Case Study: Integration of Static and Dynamic Data to Improve Geomodels in a Complex Fractured Carbonate Reservoir

Alex Assaf and Richard Steel
BG Group, Reading, UK

The theme will centre on the integration of subsurface static and dynamic production data (both rates and pressure data) to calibrate fractured carbonate geomodels. The case study addresses how workflows were developed where analytical and numerical techniques in pressure transient analysis were used to calibrate the geomodels and improve the history match and forecasting methods.

Geology and Geological Modelling
The carbonate gas condensate field is located in North Africa and is designated as the “H field”. The field is located in a North-Northwest to South-Southeast trending horst block which is bounded by a series of normal to oblique faults. These faults are thought to be highly conductive where severe fluid losses were encountered during drilling. Figure 1 depicts a top structure map with highlights from the four production wells (3 gas wells (A1-A3) and 1 oil well (A4ST3)) and a North-South cross section with a summary table from the four appraisal wells (H1 to H4). The field has a small oil rim overlain by a large gas cap. The main reservoir is comprised of variably bioturbated, dolomitised and fractured nummulitic limestones, capped directly by a transgressive package of the Compact Micrite. However, this unit was found to be heavily fractured in some of the wells and therefore may not be an obvious seal. This is further supported by core and image log data, indicating fracturing may have enhanced both horizontal and vertical permeability. The thickness of the reservoir varies between 40m and 60m and a stratigraphic pinchout is inferred to the North of the structure.

Dolomitisation is evident in wackstone to packstone units and this has lead to significant improvement to permeability. A key to understanding the reservoir potential is the estimation of the amount of dolomite present. As the limestones are generally clean (clay free), core grain density values were used to estimate the amount of dolomite present, while ICP-AES analysis was used to measure the amounts of dolomite by accurate elemental analysis. There is a strong correlation between facies and the distribution of dolomite, with originally mud-supported lithologies exhibiting the highest percentages and therefore the best poro-perm values. The primary zone of dolomitisation corresponds to reservoir Layer 3 (L3) in the model used. This observation together with the well performance, lead to a different geological interpretation and model build which is not the subject of this presentation.

Fracturing has also been identified as an important aspect of the reservoir description which has been a major contributing factor to the overall deliverability of the development wells. Comprehensive fracture studies on the appraisal well cores from H-2, H-3 and H-4 were integrated with the structural interpretation of CAST and FMI images. A complex pattern emerged whereby fracture density can be related to the formation of stylolites due to burial and compaction, and more importantly, proximity to NNW-SSE trending faults. It is evident from core analysis that the large numbers of partially open fractures within the hydrocarbon bearing formation are likely to enhance vertical permeability substantially. In addition, fracture systems related to faulting are locally important in terms of permeability. Intra-particle porosity, within nummulites tests, is common in all nummulitic facies, but it is often not effective unless compaction in grain-supported units has fractured the tests.

The formation has been divided into four main hydrocarbon bearing zones (L1 to L4). There are also two additional zones with small net reservoir units (L5 and L6) and these were not considered in this study as they appear to be not in communication with the main reservoir units (L1-L4).

The main structure is of Miocene age and it appears that the oil and gas were sourced and generated from the same source. The migration history and paths appear to be quite complex and there is a strong possibility that some form of differential migration has taken place. The trap mechanism appears to be structural in the East and West direction through possible fault seal, but stratigraphic to the North and more diagentic/facies type trap to the South.

The fluid contacts (GOC and OWC) have been interpreted from MDT pressure data and there are some degrees of uncertainty associated with the contacts.

A geological concept was developed based on the premise that fractures are due to a series of highly conductive North to Northwest trending faults. There are also small Northwest and Northeast trending fractures combined with stylolites and short fractures. The fractured model was based on the correlation between fracture intensity and distance from faults. A dual porosity model was built to model the matrix and populate the fractures. A summary of the work flow devised to capture these concepts is presented in Figure-3.

The final field grid model of 110 m x 110 m by 17 layers was built. This model was populated with the petrophyiscal properties based on the wells drilled to 2009 (Four wells during the appraisal phase and three during the development phase). One of the main uncertainties in the modelling of those properties was the water saturation, particularly in the oil leg. Subsequent work indicated the presence of a large transition zone in the oil leg, which was not explicitly modelled here. However, this aspect would only affect the oil well and would not significantly have affected the outcome of this work.

Dynamic Modelling
Since the field is severely heterogeneous, three alternate geomodels were considered:

• A “Fully Compartmentalised Model (FC)” – observed faults are artificially extended to allow complete compartmentalisation and transmissibilities across the faults are assumed to be zero. This model subsequently proved to be of no help as dynamic data indicated communication and well interferences across these faults as seen from the pressure data of the producing wells.
  
• A “Fully Open Fault Model (FO)” – observed faults are modelled as mapped; natural fault terminations are honoured as is the cross-fault transmissibility, calculated by juxtaposition of model permeabilities.
  
• A “Partially Compartmentalised Model (PC)” - as with the FO model but with zero transmissibility across the faults. This allows for tortuous communication around the fault terminations. This model is referred to as the “Superimposed Fault Model”.

Three development gas wells (A1, A2 and A3) were brought onstream. The oil well A4ST3 was also brought online but produced intermittently due to unexpected water production. The fully closed fault model (FC) could not account for the interferences seen during the drilling and subsequent production of the field. The fully open fault model (FO) could account for the interferences but showed too much communication. It should be noted that when the oil well A4 was drilled, severe wellbore stability issues were encountered resulting in many sidetracks before this was successfully drilled. Some 1.5 MM bbls of sea water was lost into the reservoir which resulted in an increase of 47 psi in the virgin pressure of the A1 well, 4 km away. This data, together with interference data during start-up has resulted in greater understanding of the fracturing network. Figure-4 depicts the PC model used in this study.

Well test pressure transient analysis (PTA) indicated a possible compartmentalisation along these faults and/or a high permeable anisotropic and fracture system.

The three gas wells were tested with downhole shut-in and all were equipped with permanent bottom hole pressure (BHP) gauges. A workflow was formulated where analytical pressure transient analysis for the three gas wells were matched against the numerical modelling of the transient behaviour based on the three simulations models built. Measured BHP data were matched using well test interpretation tools and aided in establishing near wellbore properties (permeability, Kv/Kh and skin).

Initially, the analytical interpretation was used to guide the numerical modelling of the well test and to help in the design of the local grid refinement (LGR) grids. LGR grids were built around each of the three gas wells and the well test was numerically simulated using the same well test history used in the analytical interpretation. The results from the numerical simulator (Eclipse) were then imported to the Kappa well test simulation package as if they were gauge data. The data were then compared to the measured BHP data using the log-log derivative plot. If the match is not obtained, then the Eclipse model was modified within the LGR area by varying the permeability and fracture properties until a good match was obtained. Figure-6 and Figure-7 present the outcome of workflow described, on well A2.

The PTA interpretations also helped to establish drainage geometries and whether any barriers (faults) were nearby. If sufficiently long tests are available (unplanned shut downs, close-in of wells), BHP data can also aid in establishing how far the transient response has travelled. This can help in validating the geological model.

Subsequently when more history was obtained, it was found that the results from the numerical well test models indicated that interference played major role. The field had an extended shutdown as a result of on onshore process facilities issues and the long build up data proved a useful tool to calibrate the geological models. This resulted in building one large LGR around the three wells to ensure the simulated data can be honoured to the measured data. Once this was achieved, full history matching was performed on the entire data set and a typical match, from production well A1ST1 is depicted in Figure-8.

Conclusions
The integration of subsurface static and dynamic production data proved invaluable in helping to appropriately characterise this highly heterogeneous carbonate reservoir. Incorporation of the pressure transient analysis and numerical modelling of near wellbore effects in particular, was valuable critical “feedback tool”, helping to improve geological understanding, which in turn led to achieving a better, geologically-appropriate history match. This integration of data has significantly improved forecasting and development planning.

Acknowledgement
The authors wish to thank BG Group for allowing this publication. The geological model and part of the reservoir simulation model pre history match was built with the help of TRACS-AGR.

AAPG Search and Discovery Article #120034©2012 AAPG Hedberg Conference Fundamental Controls on Flow in Carbonates, Saint-Cyr Sur Mer, Provence, France, July 8-13, 2012