AutoSum for a Generic CO2 Storage Project

Summary Working with a representative ensemble of potentially 1000s of reservoir models that captures relevant uncertainties is necessary in most Integrated Reservoir Modelling workflows. Only if known uncertainties are properly captured and understood, can robust and uncertainty aware business decisions can be made. In practice, this is unfortunately still not done often enough since current Integrated Reservoir Modelling tools were not designed to allow for efficient handling and analysis of a large ensemble of models. We are addressing the challenge of providing actionable insights into complex model behaviour, by combining traditional physics-based modelling with state-of-the-art data analytics and techniques from interpretable AI for post-processing the modelling results and representing them in a more human accessible form. We present a unifying platform to bring in machine learning approaches to Integrated Reservoir Modelling in an algorithm and simulator agnostic way.