Summary This paper presents a field case study of a faulted, multilayer carbonate reservoir to quantify the impact of grid selection on the reservoir history match. Comparison of the history match using both orthogonal and nonorthogonal grids revealed that the simple five-point scheme can lead to significant error in the history match and in the different representations of the reservoir. Introduction A simulation study of the Thamama reservoir, a faulted carbonate reservoir in Abu Dhabi, investigated flow communication across faults and evaluated development options for the field. The field is an elongated anticline that covers 62 sq miles. Geophysical and geological data and early reservoir performance indicated that the field was separated into several fault blocks by partially sealing and nonsealing faults. Two full-field simulation models were constructed that used two different approaches to represent faults:an orthogonal (regular) grid model with faults defined at block boundaries anda nonorthogonal (irregular) grid model with block boundaries defined at fault traces. The first approach results in a greater number of cells and presents difficulties in representing the fault geometry properly. However, finite-difference techniques can be used for numerical calculations. The second approach uses fewer gridblocks and provides an accurate description of the fault pattern, but transmissibility calculations could lead to significant errors because the grid is locally distorted. To quantify the impact of grid selection on the reservoir history match, both models were used and the results compared in the simulation study. History matching was performed first with the orthogonal grid model. Then, a history match using the same matching parameters was conducted on the nonorthogonal grid model. Background The Thamama reservoir comprises three zones, Zones A, B, and C. Characterized by high faulting and low permeability, this carbonate reservoir contains about 3.6 billion STB of oil in place (OIP). Zones B and C, which contain about 80% of the total OIP, have been the focus of major development since the field discovery in 1967. Fig. 1 shows the general structural configuration of the reservoir, which consists of a large dome with faults oriented north-west/southeast. Faulting in the field, which partitions the reservoir into numerous small fault blocks, has caused some concern about implementation of the full-field development plan. Although lateral and vertical communication between upthrown and downthrown faulted blocks along major faults was suspected, it was never fully identified. Throughout the field's producing life, the extent of lateral and vertical communication within the reservoir, particularly between the oil zone and the underlying aquifer, has been questioned. This question became acritical operating consideration when a peripheral water injection plan was proposed in the late 1970's. Although an extensive reservoir study was undertaken to predict the field'sfuture producing performance under different waterflood operations, no waterinjection program was implemented. We decided that further detailed reservoir studies should be conducted to understand the communication and crossflow between fault blocks fully before a full-scale injection program could be designed. A preliminary 3D, coarse, full-field simulation study of the Thamama reservoir with an in-house simulator having a rectangular (orthogonal) grid system showed the significant effect of faulting on reservoir performance. Unfortunately, the crossflow and communication levels between the fault blocks could not be determined accurately because the model could not represent the faults properly owing to simulator limitations. Therefore, we decided to use a simulator capable of handling distorted (nonorthogonal) grids, or corner-point geometry, to fit the faults and reservoir boundaries better. Two approaches were considered for constructing the model grid. The first approach uses a rectangular (orthogonal) gridding system that represents faults at block boundaries. As a result, reservoir geology and geometry cannot be represented properly. However, finite-difference techniques can be used in a straight forward manner with good accuracy. The second approach uses a nonrectangular (nonorthogonal) gridding system that accommodates accurate description of the fault pattern by placing block boundaries along fault traces. However, the nonrectangular gridding systems could lead to serious calculation errors.1 We conducted full-field simulation studies using both orthogonal and nonorthogonal grid models. The results of the history match of both models with the same input data were compared to determine the magnitude of the calculation error and differences in model performance.
Read full abstract