Efficient adjoint-based well-placement optimization using flow diagnostics proxies

Abstract Model-based optimization of placement and trajectories wells in petroleum reservoirs by the means reservoir simulation forecasts is computationally demanding due to high number simulations typically required achieve a local optimum. In this work, we develop an efficient flow-diagnostics proxy for net-present-value (NPV) with adjoint capabilities computation well control gradients approximate sensitivities respect placement/trajectory parameters. The suggested flow-diagnostic consists numerically solving single pressure equation given scenario solution few inter-well time-of-flight steady-state tracer equations, achieved seconds model medium size. Although may not be particularly good approximation full response, find that cases considered, correlation very hence suitable use loop. which provides similar computational complexity as forward (a seconds). We employ version generalized reduced gradient handling individual constraints (e.g., bottom-hole-pressures rates). As result, are enforced within computations, every parameter update becomes feasible without sacrificing information. present two numerical experiments illustrating efficiency performance approach problems involving models realistic complexity. placements evaluated using simulations. conclude discussing limitations possible enhancements methodology.