Reservoir Characterization and Modelling of Stacked Fluvial/Shallow Marine Reservoirs: What is Important for Fluid-Flow Performance and Effective Reservoir Prediction?

Abstract

Earth modelling and field implementation experience in fluvial/shallow marine clastic reservoirs suggests that relatively large, km to m scale geologic features such as stratal framework and facies architecture dominate the model behaviour and influence the accuracy of performance forecast more than numerically driven pore scale heterogeneity. Twenty seven reservoir models were built by using different modelling methods and varying a combination of more uncertain reservoir characterization parameters ranging in scale from field to voxel to pore level. The models were largely influenced by deterministic geological controls rather than numerically driven stochastic variations. Guided by regional work, core and image log integration, the large scale stratal framework, the facies scheme and the zonal variograms were kept the same in all the models while the rock fabric and pore scale heterogeneity were varied using facies models, reservoir trend maps, and permeability contrast scenarios. These 27 models were all subjected to history match constrained fluid-flow simulation with a partially active aquifer, and peripheral and pattern waterflooding. Static and dynamic parameter uncertainties were handled by using experimental design. The results show that oil recovery from existing wells at 98% water cut ranges from 25 to 40% in these 27 models. Also, there are differences (greater than 20%) in water-breakthrough time, water cut, and cumulative water production. While the large scale stratal framework defined the flow units and provided fundamental control on zonal production behaviour (e.g. differential depletion, early water breakthrough, and conformance control), the differences in oil recoveries and water cut among all models were found to be related to variation in m-scale rock heterogeneity rather than pore scale heterogeneity and permeability contrast. Do we need more complex, numerically exhaustive models to predict flow performance in channelised reservoirs with reasonable well control? Some of the simple, but geologically constrained models we constructed provided superior flow-simulation results, impacted development concepts and led to accurate prognosis from new drills.