B018 MASS CONSERVING FORWARD & ADJOINT BLACK-OIL RESERVOIR SIMULATION FOR (UNDER)SATURATED OIL

Reservoir simulators that efficiently calculate gradients are not yet widely used, because for few control parameters (or little data available), it is feasible to approximate gradients by numerical perturbations. As more parameters need to be handled, the need for efficiently calculating gradients increases. In this paper it is shown how to implement a Black-Oil reservoir simulator for (under)saturated oil with gradient capabilities. The usefulness of implementing reservoir simulators with Jacobian capabilities is shown by embedding reservoir simulation in a closed loop setting. A gradient-based optimization routine can determine how inflow control valves in smart wells should be handled in order to optimize oil recovery or net present value. Moreover a gradient-based data assimilation routine improves the reservoir simulator's predictive capabilities by updating parameters every time new measurements become available.