3D stochastic modelling of fault zones in siliciclastic reservoirs. Implications for reservoir description and fluid flow modelling
Not peer reviewed
MetadataShow full item record
Fault zones in siliciclastic rocks form distinct volumetric entities and typically exhibit great structural complexity. While it has been considered that the main impact of fault zones to fluid flow is to baffle or block the flow, observations suggest that fault zones can (i) act as dual baffle-conduit conduit systems, (ii) host long-distance flow in both along-dip and along-strike directions, (iii) act as pure conduit systems exhibiting higher flow rates than the surrounding host rock. However, the complex flow behaviours inside fault zones cannot be readily captured or forecast in traditional reservoir models, as faults are conventionally implemented as 2D modelling objects. A comprehensive and integrated approach has been adapted for providing a new feasible method, i.e. “fault facies modelling”, for volumetric fault representation in industrial reservoir models. To date, research on fault facies modelling is in many ways still in its early stages of development, with emphasis being placed on establishing frameworks for fault zone characterization suitable for modelling purposes and modelling algorithms for implementing them. The principal aims of the present study are to:
1. Improve the key aspects of fault facies modelling, i.e. the use of outcrop-based facies maps, 3D gridding, property modelling, upscaling, and application on field-scale models.
2. Advance current knowledge on the impact of fault zone structure on reservoir fluid flow.
To increase the capability of fault facies modelling with regard to reproducing outcrop observations, a new way of utilizing outcrop observations is introduced. Prior to modelling, outcrop-based fault facies maps are created by discretising outcrop observations using grid resolutions that are deemed to be high enough to resolve fault facies units depending on flow scale considered. The maps form conceptual templates for (i) selecting the most suitable stochastic modelling techniques and workflows, (ii) establishing representative geostatistical properties of fault facies and fault facies associations, (iii) assessing the quality of the resulting fault facies models.
With respect to fault zone property (facies and petrophysical) modelling, the present study focuses on the applications of fault zone displacement functions and stochastic modelling methods. Automated scripts are employed to define fault zone displacement functions to allow more flexibility in the application of the functions. This, in turn, makes it easier to handle outcrop-based statistics such as (i) damage zone width as a function of fault throw, (ii) fault core thickness as a function of fault throw (iii) the fractions of total fault throw accommodated by fault core and damage zones, and (iv) the types of displacement function (quadratic, cubic, and quartic) in fault core and damage zones.
Truncated Gaussian simulation (TGS) is used for modelling damage zone features. This has not been tried before. In different fault system configurations, i.e.: (i) single isolated faults, (ii) single-tip interacting faults (branching of two fault segments), and (iii) double-tip interacting faults forming a relay ramp, the applications of TGS are constrained by conceptual and outcrop-based conditioning parameters. Comparison between the resulting models and the discretized damage zone observations show that important observed characteristics such as variations of fault facies thickness, -length, - adjacency, and compartment geometries are reproduced by the models.
For fault core modelling, a hierarchical approach employing object-based simulation technique is demonstrated. To reproduce a complex fault core lens configuration where lenses are stacked one to another, lensoid objects are populated by setting very high volumetric proportion of the simulated objects (hence, very low background volumetric proportion). Subsequently, the lens models are refined non-proportionally (for creating thin and thick cells) and slip zones are implemented in thin cells separating two different lensoid objects. Comparison between the resulting models and the discretized fault core observation show that the models reproduce important fault core characteristics such as (i) lenses with varied thickness and aspect ratio and (ii) lens size distribution that shows a higher number of smaller lenses, and (iii) the distribution of slip zones, i.e. flow baffles/conduits, that 9 is closely related to the lenses distribution. Moreover, combined with flow-based upscaling, the hierarchical approach makes it possible for employing reliable and accurate fault facies modelling on a full-field scale within the framework of existing reservoir modelling tools.
Work performed on flow simulation and sensitivity testing and analysis have provided new insights into the impact of different fault zone parameters on reservoir fluid flow. When both fault core and surrounding damage zones are considered, fault facies modelling parameters can be ranked according to their impact on reservoir responses in the following descending order: (i) fault core thickness, (ii) the type of displacement function, (iii) sedimentary facies configuration, (iv) fault core throw percentage, (v) fault system configuration, (vi) maximum damage zone width. When considering the damage zone separately, the impact of fault facies modelling factors can be ranked in descending order as: (i) fault facies volumetric proportion, (ii) deformation band frequency, (iii) damage zone width, (iv) deformation band permeability, and (v) fault facies cluster extent.
Flow simulation study performed on outcrop-based relay zone models corroborates conclusions from earlier studies that the main factor controlling fluid flow in relay zones is deformation band permeability. Observations on variations in shape and migration velocities of fluid fronts in the relay zone show that, within khr/kdb range of 101 to 105 (where khr and kdb are host rock and deformation band permeability, respectively), increasing khr/kdb causes increased fluid flow complexity as indicated by amplified deviation from piston-like fluid displacement. High deformation band permeability allows injected water to saturate the entire oil column with ease and hence relay zones with high permeability bands provide good vertical sweep efficiency. For lateral sweep efficiency, however, relay zones with medium deformation band permeability show better performance due to enhanced flow tortuosity in the presence of pervasive but weak flow baffles.
Field-scale flow simulations indicate that fault cores with membrane slip zone clusters act as dual baffle-conduit systems. Sweep efficiency in reservoirs with baffle-conduit fault cores is closely related to injector-producer configuration. Baffle-conduit fault cores subparallel to injector-producer pairs focus injected fluids, hence decreasing sweep efficiency. Baffle-conduit fault cores perpendicular to injector-producer pairs partition and distribute the injected fluids and therefore increase overall sweep efficiency. Reservoirs with baffle-conduit fault cores exhibit two behaviours: (i) water breakthrough occurrence, time, and sequence are not sensitive to the choice of injectionproduction scheme and (ii) sweep efficiency decreases when field production rate is lowered. Fault cores with conduit slip zones, however, act as thief pathways. Sweep efficiency in reservoirs with conduit fault cores is less dependent on injector-producer configuration. Two distinct behaviours of reservoirs with conduit fault cores are: (i) water breakthrough occurrence, time, and sequence are highly sensitive to the choice of injection-production scheme and (ii) sweep efficiency increases when field production rate is lowered. These simulation results show that the improved realism added by incorporating volumetrically expressed fault cores substantially influences forecasts of field behaviour, and consequently should be considered during oil/gas production planning.