In this paper, we present a new and automatic lumping methodology to improve the accuracy and CPU efficiency of compositional reservoir simulation. Typical equation of state (EOS) based models for prediction of PVT experiments employ a large number of components. Such detailed EOS models are subsequently lumped into a reduced number of pseudocomponents for use in compositional reservoir simulation. Traditional lumping approaches do not include any information regarding the mixture compositions will form during a gas injection process. These mixture compositions can differ significantly from those investigated by standard PVT and swelling test experiments. Here we demonstrate the benefit of including displacement characteristics obtained from semi-analytical displacement calculations in the selection of pseudo-components for compositional reservoir simulation and propose an automated scheme for grouping/lumping.
Components with close proximity in terms of specific velocities and composition space are lumped subject to the observed variation in equilibrium K values as defined by displacement dynamics. Lumping is performed starting from the most detailed fluid description in a sequential manner to gauge the quality of the fluid description as we reduce the number of components in the fluid description. Examples are provided for multicomponent reservoir fluids and CO2 injection where a substantial amount of data is available (PVT, swelling test and slimtube MMP). We demonstrate that the proposed lumping strategy provides for additional control of the predictive quality of the reduced EOS model that may not be possible with currently available lumping approaches.
A majority of oil fields currently under production are candidates for EOR processes such as tertiary gas floods. The work presented in this paper is directly applicable to the study and design of such EOR processes. In addition, this is to our best knowledge the first paper that combines displacement dynamics from the analytical theory of gas injection processes with reservoir fluid characterization towards improved accuracy/efficiency of compositional reservoir simulation.