We previously published a two-dimensional data-driven model (INSIM-FT) for history matching waterflooding production data and to identify flow barriers and regions of high connectivity between injector-producer pairs. This two-dimensional INSIM model assumed vertical wells. The history-matched models can be used for prediction of waterflooding performance and life-cycle waterflooding optimization. The INSIM-FT-3D model presented here extends INSIM-FT to three dimensions, considers gravity and enables the use of arbitrary well trajectories. INSIM-FT-3D places nodes at each well perforation and then adds nodes throughout the reservoir. Flow occurs through "streamtubes" between each pair of connected nodes. Mitchell's best candidate algorithm is used to place nodes and a three-dimensional (3D) connection map is generated with Delaunay triangulation. Pressures and saturations at nodes, respectively, are obtained from IMPES-like pressure equations and a Riemann solver that include gravity effects. With history-matched model(s) as the forward model(s), we estimate the optimal well controls (pressure or rates at control steps) that maximize the life-cycle net-present-value (NPV) of production under waterflooding using a gradient-based method that employs a stochastic gradient. Two 3D reservoirs are considered to establish the viability of using INSIM-FT-3D history-matched models for waterflooding optimization, a channelized reservoir and the Brugge reservoir. Unlike history-matching and waterflooding optimization based on reservoir simulation models, INSIM-FT-3D is not a detailed geological model. Moreover, the time required to run INSIM-FT-3D is more than one order of magnitude less the cost of running a comparable reservoir simulation model.

You can access this article if you purchase or spend a download.