Ensemble based History Matching of Reservoir Flow Models using 4D seismic, Production Data and LWD - A Feasibility Study

Abstract

In the presented work, history matching methodology is explored with a particular focus on ensemble based methods. Real data, both production data and 4D seismic together with 3D Model Repository, has been used to update a reservoir model by utilization of the Ensemble Smoother method. The Ensemble Smoother method is well suited for inclusion of different data types and to accomplish this the Ensemble based Reservoir Tool has been applied. A rock physics model has been used to generate simulated time strain which is added to the history matching procedure where time strain obtained from 4D seismic serve as observations. The initial ensemble comprises parametrization of lateral permeability as well as the addition of a fault, which was detected by LWD measurements and confirmed by the Model Repository, providing various plausible model realizations.

Ensemble Smoother updated realizations show a trend of choosing low values for the fault transmissibility multiplier in most subdivisions of the fault indicating strong influence of a sealing fault. Where the fault transmissibility multiplier has higher values the modified permeability is consistently reduced, hence compensating for the lack of reduced transmissibility in the relevant area, limiting flow across the proposed fault. The field of modified permeability closest to one of the production wells is also consistently reduced. When the span of the initial ensemble is sufficiently wide the Ensemble Smoother update narrows this span closer to observed data. In situations where this span is not sufficiently wide the Ensemble Smoother method is still working as intended seeing how updated model realizations move in the right direction and are situated as close to the observations as possible given the initial parametrization.

The usefulness of ensemble based methods extend beyond history matching. The possibility for continuous updates of the reservoir flow model with new data acquired from multiple sources is interesting in relation to determining optimal well placement and completion with respect to production optimization in both pre-drilling- and while-drilling perspectives.