Robust optimisation of water flooding using an experimental design-based surrogate model: A case study of a Niger-Delta oil reservoir

Reservoir simulations are often applied in oil field development to predict and optimise the performance of the production process and improve the decision-making process. This process is made complex by the need to account for uncertainties in the simulation model input parameters. The classical Markowitz optimisation model is an approach that can be applied for reservoir performance optimisation considering uncertainty. However, numerous full-physics simulations are required for the optimisation procedure to work. To overcome this challenge, we deploy a surrogate model of total oil production for the robust optimisation of waterflooding in a Niger-Delta reservoir. Using this approach, we reduce computational costs and consider four uncertain geological variables in a robust application of the optimisation routine. To adequately capture the uncertainties in the geological model, we apply 100 realisations of the reservoir model. In our presented case, we show that by applying a proxy model of the cumulative oil production in the study, the computational costs of the optimisation routines are significantly reduced. Overall, we perform more than 10000 proxy model evaluations requiring 4 days and saving 417 days of computation time which is required if full-physics simulations were applied in the optimisation routine. As a result, we can make decisions accounting for uncertainty and the potential implications for variation in the development and operation of the field.