Central upwind finite volume method in unstructured meshes: Application for Polymer injection in porous media
Central-Upwind; Unstructured mesh. Finite Volume. Adsorption. Retention. Least-Squares
Polymer flooding in oil reservoirs has been one of the enhanced oil recovery (EOR) methods most applied in the last decades. Its implementation aims mainly to adjust the unfavorable mobility ratio, increasing sweep efficiency. However, the polymer used to viscosify the injected solution is commonly subject to adsorption and retention effects, which may result in damage to the formation, and as a consequence the injectivity loss. In this context, due to the complexity and high cost of the EOR method, engineering has relied on numerical simulations to assess technical feasibility. Yet, available computational tools are generally limited to work with pre-defined mathematical models, which may not account realisticly the main physical phenomena. Therefore, deriving mathematical and computational models to describe optimal scenarios for the recovery of oil by polymer flooding is imperative. Hence, in this work, a new mathematical and computational model is proposed for the polymer flooding in oil reservoirs. The model includes the Darcy law for water and oil phase, and the water viscosity is determined by an experimental correlation that establishes its dependence with polymer concentration. Furthermore, mass conservation governs the two-phase flow for oil and water, expressed by their saturation. Finally, a convection-diffusion-reaction differential equation is used to model polymer transport in porous media. For numerical aspects, aiming to obtain a model that is able to compute with accuracy the effect of formation damage in the well vicinity, a triangular unstructured mesh is considered in the discretization. The discreet solution to the water saturation and to the polymer transport is obtained using the high-order finite volume Central-Upwind method along with Runge-Kutta method. In order to analyse the accuracy and stability, different hyperbolic problems are simulated and their results compared to analytical solutions. Therefore, firstly simulations to canonical hyperbolic problems in a rectangular domain discretized with a structured triangular mesh. In this doing this, one is able to analyze the accuracy of different methods to compute the gradient. In all scenarios, the method based on least-squares was superior to the other one studied. Similarly, solutions to the same problems are obtained using least-squares in a radial domain with a local refinement in the vicinity of the well. For the polymer system, the proposed adsorption and retention effect are considered. Although the mesh is coarse away from the well, numerical solutions are accurate and stable, with small numerical diffusion near the discontinuities.