Abstract
The efficiency of streamline simulation, and in particular compositional streamline simulation, can be improved greatly if the streamline density is allowed to be locally adapted.A dense coverage is generally needed near composition fronts while in other areas of the flow domain a coarse streamline coverage may suffice to give accurate saturation and compositions. In the traditional streamline method,streamlines are viewed as fluid carriers. The associated low order mapping of compositions from the streamlines to the background grid used to solve for flow requires each background grid cell to be crossed by at least one streamline. This constraint prohibits local streamline adaptation and often leads to an unnecessarily high density of streamlines in large parts of the reservoir. For compositional problems, for which most of the computation time is spent on flash calculations along the streamlines, this approach leads to high and unnecessary costs. In such applications, it is crucial to minimize the number of streamlines, while ensuring a minimum streamline density to achieve the desired solution accuracy.
We view streamlines as forming a flow-based unstructured grid for transport. This viewpoint, together with an improved accurate mapping from streamline grid to background grid proposed by Mallison, Gerritsen and Matringe (2004)1 allows us to locally increase or decrease the streamline coverage. The new framework also allows partial streamlines (that do not start and end at wells) to be used to dynamically increase the local streamline density where necessary. To determine what streamline density is needed in different regions of the flow domain we designed an indicator that a priori detects areas of potential high mapping errors. The indicator takes into account the local density of streamlines as well as the proximity of solution fronts. This new streamline coverage control algorithm, combined with Adaptive Mesh Refinement (AMR) along the streamlines, can be viewed as an AMR strategy for the streamline grid.
