into two groups: Group A, composed mainly of dolostone and Group B, comprised exclusively of calcareous dolostone. Regression analyses of plug porosity-permeability values are characterized by one regression line for each group by which dolostone represents a higher permeability trend relative to calcareous dolostone. Core-plug scaling is used to scale-up the porosity-permeability relationships from core plug- to model-scale (100 m by 100 m cells). The two regression lines accurately capture the permeability contrast within the dolomitized unit.
To extend the method into a full-field model, it is necessary to calibrate the well logs to the core data. Comparison of cores with various log responses indicates the porosity log is the most useful tool to achieve this. Group A, characterized by higher dolomite content, is distinguished by a distinct decrease in the porosity due to progressive dolomitization.
Porosity logs from 499 wells are interpreted and permeability values are assigned using the regression lines based on the detailed distribution map of both groups. The modeling approach using hundreds of well logs calibrated to cores yields a more detailed picture of the spatial permeability variations of the dolomitized unit. Dynamic data from ongoing history matching is also used to implicitly adjust the first-pass static model.
Brantferger, Kenneth M. (Zakum Development Co.) | Al-Jenaibi, Haitham (ZADCO) | Patel, Harshad (ZADCO) | Al-Harbi, Amal Saeed (Zakum Development Co.) | Kompanik, Gary (ZADCO) | Mubarak, Magdi Ibrahim (ZADCO Petroleum Co)
Yamamoto, Kazuyuki (Zakum Development Co.) | Al Zinati, Osama (ZADCO) | Ottinger, Gary (ZADCO) | Edwards, Henry Ewart (Zakum Development Co.) | Kompanik, Gary (ZADCO) | Al Ameri, Mohamed Braik (Zakum Development Co.)