| Popis: |
Abstract Many fields show strong compositional variations of the reservoir fluid both with depth and with time during production. With such fluids, which are volatile oils or gas condensates, compositional models are needed instead of black-oil models. When representing these fluids in a compositional reservoir simulation, a limited number of pseudocomponents are required to avoid a prohibitive computation time. To represent the same fluid in a process simulation, more components are required to be able to design correctly the surface facilities. The objective of this paper is to show, through several field case examples, how a consistent fluid representation within the reservoir and through the surface facilities can be integrated into the compositional reservoir simulation. First of all, this paper shows through a complex field case example that a seven pseudocomponent fluid representation is sufficient to model accurately a 150 meter thick reservoir fluid column. Subsequently, this study tackles the problem of delumping the reservoir compositional profiles to feed process simulators: tests have been performed to assess the influence of component lumping in reservoir simulations and a simple delumping method is proposed. Finally, an evaluation of the different possibilities of representing a process in a compositional reservoir simulation is performed through several tests on field case examples, and a method is proposed to minimise the deviations from the actual process. Introduction Petroleum fluids are composed of a huge number of hydrocarbons, which are partitioned into several phases (gas, liquids, solids) according to the laws of thermodynamic equilibrium. To model correctly the behaviour of these fluids, it is not necessary to consider individually each compound: most of the pure compounds are grouped together into so-called pseudo-components, mainly to save computation time. For the sake of simplicity, both pure compounds and pseudo- components will be called components. In black oil reservoir simulations, only two components are modelled: gas and oil. In compositional reservoir simulations, applied to volatile oil and gas condensate reservoirs, several components are needed to model the fluid behaviour correctly; indeed, with these kinds of fluids, strong compositional variations are observed with depth and with pressure, inducing strong property changes. In process simulations, an even more detailed description of the fluid is often required, especially for the light components, since the outlet specifications (sales gas composition, LPG recovery, export liquid RVP…) concern individual pure compounds and are very sensitive to phase equilibrium and enthalpic calculations over a wide range of temperature and pressure. Moreover, the representation of the surface process in reservoir simulators is extremely simplified in comparison with the actual process. The objective of this paper is to investigate the deviations due to different fluid representations (lumped, delumped or detailed), and to different process representations (actual or simplified) in compositional reservoir simulations. For this purpose, many tests have been carried out on several field case examples. Moreover, methods are proposed to reduce and minimise the observed deviations. The first part deals with fluid modelling for reservoir and process simulations: a detailed and a lumped fluid representations are used to model a whole reservoir fluid column. The second part addresses the problem of «delumping» the reservoir compositional profiles to feed process simulators. The third part evaluates the different possibilities of representing a process in a compositional reservoir simulation. |
