Conceptual model building provides a key to success in fractured reservoir characterization.

There is a common misconception in fractured reservoir modeling that "all fractures are equal." This often drives the characterization process down the road of fracture prediction as the quest becomes simply to find more fractures without really paying attention to how the fractures are actually behaving. Experience has shown that it is not just the presence of fractures that is important, it is their dynamic behavior and connectivity of the fracture system that ultimately determines their contribution to overall reservoir performance. It is this recognition that not all fractures are equal that motivates us to invest effort in developing a conceptual fracture model: a basic framework of understanding of the likely geometric and dynamic properties of the various fracture components seen within the reservoir. It is only through the development of a verifiable conceptual fracture model that reservoir planning decisions can be optimized and the dynamic effects of the fractures fully implemented within the reservoir simulator.

As part of a recently completed reservoir characterization study within one of the Cretaceous reservoir units of the Awali Field in Bahrain, a major effort was made to develop a suitably conditioned fracture model of the Amadi reservoir, a tight, thin, naturally fractured carbonate. Having little reservoir thickness and a predominance of vertical wells, there was little in the way of direct fracture information with which to develop the fracture model.

Instead of focusing on a model of fracture prediction, an approach was decided on that involved the development of a conceptual model that described the various fracture components in terms of their geometry and behavior. This understanding was tested by building discrete fracture network (DFN) models that allowed the integration of available dynamic data with the static fracture geometries. The ultimate goal was to help constrain the simulation model for this unit, but it was also important to be able to make reservoir planning and development decisions to maximize fracture production.

Background

The Awali Field lies beneath the center of Bahrain island, to the east of Saudi Arabia and to the west of the Qatar arch (Figure 1). The Bahrain field was discovered in 1932 and is the oldest field in the Gulf. The broad geological structure of the Awali reservoirs consists of an asymmetric elliptical dome that is elongated in the north-south (N-S) direction. The fault pattern is characterized by a N-S graben with extensional faults along the crest of the dome. According to a paper by J.A. Samahiji and A.N. Chaube, an additional set of later-formed NW-trending faults are seen on the flanks of the reservoir. There is some debate about the origin of the Awali anticline, with the traditional explanation being it is induced by salt doming with a more recent explanation being that the anticline represents a drape fold over basement faulting. The main productive levels within the Awali Field are the Cretaceous Wasia Group, Jurassic Arab and Permian Khuff formations. Full reservoir characterization of a number of the productive reservoirs was being undertaken by CGG, with the fracture characterization work being undertaken by Golder Associates and focused during this phase on the Cretaceous Ahmadi carbonate reservoir.

The starting point of conceptual fracture model development is to identify the broad structural drivers of the various fracture components thought to be present and hydraulically significant within the reservoir. In terms of the Awali Field, this is not straightforward because of the uncertain origin of the anticline structure and fault system. However, it is by understanding the structural drivers that create the fracture systems that it becomes possible to predict the likely small-scale geometries that may be present
There are several routes by which fracture intensity and orientation have been predicted across a structure when there is little direct fracture data, including techniques such as curvature analysis, palinspastic reconstruction or elastic dislocation modeling. However, all of these different approaches are based on the belief that the contribution of fractures to reservoir performance is related to the broad shape or deformation of the structure, automatically assuming that all fractures are hydraulically important and fracture intensity ultimately controls behavior. It is equally likely that the different fracture components present will have vastly different hydraulic behaviors that are not controlled by the shape of the structure. An approach is needed that can resolve the hydraulic behavior of the various fracture components in order to help determine the appropriate conceptual fracture model. This is achieved firstly by looking through the available evidence before synthesizing this into a conceptual understanding.

Fracture data analysis

As with many mature fractured reservoirs, the Awali field suffers from a lack of direct fracture data due to both the reservoir development predating the widespread introduction of image log technology and also because vertical wells are very poor at sampling sub-vertical fractures. However, there is still a considerable amount of information that can help determine how the fractures are influencing the reservoir performance.

Fracture Data. Available fracture data for the Awali Field is limited to three sources: wellbore fracture data and core and surface analogs. The limited image log data revealed a small number of steeply dipping fault parallel fractures also seen in the cores. Surface mapping has revealed a pair of orthogonal joint sets trending NE-SW and NW-SE, according to a report presented by Yahya Al Ansari Hustedt and J. Mattner. This joint system was regularly spaced at a distance of approximate bed thickness with a long-length scale in the NE-SW direction. A series of short orthogonal joints provided limited connectivity of the fracture system in the NW-SE direction.

Well Test Data. Well tests provide the best indirect assessment of the nature of a fracture network present within a reservoir. A pressure transient test provides the only method of investigating the fracture network away from the well, and the response of the pressure derivative curve has been shown to be a diagnostic tool for characterizing the nature of the fracture network (Figure 2). A considerable number of well tests were available from across the reservoir structure, and by plotting up the well test curves in a rate-normalized fashion not only can the overall transmissivity of each well test be compared but also the style of flow. This quantitative method of well test analysis has proved critical in helping to develop the fracture conceptual model and also to resolve the spatial extent of the conceptual models. From this analysis it was seen that there were two broad styles of pressure derivative that corresponded to the flanks and graben of the structure. Of course, well tests are non-unique and need to be considered within the context of the likely fracture geometries seen by the test. The seismic fault system was considered permeable based upon the presence of late-stage constant pressure boundaries in the well test data and also the presence of the highest transmissivity values being located proximal to major faults.

Production Data. Looking at trends in production data provided a first-order approximation of the broad connectivity of the reservoir. Analysis of a number of clusters of wells in close proximity showed that there was a consistent relationship between wells on a NE-SW trend which was believed to be caused by the presence of the layer-bound jointing seen in the surface outcrop. This was also in agreement with a gas breakthrough that traveled almost 4,000 ft (1,200 m) between two wells along the same trend in 7 days. The production correlations consistently extended across the NW-SE faults, suggesting the joint system has a limited interaction with the seismic faults.

Conceptual model development and testing

All of the data analysis provided insight into the geometry and dynamic behavior of the fracture network, with skill being required to weigh up which evidence is important and needs including within the conceptual model and which doesn't. The main elements of the fracture system as revealed from the data analysis are:

• The main fracture system appears to be a NE-SW layer-bound joint system.
• The joint system is thought to have a limited connectivity in the NW-SE direction.
• The fault parallel fractures are thought to be hydraulically insignificant because they provide no connection to the reservoirs above or below and also no connectivity along the trend of the fault.
The seismic faults are conductive and link the overlying reservoir together.

Once the elements of the conceptual model are in place, this can be turned into a DFN model that captures these key components of the fracture model. DFN models are increasingly being seen as a highly effective way to model fractured reservoirs where complex geometries result in unusual and often highly directional flow regimes. According to an SPE paper by B.P. Dershowitz et al, 1998, it is their ability to accurately capture the complex spatial distributions and connectivity of fractured reservoirs that make them more effective than traditional continuum methods. In the case of the Ahmadi reservoir, a series of DFN realizations were developed that represented the different structural domains present within the reservoir, namely the flanks and graben. On the flanks of the Awali structure, the fracture network appears to be represented by a poorly connected system of layer-bound joints that connect to a system of conductive seismic faults (Figure 3). Within the crestal graben, a network of connected faults was believed to provide the main fracture contribution.

Ultimately the only way to gain confidence in the conceptual model is to test it against available dynamic data. One of the strengths of the DFN approach is the ability to convert the models to flow grids, allowing the simulation of well tests through them. According to an SPE paper by L. Wei, this is a critical step as the geometry of the fracture network has been shown to influence the pressure derivative from a well test. By matching the actual and simulated pressure response of a well test, the geometry of the static model can be tested against the dynamic data, increasing the confidence of the conceptual model. Figure 4 shows pressure snapshots of a well test with the pressure seen to be diffusing through the poorly connected joint system until it reaches the conductive fault system.

This process was repeated for a series of well scale models on both the flanks of the structure and within the crestal graben, enabling the conceptual fracture model to be tested against a range of dynamic data. It was found that these two broad conceptual models were capable of producing the correct dynamic response and also enabled a restricted range of engineering properties for each of the main fracture components to be derived, namely the layer bound fractures, the NW-SE oriented faults and the N-S oriented faults. These engineering properties were implemented within the simulator model by defining zones where the fracture components were present and applying systematic permeability multipliers to the matrix values based upon the results of the DFN simulation. This approach, concentrating firstly upon understanding fracture geometry and behavior, resulted in an accurate and accelerated history match of reservoir performance.

The conceptual model isn't static and needs to be constantly tested against available data. The current model provides a framework of understanding based upon available static and dynamic data. However, it is always important to refine, update or even reject the model as new data are acquired. As the certainty of the model increases, it becomes appropriate to use the 3-D fracture prediction methods described earlier as they become a sensitive tool for reservoir modeling rather than the coarse instrument of blind characterization when applied without appropriate forethought.

Comments