Reservoir simulation has entered a new era with the advent of massively parallel processing (MPP) and its ability to handle millions of cells. In this study, a simulation model was constructed from a geologic model without scaleup, with the resulting model having 1.38 million cells. The majority of the effort in reservoir simulation was spent on reviewing geologic and engineering data. History-match changes were made at the well level and interpolated in k H and k V arrays with a deterministic mapping routine. Despite the complexity, the model size, and 30 years of history, a successful history match was achieved after 40 runs. The model has been used to identify well locations for bypassed oil recovery. To date, these locations have not been drilled. IntroductionReservoir simulation projects are often a balance between the complexity of (presumably) producing more accurate results and reasonable computer turnaround time. Because of computer resource constraints, conventional simulation technology usually requires upscaling of high-resolution geology. Recent developments in MPP computer technology have dramatically increased computer horsepower, giving us the choice of either enhancing geologic and/or engineering complexity or enjoying faster turnaround. [1][2][3] Is more model detail worth it? Will it better replicate the physics in the reservoir? Will we be able to make better decisions with a high-resolution model? In this case study, the Hadriya reservoir in Berri field was chosen to test this concept and to act as a test case for Saudi Aramco's new POWERS MPP simulator. 4 Background Berri Hadriya has been on production since 1970 and on peripheral waterflood since 1975. Water encroachment into the reservoir is very complex, with extensive water-over-oil bypassed areas. Over the years, a comprehensive well-logging program has monitored this water encroachment. Remaining development options are mainly in bypassed areas behind the flood front or in dry areas of lower rock quality. The overall objective of this project was to guide development drilling, especially in the bypassed oil areas.The Berri Hadriya reservoir is characterized by highly permeable intervals separated by thin, tight streaks that influence water encroachment. To maintain reservoir character and capture the complex water encroachment, the decision was made to simulate the reservoir at the scale of the geologic model. The resulting simulation model has 1.38 million cells.During the history match, only the permeability was changed; structure, cell thickness, porosity, and facies from the geologic model remained the same. The overall strategy was to make changes only at well locations and to use a mapping program to rebuild the model for each run. Where available, core data, pressure buildups, and flowmeters would be honored. GeologyThe Hadriya reservoir in Berri field is a north/south-oriented wedge-shaped accumulation of grainy carbonates deposited in a distally steepened outer-ramp environment. 5 Facies patterns devel-

This paper describes the technical considerations that set the major design parameters for the Prudhoe Bay Miscible Gas Project (PBMGP). This project was planned as a large hydrocarbon miscible flood. The basic project concept is to manufacture a miscible injectant from the field separator off-gas. This injectant is compressed and delivered to the EOR project area for injection in a water-alternating-gas (WAG) mode. The miscible gas supply will vary, generally increasing with time. Over the first 10 years of the project, an average supply of 200 MMscflD [5.66x 10 6 m 3 /d] is anticipated. The gas processing plant and the factors affecting miscible gas supply are described. The EOR froject area was selected by a screening process. This led to a project area definition encompassing 4.9xlO RB [779 X 10 6 m 3 ] PV tontaining 2.2xI09 STB [350 X 10 6 m 3 ] original oil in place (OOIP). The planned cumulative volume of miscible gas injected will be 10% PV.Reservoir studies indicate an incremental oil recovery by miscible flooding of some 5.2% OOIP or 115x 10 6 STB [18.3X10 6 m 3 ]. Aspects of these reservoir studies are described.The Sadlerochit reservoir is both the ultimate source of the miscible solvent and the target reservoir. This introduces several reservoir/facility interaction effects. The planning of a major EOR project in the Arctic has also involved technical considerations not routinely encountered in conventional oilfield projects. Both aspects are discussed in the paper.

The Prudhoe Bay Field-Sadlerochit Reservoir presents many complexities which make reservoir simulation of this large oil accumulation difficult. The presence of major faults, the existence of large shale complexes, the enormous gas cap and associated coning of wells, and the heavy oil/tar mat could cause, even if each existed separately, major problems for a standard reservoir simulation model. This paper describes the techniques which were paper describes the techniques which were used by ARCO Oil and Gas to overcome these difficulties in modeling the Prudhoe Bay Field. Because of the large areal extent o the field, a unique grid in radial coordinates was used to reduce the number of grid cells to a manageable number. A new form of velocity-dependence was introduced into the gas/oil relative permeability and capillary pressure pseudo function relationships in order to more accurately model the gravity drainage and coning behavior in the field. A special pressure solution technique was developed to handle the extensive faults throughout the reservoir. Finally, an overall well and facility management routine was added to control the production/injection of the field and to give realism to the complex boundary conditions involved in these controls. Validation of the modeling techniques was carried out in small scale models, larger cross-sectional models, and finally through history matching of the field. These validations gave excellent results. Introduction The Prudhoe Bay Sadlerochit reservoir (Figure 1) discovered in 1968, represents the largest accumulation of oil found to date in the North American Continent with a proven volume in excess of 20 billion proven volume in excess of 20 billion barrels (3.2 × 109 m3) of oil and 23 trillion cubic feet (6.5 × 1011 m3) of free gas originally in place. Reservoir simulation studies were begun by ARCO Oil and Gas soon after discovery, however, the original cross-sectional models which were developed, although adequate for recovery calculations, were inadequate to completely define a real variations in field performance. Prediction of those areal variations would be necessary for planning of production facilities. In addition, the complexity of the reservoir indicated that several different but simultaneous mechanisms might be optimal for maximum recovery. Only a model with areal variations could adequately handle this. It was, therefore, decided in January, 1978, to undertake the development of a full field three dimensional reservoir simulation model for the Prudhoe Bay Sadlerochit reservoir. The criteria for the development of the model were that it should be relatively fast computationally yet be accurate enough to properly predict recovery mechanisms in the properly predict recovery mechanisms in the reservoir. The required computational speed of the model (results within one-two days) necessitated that there be fewer than 10,000 cells in the model with the computational facilities available at that time at ARCO Oil and Gas Co. A minimum of five model layers were required to properly describe the mechanisms of fluid movement because four major shale complexes existed within the main reservoir. Since field development was initially planned on a 160 acre spacing, the maximum areal cell size for the model was limited to this value.

Reservoir simulation has entered a new era with the advent of massively parallel processing and its ability to handle millions of cells. In this study a simulation model was constructed from the geologic model without scale up, the resulting model having 1.38 million cells. The majority of effort was spent on reviewing geologic and engineering data. History match changes were made at the well level and interpolated in Kh and Kv arrays using a deterministic mapping routine. Despite the complexity and model size, a successful history match was achieved in 40 runs. The model has been very successful in identifying well locations for bypassed oil recovery. Introduction Reservoir simulation projects are often a balance between complexity to (presumably) produce more accurate results and reasonable computer turn-around time. Because of constraints in computer resources conventional simulation technology usually requires upscaling of high-resolution geology. Recent developments in massively parallel processing (MPP) computer technology have dramatically increased computer horsepower giving us the choice of either enhancing geologic and/or engineering complexity or enjoying faster turn-around.1,2,3 Is more complexity worth it? Will a high-resolution model better replicate the physics in the reservoir? Will we be able to make better decisions with a high-resolution model? In this case study, the Hadriya reservoir in Berri field was chosen as a test case for Saudi Aramco's new POWERS MPP simulator.4 This large carbonate reservoir has been on production since 1970 and peripheral waterflood since 1975. Water encroachment into the reservoir is complex with extensive water-over-oil bypassed areas. The decision was made to simulate this reservoir at the scale of the geologic model. Background Berri Hadriya has been on production since 1970 and peripheral waterflood since 1975. Remaining development options are mainly in bypassed areas behind the floodfront or in dry areas of lower rock quality. The overall objective of this project was to guide development drilling especially in the bypassed oil areas. The Berri Hadriya reservoir is characterized by highly permeable intervals separated by thin tight streaks that influence water encroachment. To maintain reservoir character and capture the complex water encroachment, the decision was made to simulate the reservoir at the scale of the geologic model. The resulting simulation model has 1.38 million cells. During the history match only permeability was changed. Structure, cell thickness, porosity and facies from the geologic model were unchanged. The overall strategy was to make changes only at well locations and use a mapping program to rebuild the model for each run. Where available, core data, pressure buildups and flowmeters would be honored.