Impact of eliminating fracture intersection nodes in multiphase compositional flow simulation

Algebraic elimination of nodes at discrete fracture intersections via the star‐delta technique has proven to be a valuable tool for making multiphase numerical simulations more tractable and efficient. This study examines the assumptions of the star‐delta technique and exposes its effects in a 3‐D, multiphase context for advective and dispersive/diffusive fluxes. Key issues of relative permeability‐saturation‐capillary pressure (kr‐S‐Pc) and capillary barriers at fracture‐fracture intersections are discussed. This study uses a multiphase compositional, finite difference numerical model in discrete fracture network (DFN) and discrete fracture‐matrix (DFM) modes. It verifies that the numerical model replicates analytical solutions and performs adequately in convergence exercises (conservative and decaying tracer, one and two‐phase flow, DFM and DFN domains). The study culminates in simulations of a two‐phase laboratory experiment in which a fluid invades a simple fracture intersection. The experiment and simulations evoke different invading fluid flow paths by varying fracture apertures as oil invades water‐filled fractures and as water invades air‐filled fractures. Results indicate that the node elimination technique as implemented in numerical model correctly reproduces the long‐term flow path of the invading fluid, but that short‐term temporal effects of the capillary traps and barriers arising from the intersection node are lost. Key Points Star‐delta elimination technique shows first‐order convergence for diffusive/dispersive transport and multiphase advective flow Elimination at intersections precludes conduit flow and short‐term capillary traps, but redirects invading NAPLs appropriately Elimination removes short‐term capillary barriers and is not recommended for transient simulation of water invading dry fractures