Optimization of Well Placement in Carbon Capture and Storage (CCS): Bayesian Optimization Framework under Permutation Invariance

Abstract

Carbon Capture and Storage (CCS) stands as a pivotal technological stride toward a sustainable future, with the practice of injecting supercritical CO2 into subsurface formations being already an established practice for enhanced oil recovery operations. The overarching objective of CCS is to protract the operational viability and sustainability of platforms and oilfields, thereby facilitating a seamless transition towards sustainable practices. This study introduces a comprehensive framework for optimizing well placement in CCS operations, employing a derivative-free method known as Bayesian Optimization. The development plan is tailored for scenarios featuring aquifers devoid of flow boundaries, incorporating production wells tasked with controlling pressure buildup and injection wells dedicated to CO2 sequestration. Notably, the wells operate under group control, signifying predefined injection and production targets and constraints that must be adhered to throughout the project’s lifespan. As a result, the objective function remains invariant under specific permutations of the well locations. Our investigation delves into the efficacy of Bayesian Optimization under the introduced permutation invariance. The results reveal that it demonstrates critical efficiency in handling the optimization task extremely fast. In essence, this study advocates for the efficacy of Bayesian Optimization in the context of optimizing well placement for CCS operations, emphasizing its potential as a preferred methodology for enhancing sustainability in the energy sector.

Keywords

• CCS
• machine learning
• Bayesian optimization
• Gaussian process
• regression
• predictive modeling