A multi-period injection strategy based optimisation approach using kriging meta-models for CO2 storage technologies

Abstract

This study aimed at addressing the problem of maximising the trapped amount of CO2 in a reservoir with heterogeneous and uncertain petro-physical properties. A multi-period injection strategy is adopted and an optimisation algorithm is employed to find out the injection quantities in different wells at different spans of time. A number of realisations of permeability and porosity are generated to represent the uncertainty in the model and are integrated into the optimisation framework. The optimisation algorithm employs a kriging based meta-modelling technique in each iteration to decide on the best sampling point for the successive iteration. This technique reduces the computational run time as the number of function calls to the actual flow solver is drastically reduced. As compared with Genetic Algorithm, the methodology gives better results in terms of computational time and convergence characteristics.