Accurate characterization of subsurface heterogeneity is important for the safe and effective implementation of geological carbon storage (GCS) projects. This paper explores how machine learning methods can enhance data assimilation for GCS with a framework that integrates score-based diffusion models with machine learning-enhanced localization in channelized reservoirs during CO2 injection. We employ a machine learning-enhanced localization framework that uses large ensembles (Ns=5000) with permeabilities generated by the diffusion model and states computed by simple ML algorithms to improve covariance estimation for the Ensemble Smoother with Multiple Data Assimilation (ESMDA). We apply ML algorithms to a prior ensemble of channelized permeability fields, generated with the geostatistical model FLUVSIM. Our approach is applied on a CO2 injection scenario simulated using the Delft Advanced Research Terra Simulator (DARTS). Our ML-based localization maintains significantly more ensemble variance than when localization is not applied, while achieving comparable data-matching quality. This framework has practical implications for GCS projects, helping improve the reliability of uncertainty quantification for risk assessment.

Since the early 2000's there has been a significant focus from many groups around the world towards the development and application of innovative technologies in order to improve reservoir management strategies and optimize field development plans. Benchmark studies are a very valuable way of evaluating and demonstrating the status and potential of developing technology. Numerical optimization is seen as a valuable technology for decision support in various stages of the life cycle of hydrocarbon fields. Its potential has been demonstrated in previous benchmark studies such as the 2008 Brugge study on Closed-Loop Reservoir Management albeit for primarily well control problems. Additionally since the Brugge benchmark exercise also involved history matching it was difficult to separate and thus draw significant conclusions about the performance of the optimization methods. Thus the OLYMPUS optimization benchmark challenge was setup and aimed at field development (FD) optimization under uncertainty. In this talk we will provide an overview of the OLYMPUS case and the optimization problems defined. In addition we aim to provide an anonymized overview of validated results from the participants for the OLYMPUS workshop which takes place the day after ECMOR.