An ensemble-based production optimization technique is applied to a simulation model of OMERS Energy's Chigwell Viking ‘I’ Pool in order to determine optimal CO2-WAG cycle length, injection rates and production bottom hole pressures (BHPs). An ensemble-based approximate gradient calculation is used in an expected net present value (NPV) maximization. A single model was fist used to contrast the individual optimization of injection rates and the injection cycle lengths with the combined optimization of rates and WAG cycle lengths in order to determine the best parameters to consider for WAG optimization. By combining cycle length, injection rate and production BHP controls, a significant increase in the NPV is observed relative to using injection rate and production BHP control only. The model's non-uniform well placement and geological properties require full individual controllability of the wells to realize the optimal sweep. The controllability offered by combining cycle length, injection rates and production BHP as controls for individual wells is seen to lead to solutions where some wells are under gas-only injection and other wells are under water-only injection for some time. The obtained solutions in general require fewer switches between injection phases and therefore offer a reduction of the operational costs and risks.
The optimization workflow and control parameterization are also applied to an ensemble of model realizations obtained by generating samples of the uncertain model parameters. The improvements in expected NPV demonstrate the practical applicability of ensemble-based approaches for optimization under uncertainty to real field cases. If CO2 storage credits are added to the objective function, a different control strategy is found that also leads to an increase in NPV. This result highlights the potential for economic incentives to increase both CO2 storage and oil recovery. We also demonstrate that the availability of CO2 (or, similarly, its price) will influence the optimal strategy, and therefore that strategies that work in one CO2 availability/price scenario may not necessarily be optimal in another one. The techniques discussed in this paper, however, can be applied to determine the optimal strategy for each particular operational scenario.