Dynamic water flood optimization with smart wells using optimal control theory

摘要: During the oil production process water is generally injected into reservoir to maintain pressure and displace towards wells. Ideally, as continues will slowly move through in direction of producers, meantime sweeping between. However, because rock properties vary spatially, displacement does not occur uniformly. There may be preferential, high permeability flow paths which channels producer. Oil outside these a result displaced by water. Because this non-uniform often starts at an early stage. With conventional wells there little that can done remedy without significant costs. This also hardly prevented, identification possible preferential difficult due reservoir's limited accessibility. As uncertainty lack control on typically only relatively small percentage (30-40 percent) present recovered economically. Hence, world's recoverable reserves increased if larger from reservoir. In de last few years variety technologies better measure have been developed. These are installed within well they operated remotely. A equipped with type measurement technology referred smart, intelligent or instrumented well. down-hole valves isolating packers split up segments controlled separately. enables fluid out By manipulating valve-settings it some degree change distribution thereby The objective thesis work examine doing increase An important part study comprises calculation optimize net value process. To able do for various reservoirs numerical model was used, instead real optimal gradient-based optimization routine, derivative information calculated theory. results show improvement flooding achieved dynamically controlling injection fields smart realized valves, vertical surface. depends spatial variation properties. scope constraints operating conditions, increases increasing available inject produce fluids. addition relative locations, partly determine what extent affected. fact dynamic negative impact geological features mitigated. poorly known reality, must based estimated We investigated data. end routine combined ensemble Kalman filter data-assimilation method, developed RFRogaland Research. used frequently update pressure-, saturation, reservoir, After each model, optimum were recalculated remaining producing period. First indicate improvements closed-loop approach.