Combining the Perpendicular Bisector (PEBI) grids with the Two Point Flux Approximation (TPFA) scheme demonstrates a potential to accurately model on unstructured grids, conforming to the geological and engineering features of real grids. However, with the increased complexity and resolution of the grids, the PEBI conditions will inevitably be violated in some cells and the approximation properties will be compromised. The objective is to develop accurate and practical grid quality measures that quantify such errors.
We critically evaluated the existing grid quality measures and found them lacking predictive power in several areas. The available k-orthogonality measures predict error for flow along the strata, although TPFA provides an accurate approximation. The false-positive results are not only misleading but can overwhelm further analysis. We developed the so-called "truncation error" grid measure which is probably the most accurate measure for flow through a plane face and accurately measures the error along the strata. We also quantified the error due to the face curvature. Curved faces are bound to exist in any real grid.
The impact of the quality of the 2-D Delaunay triangulation on TPFA approximation properties is usually not taken into account. We investigate the impact of the size of the smallest angles that can cause considerable increase of the condition number of the matrix and an eventual loss of accuracy, demonstrated with simple examples. Based on the analysis, we provide recommendations. We also show how the size of the largest angles impacts the approximation quality of TPFA. Furthermore, we discuss the impact of the change of the permeability on the TPFA approximation. Finally, we present simple tools that reservoir engineers can use to incorporate the above-mentioned grid quality measures into a workflow. The grid quality measures discussed up to now are static. We also sketch the further extension to dynamic measures, that is, how the static measures can be used to detect change in the flow behavior, potentially leading to increased error.
We investigate a comprehensive set of methods, several of them new, to measure the static grid quality of TPFA on PEBI grids and possible extension to dynamic measures. All measures can be easily implemented in production reservoir simulators and examined using the suggested tools in a workflow.