RESERVOIR PRODUCTION SETTINGS OPTIMIZATION UNDER MODEL UNCERTAINTIES

Secondary recovery by engineered drives is a popular production process in the recovery of oil from subsurface reservoirs. It is mainly achieved by waterflooding technology - which involves the injection of water into the hydrocarbon reservoir. Studies on model-based optimization of waterflooding strategies have shown that there exists a significant potential to increase the recovery factor, measured in terms of the Net Present Value (NPV) of the reservoir. However, there are inherent draw backs to achieving optimal production procedures. These include the high dimension of the search space, the computational requirements associated with the objective function evaluations (which basically entail full reservoir simulations); and most importantly, the effects of uncertainties which arise from different sources within the reservoir model. The presence of these uncertainties inevitably translates to the fact that reservoir models are only a (very) crude approximation of reality. Their predictive capability in the context of production optimization is therefore very limited and moreover, tends to deteriorate over time. In this paper, we investigate the effects of uncertainties in reservoir models. We show that effective and robust optimization of reservoir-model production settings can reduce the negative effect of model uncertainties; we also show that they can be used to find production settings that are robust against a continuous range of uncertainties.