Reducing computational space by an adaptive moving boundary for convective dominated flow

This work presents a computationally efficient method utilizing a flood front tracking technique “moving boundary” for simulating convective dominated flow problems. The method segregates the reservoir into two regions, active and inactive; the active region forms the curtailed system whose boundary conditions represent the inactive region. These regions adaptively change with the advancement of the floodfront until they eventually merge to cover the entire reservoir model. In essence and as the name hints, the method reduces the computational size of the reservoir by virtue of artificial, or virtual, boundaries. The technique correctly identifies regions of the reservoir undergoing rapid saturation changes and introduces either Neumann or Dirichlet boundary conditions at some downstream distance from the flood front. The collocations of points where the boundary conditions are applied constitute the moving virtual boundary. Once the flood front comes in the vicinity of a virtual boundary, the boundary is shifted to the next appropriate position, hence giving the method its name Moving Boundary. The method is akin to streamline simulation whereby streamlines, traced from a pressure gradient surface, are frozen for extended temporal lengths; whereas here we freeze some boundary conditions enabling us to effectively reduce the computational requirements. As in streamline simulation, our method is best suited for convective dominated flow problems. Two dimensional, highly heterogeneous, reservoir simulation runs utilizing the moving boundary method are presented and compared to those results obtained from conventional full model runs. The results indicate an excellent match along with the added computational efficiency.