Use of PEBI Grids for Complex Advanced Process Simulators

Abstract PEBI ("perpendicular bisector") grids have been shown in the past to have the potential to reduce computational times for the simulation of relatively straight-forward processes. It is the purpose of this paper to investigate PEBI-based gridding for a much more complex thermal process in a full field setting using commercial simulation products. The goals are to study both computing efficiency and accuracy by comparing results obtained by modelling a field using the more conventional corner point-based gridding with local grid refinement to those obtained using a PEBI gridding approach. The field in question has been produced for about 20 years, with an operational history that includes cyclic steam stimulation in portions of the field. Many vertical faults have been mapped in the reservoir, and over 140 wells, many of them deviated or horizontal, have been drilled. A simulation study had been done using approximately 170 000 active corner point cells. It was found that the inclusion of foamy oil behaviour was necessary to get a good match of the production history. It appeared though that this simulation was lacking in accuracy near horizontal wells. This was probably due to the inability of the model to align its locally refined grids properly with many of the wells, and the difficulty in getting reasonable refinement levels where cyclic steam stimulation was being carried out, without making the problem considerably larger. A second simulation study was carried out using a PEBI-based grid. The more flexible aspects of this gridding system allowed construction of a better aligned grid, especially near the horizontal wells and near faults. Moreover, the characteristics of the PEBI-based grid also allowed efficient grading of grid cell sizes, so that particularly fine-scale gridding could be used near wells, while still maintaining an overall model size that was about half that of the corner point model. All processes modelled in the original simulation were replicated in the PEBI-gridded model, including the foamy oil and thermal aspects. The PEBI gridded model ran in about a third of the time of the original corner point model, and showed accuracy improvements which were attributable to the better placement of grid cells. These results showed that the simulator's ILU-based sparse matrix solver technology was very capable of computing in an unstructured grid environment, even while using cells near wells that were smaller than those used in the corner point model. Thus, PEBI-based gridding can be used to efficiently model complex processes in a full field setting. These grids can demonstrably improve accuracy and are much more adaptable for modelling near wells and in a complex geological setting. A three-fold run time improvement was noted for the field in question when comparing to a more conventional, corner point gridded model.