No abstract

No abstract

A reliable prediction of two-phase flow through porous media requires the development and validation of models for flow across multiple length scales. The generalized network model is a step towards efficient and accurate upscaling of flow from the pore to the core scale. This paper presents a validation of the generalized network model using micro-CT images of two-phase flow experiments on a pore-by-pore basis. Three experimental secondary imbibition datasets are studied for both sandstone and carbonate rock samples. We first present a quantification of uncertainties in the experimental measurements. Then, we show that the model can reproduce the experimental fluid occupancies and saturations with a good accuracy, which in some cases is comparable with the similarity between repeat experiments. However, high-resolution images need to be acquired to characterize the pore geometry for modelling, while the results are sensitive to the initial condition at the end of primary drainage. The results provide a methodology for improving our physical models using large experimental datasets which, at the pore scale, can be generated using micro-CT imaging of multiphase flow.

Summary The saturation distribution after unstable waterflooding into highly viscous oil may have a decisive effect on the efficiency of tertiary polymer flooding, in particular because of hysteresis effects associated with oil banking. In this work, we model waterflood and tertiary polymer-flood experiments performed on Bentheimer sandstone slabs with heavy oils of approximately 2,000 and 7,000 cp, and compare the numerical results with experimental production, pressure, and X-ray data. The unstable waterfloods are initially simulated in two dimensions with our parallel in-house research reservoir simulator (IHRRS) using a high-resolution discretization. In agreement with existing literature, we find that Darcy-type simulations dependent on steady-state relative permeabilities—inferred here from a 3D quasistatic pore-network model (PNM)—cannot predict the measured waterflood data. Even qualitatively, the viscous-fingering patterns are not reproduced. An adaptive dynamic PNM is then applied on a 2D pore network constructed from the statistics of the 3D network. If the fingering patterns simulated with this 2D PNM are qualitatively in good agreement with the experimental data, a quantitative match still cannot be obtained because of the limitations of 2D modeling. Although 3D dynamic PNMs at the slab scale would currently lead to prohibitively high computational cost, they have the potential to address the deficiencies of continuum models at highly unfavorable viscosity ratio. For the tertiary polymer floods characterized by a much more favorable mobility ratio, Darcy-type modeling is applied, and history matching is conducted from the end of the waterfloods. We find that unless hysteresis caused by oil banking is accounted for in the relative permeability model, it is not possible to reconcile the experimental data sets. This hysteresis phenomenon, associated with oil invasion into previously established water channels, explains the rapid propagation of the oil bank. For the considered experiments, a simultaneous history match of good quality is obtained with the production and pressure data, and the simulated 2D saturation maps are in reasonable agreement with X-ray data. This paper addresses the challenges in modeling highly unstable waterflooding, using both a conventional Darcy-type simulator and adaptive dynamic PNM, by comparing the simulated results with experimental data including saturation maps. It also highlights the important role of relative permeability hysteresis in the tertiary recovery of viscous oils by polymer injection.