Corner-point geometry has run reservoir simulation for decades, and on a gently folded, lightly faulted field it earns its place: a simple i, j, k index, fast to build, supported by every major simulator and geomodeling package. The trouble starts when the geology stops being gentle.
What a corner-point grid does at a fault
A corner-point grid describes the reservoir with near-vertical pillars on a logically Cartesian map, each cell pinned by eight corner points. To model a fault, corner points slide along the pillars until layers are offset, and where the offset juxtaposes non-adjacent layers the simulator stitches them with non-neighbor connections. It works — up to a point. Because the grid can never abandon its Cartesian index, a fault that cuts diagonally across the map is represented as a stair-step: a zig-zag of cell faces standing in for a plane that is neither horizontal nor vertical.
Around that stair-step, cells distort. They turn non-orthogonal, stretch to high aspect ratios, twist, and — at pinch-outs — collapse to zero thickness.
Why the distortion costs you accuracy
Two things break at once. First, most simulators compute flow with a two-point flux approximation that is only accurate on K-orthogonal cells; the twisted cells around a fault are exactly where that assumption fails, so grid-orientation error concentrates where the geology matters most. Second, the fault plane you interpreted so carefully is not actually a face in the grid, so the fault transmissibility that decides whether fluids cross it is applied to a stair-stepped approximation rather than the real surface.
For CO₂ storage the stakes are sharper. A fault can seal a plume or channel it toward the seal edge; a caprock either contains buoyant CO₂ or it does not. If the grid does not sit on those surfaces, the migration and trapping you simulate are not quite the ones in the ground.
What PEBI grids do differently
An unstructured PEBI grid — PEBI for perpendicular bisector, the reservoir name for a Voronoi grid — drops the Cartesian index entirely. You place node points wherever the geology demands, and each cell face lands on the perpendicular bisector between neighbors. Two consequences follow directly:
- The grid is orthogonal by construction for isotropic permeability, which is exactly what the two-point flux scheme wants.
- You can place nodes so faces fall on the fault and horizon surfaces, making the grid fault-conforming by design. The fault becomes a real internal face with its own transmissibility, and the cells around it stay well-shaped.
Conform to the geology and honor the solver’s math — a PEBI grid does both, where a corner-point grid is forced to trade one against the other.
The catch, and what changed
If PEBI grids are so well suited, why has corner-point stayed the default? Because building a conforming Voronoi grid was the hard part. Traditional methods clipped cells against fault and boundary surfaces, and clipping breeds slivers and degenerate cells — right where quality matters. That is the barrier VoroCrust (Dr. Mohamed Ebeida et al., ACM Transactions on Graphics, 2020) removed: a provably conforming polyhedral Voronoi mesh with no clipping, so faces land on the surfaces by construction.
That advance turns “PEBI is what the simulator wants, but nobody can build it fast” into an automatic step.
For oil & gas and subsurface teams whose fields keep getting more faulted, that is the difference between fighting the grid and trusting it. If corner-point grids keep stair-stepping through the structure that drives your flow and trapping, we would like to show you a grid that conforms to it instead. Book a technical walkthrough.