Abstract
Figures
Introduction
Methodology
Example
Conclusions
References




















Abstract
Figures
Introduction
Methodology
Example
Conclusions
References




















Abstract
Figures
Introduction
Methodology
Example
Conclusions
References




















Abstract
Figures
Introduction
Methodology
Example
Conclusions
References




















Abstract
Figures
Introduction
Methodology
Example
Conclusions
References




















Abstract
Figures
Introduction
Methodology
Example
Conclusions
References




















Abstract
Figures
Introduction
Methodology
Example
Conclusions
References




















Abstract
Figures
Introduction
Methodology
Example
Conclusions
References




















Abstract
Figures
Introduction
Methodology
Example
Conclusions
References

 Introduction

Faults can play a major role in restricting or enhancing communication within a reservoir. Unfortunately, the accuracy with which it is possible to define both the properties (thickness and permeability) and the geometry is variable. In a significant number of cases, modifications of the geometry within realistic uncertainty bounds has a marked effect on the simulated performance of the reservoir (e.g. Freeman et al., 2008a, b). Integrating these structural uncertainties is therefore critical if the true variations in potential flow patterns are to be understood. Several reservoir modelling packages allow the incorporation of fault rock properties in the modelling and simulation process (e.g. Petrel^TM, ROXAR RMS^TM). It is therefore relatively straightforward to perform a series of simulations with modified fault rock properties and discern the impact that this has (e.g. Manzocchi et al., 2008). What is more challenging is the realistic inclusion of geometric uncertainties.

The offset modelled in a geocellular grid will be uncertain for a number of reasons. These include: (i) seismic imaging quality or noise; (ii) acoustic impedance variations; (iii) seismic imaging resolution (lateral and vertical); and (iv) Fresnel zone imaging limitations. These factors mean that the architecture defined around a fault from seismic data will not be a direct image of the geological architecture. To further complicate the situation, fault geometries are heterogeneous and a degree of the deformation is usually accommodated in the wall rocks to the main fault plane. This is usually displayed as folding and/or minor faults (damage zones) around the main slip surface (Figure 1). This distributed deformation cannot usually be imaged by seismic data. In an attempt to correct for these factors the geocellular modeling process usually ignores the form of seismic markers immediately adjacent to a fault. Instead, it uses data from outboard of the fault and projects this into the modelled fault surface. The architecture generated by this process, which is inherent in the majority of reservoir simulation grids currently in use, is clearly therefore uncertain. Vertical uncertainties of tens of metres are believed to be common. This amount of uncertainty is often believed to be irrelevant on the bulk reservoir scale because it does not greatly affect the overall overlap area of the bulk reservoir:reservoir juxtaposition. These geometric uncertainties are therefore often overlooked later in the process. Although this uncertainty does not drastically affect the bulk form of reservoir juxtapositions it does affect the detailed distribution of permeability connections that are developed across the fault. In the majority of cases where the permeability is heterogeneous within the reservoir and the vertical permeability is relatively low in comparison to the horizontal permeability, this change in permeability connections across the fault can have a significant impact on simulated cross-fault fluid flow.

In this contribution we present data that shows the scale of the impact of modifying the throw on faults in a high net-to-gross reservoir. The system of throw modification developed and implemented within Petrel^TM allows fault offsets within fully populated geocellular grids to be automatically modified on-the-fly and the results rapidly fed through to simulation. The throw modification typically has a significant impact on flow. The technique developed allows for improved geometric uncertainty to be routinely incorporated in the day-to-day reservoir evaluation process.

Methodology

Previous attempts to incorporate structural uncertainty in reservoir models have tended to require the generation of multiple structural architectures or grids as one of the earliest steps in the reservoir modelling process. Each of the structural models is then populated with horizons and stratigraphies. This route is manually intensive, complex to track and time consuming. The preferred means of uncertainty incorporation would be more automatic. The approach taken here is to operate on the final populated geocellular grid and modify the throws automatically within that grid. This has the advantage that it is quick to apply and integrate within the simulator and that the time consuming stratigraphic and petrophysical property population stages have already occurred. The approach allows numerous structural architectures to be compared for the same stratigraphic model. The implementation used allows throws to be modified between different structures and to vary along individual structures. The throws can be varied as static bulk shifts (e.g. +/–10m), percentage changes (e.g. 80% of the original throw) or variably across the area using spatial variability rules. Figure 2 shows an example of the process. The geocellular grid is only modified immediately adjacent to the fault. This has the advantage that only minor modifications are required on the simulation deck to integrate the modified grid.

To further simplify the integration process in the simulation workflow an alternative approach has also been implemented. Rather than physically modifying the geocellular grid the throw modification process leaves the grid unchanged but outputs non-physical non-neighbour connections that capture the change in juxtapositions developed. These new connections can be added into the simulator as new INCLUDE files while keeping all of the rest of the simulation files unchanged. With this approach the physical geometry is not altered, so well placements within the grid are still exactly honoured even if they occur immediately adjacent to faults.

The degree to which the throw should be modified is a source of uncertainty. Outcrop data can help to inform on the amount of throw or the range of uncertainties that should be applied. Figure 3 shows the relationship between the main slip surface throw within a fault zone compared to the net throw seen on all of the fault elements within the fault zone. The graph shows that the main slip surface commonly accommodates 50-95% of the throw of the fault zone.

In general, the observed relationship is that the throw modelled by the typical re-projection process used in the geocellular modelling workflow will tend to be equivalent to the actual throw or over-estimate the throw on the fault (usually over-estimate). The Fresnel zone imaging issues in seismic data will also tend to have the same impact. In general, if any accommodation of deformation occurs around the main slip surface then it will tend to reduce the offset on the main slip surface compared to the total horizon offset observed across the fault zone.

Example

To evaluate the flow impact of these relationships a geocellular model has been taken and the throws modified to between 50-120% of the original throws on the faults. These models have then been taken forward for reservoir simulation. In this suite of models the same water injector-producer pair was used. The permeabilities populated in these models were the same and ranged between 10-300 mD. The stratigraphy is a high net-to-gross system. Figure 4 shows the water breakthrough times for the different simulation runs. The graph shows that the higher throws generated earlier water breakthroughs at the producer whereas the low throw models generated later breakthroughs. The higher throws led to greater offsets and the isolation of sections of the reservoir immediately adjacent to the faults. The same volume of fluid flow therefore occurred through a smaller total volume of the reservoir, leading to more localized flow between the injector and the producer.

Conclusions

Geometric structural uncertainty can have a large impact on simulated reservoir performance. Historically it has been difficult to integrate this component in the usual modelling workflow. In this contribution we present a technique that provides the ability to rapidly modify the architecture of existing geocellular reservoir models and pass those forward for flow simulation. This allows the impact of this uncertainty to be rapidly and routinely evaluated. By changing only the geometry adjacent to the faults or alternatively only changing the fault connections data within the simulator, this allows only minor changes to be applied to the simulation deck prior to repeat simulations. This provides a time efficient technique to conduct this type of analysis. The typical seismic imaging limitations combined with the accommodation of strain in the wall rocks to main slip surfaces within fault zones both combine to generate reservoir simulation models that tend to over-estimate the true throw along structures. Reducing the throws to ~70% of the original throw appears to be a useful guide. Integrating this scale of throw reduction within simulation models appears to have a marked effect on the modelled simulated performance. Even in a relatively high permeability, high net-to-gross reservoir model, these minor throw variations can induce significant changes in modelled water breakthrough times (~10-20%). It should be noted that reducing the throw on the faults within our simulation model had the effect of enhancing the total connectivity of the system and led to more efficient sweep patterns than were predicted by having the original faults in the model as completely open to flow (no fault rock) or removing the faults completely from the grid. This workflow provides an efficient mechanism to routinely define the likely impact of structural uncertainty on reservoir simulation models.

References

Freeman, S.R., S.D. Harris, and R.J. Knipe, 2008a, Fault seal mapping - incorporating geometric and property uncertainty, in A. Robinson, P. Griffiths, S. Price, J. Hegre, and A. Muggeridge, (eds), The Future of Geological Modelling in Hydrocarbon Development, The Geological Society, London, Special Publications 309, p. 5–38.

Freeman, S.R., S.D. Harris, and R.J. Knipe, 2008b, Integrating detailed cross-fault fluid flow behaviour into existing reservoir simulation models, AAPG International Conference & Exhibition, Cape Town. http://www.searchanddiscovery.net/abstracts/html/2008/intl_capetown/abstracts/471534.htm

Manzocchi, T., A.E. Heath, B. Palananthakumar, C. Childs, and J.J. Walsh, 2008, Faults in conventional flow simulation models: a consideration of representational assumptions and geological uncertainties Petroleum Geoscience, v. 14, p. 91-110.

Return to top.

Copyright © AAPG. Serial rights given by author. For all other rights contact author directly.