Multi-objective optimisation of Water-Alternating-Gas injection in fractured carbonate reservoirs : from theoretical considerations to application in a low permeability giant carbonate reservoir

Abstract

Water-Alternating-Gas (WAG) injection can be an efficient way to improve recovery factors in carbonate reservoirs. However, the complex geology of carbonate reservoirs and the complexity of the WAG process itself create uncertainty when predicting the efficiency of WAG. It is hence difficult to select engineering pa rameters such that recovery during WAG is optimised. In this thesis, Polynomial Chaos Expansion (PCE) based proxy modelling is coupled with the Non-Sorting Genetic Algorithm-II (NSGA-II) multi-objective optimisation workflow to identify engineering parameters, i.e. well controls, that enable us to optimise WAG injection in both, synthetic and real fractured carbonate reservoirs. The contribution from this thesis is the implementation, testing, and application of an efficient framework to optimise WAG injection. In particular, this thesis aims to understand if a workflow that combines PCE-based proxy modelling coupled with the NSGA-II algorithm can be applied effectively in a real-world scenario. The application of this workflow becomes more difficult as the physics of the recovery mechanism become more complex, the range of engineering controls increases, and the reservoir geology becomes more complicated. This thesis therefore evaluates how the order of the PCE changes, how the proxy models need to be enriched, and how the number of iterations need to be adjusted in order to obtain reliable optimal WAG injection designs in different geological reservoir scenarios. The geological scenarios range from a synthetic analogue model for a Middle-Eastern fractured carbonate reservoir to a complex, giant, and faulted onshore carbonate reservoir in the UAE that comprises a thick transition zone. Each geological scenario considers increasing levels of complexity and uncertainty with respect to reservoir characterisation. WAG designs range from immiscible to pseudo-miscible (for the synthetic reservoir cases) to fully miscible (for the real-field application). As a key result, this thesis clearly illustrates the potential of integrating proxy modelling and multi-objective optimisation with new reservoir characterisation methods to identify WAG injection designs that have the potential to increase reservoir performance in fractured carbonate reservoirs.