Modeling compositionally-sensitive reservoirs
MetadataShow full item record
This research work provides guidelines for choosing the PVT model (black-oil or EOS compositional) for full-field reservoir simulation of volatile/near-critical oil and gas condensate fluid systems produced by depletion and/or gas injection. The main issues covered in this dissertation are: 1. Modeling of PVT and flow behavior. 2. Ensuring consistency between black-oil and compositional simulation, particularly for in-place surface volumes. 3. Situations when black-oil models can and should be used, and when compositional models are required. The first task in this research was to select a wide range of fluid systems varying from medium-rich gas condensate; to near critical fluid; to volatile oil; to slightly volatile oil; to low-GOR oil. In this way, it was desired to have different fluid systems one can expect in any petroleum reservoir. The one way to get different fluid systems along with corresponding equation-ofstate (EOS) models was from published literature. This will require different EOS models for different fluid systems ― i.e. a large number of fluid systems and corresponding EOS models. This approach was not considered practical. Another way was to select a complex fluid system from a complex petroleum reservoir, thereafter derive different fluid systems from the complex fluid system. In this way, only one EOS model would describe all the fluid systems required in this study. The complex fluid system selected was from an actual North Sea reservoir. This fluid system varied from medium-rich gas condensate to slightly volatile oil, with a critical-fluid gas/oil transition. The EOS model was obtained by the Pedersen et al. characterization procedure with the SRK EOS. The decanes-plus fraction was split into 9 fractions using a commercial PVT simulation program. The final detailed-EOS model contained 22 components with 12 heptanes-plus fractions. Since it is impractical to use such a detailed 22-component EOS model in fullfield simulation, the detailed-EOS was lumped to fewer components in a stepwise procedure. It was possible to lump the detailed-EOS from 22 components to as few as 6 components. The lumped-EOS model predicted all PVT properties with reasonable accuracy ― even including depletion and gas injection performance for the near critical compositionally grading fluid system. The minimum miscibility pressures predicted with the lumped-EOS were also quite accurate, when compared with the detailed-EOS calculations. For black-oil simulation models, the black-oil PVT tables can be generated using different PVT experiments. It is recommended to use a depletion-type experiment (CCE, CVD, or DLE) for oil and gas condensate reservoirs. The black-oil PVT tables for compositionally grading reservoirs with undersaturated GOC should be generated from CCE experiment using the GOC fluid. For saturated GOC reservoirs, the oil PVT tables should be generated from the GOC oil and the gas PVT tables should be generated from the GOC equilibrium gas. The surface oil and gas densities should be modified to obtain correct reservoir oil and gas densities at the GOC conditions. Different methods are also described for generating black-oil PVT tables for gas injection, where the black-oil PVT tables are required to be extrapolated to higher pressures. The aim of the research was also to get consistent in-place surface volumes. The compositionally grading reservoir was initialized using the lumped-EOS model for obtaining initial in-place volumes. Different methods were used for getting compositional gradient for the lumped-EOS model. The lumped-EOS model inplace volumes were compared with the detailed-EOS model in-place volumes. It was possible to get quite accurate in-place volume in the lumped-EOS model with proper selection of the compositional gradient. The lumped-EOS compositional model in-place volumes were also compared with that of the black-oil model. For the black-oil model initialization, the blackoil PVT tables were generated from the lumped-EOS using the GOC fluid. The compositional gradient in the black-oil model i.e. solution gas-oil ratio and oilgas ratio versus depth were obtained from the compositional model. The black-oil model in-place volumes were quite accurate with proper selection of the black-oil PVT table and compositional gradient. For reservoir simulation studies, a 3D dipping reservoir with 99 layers was used. The reservoir had dip angle of 3.8 degree. The permeability from the top layer to the bottom layer were either monotonically increasing or decreasing. The reservoir layer permeabilities were varied based on Dykstra-Parsons model. Different average reservoir permeability was used to quantify the effect of gravity. The simulated reservoir performance was analyzed for different fluid systems for both depletion and injection cases. Furthermore, the possibility for reducing the number of numerical layers without loosing the “accuracy” was examined since it is not practical (due to excessive CPU time) to use 99-layer simulation model for comparison of the production performance from the compositional and the black-oil models. Different fluid systems and permeability distributions were used for comparing the simulated performance from a model with a reduced number of layers with the results from the model with 99-layer. Based on the analysis, it was found that 10 numerical layers with equal flow capacity were sufficient to reproduce the production performance from the 99- layer. For comparing black-oil and compositional simulation performance results, the 3D model was used to obtain areal and vertical sweep efficiency correctly. Three different non-communicating geologic units were used, each geologic unit with ten numerical layers and different horizontal permeabilities. The layer permeabilities were distributed in different ways ― e.g. highest k at the top, highest k at the bottom or highest k in the middle. Reservoir performance was analyzed for different simulation cases. Each “reservoir” was simulated using black-oil and compositional models for various depletion and gas injection cases. The simulated performance for the two PVT models was compared for fluid systems ranging from a medium-rich gas condensate to a critical fluid, to slightly volatile oils. The initial reservoir fluid composition was either constant with depth, or a vertical compositional gradient. Both saturated and undersaturated GOC’s were considered. The reservoir performance for the two PVT models was also compared for different permeability distributions. Reservoir simulation results show that the black-oil model can be used for all depletion cases if the black-oil PVT data are generated properly. In most gas injection cases, the black-oil model is not adequate ― with only a few exceptions. Most of the results presented in this dissertation have been presented in the following paper, included as Appendix A: Fevang, Ø., Singh, K., and Whitson, C.H. : “Guidelines for Choosing Compositional and Black-oil Models for Volatile Oil and Gas Condensate Reservoirs,” paper SPE 63087 presented at the 2000 Annual Technical Conference and Exhibition, Dallas, Texas, 1-4 October 2000.