A Coarsening Approach to the Troll Aquifer Model

TL;DR

This paper introduces a coarsening method for reservoir simulation models that simplifies complex subsurface representations, significantly reducing computational costs while maintaining key physical properties, demonstrated on the Troll aquifer model.

Contribution

The paper presents a novel coarsening approach that preserves physical properties and enables large-scale simulations efficiently, applied to the Troll aquifer model with open-source implementation.

Findings

Coarsened model improves pressure distribution predictions.

Approximately 100,000 simulations completed in five days.

Method reduces computational time from seven months to five days.

Abstract

Effective management of pressure communication and interference between concurrent CO$_2$ storage operations is essential for the development of gigaton-scale storage hubs. Coarse models in reservoir simulation offer a simplified representation of the subsurface to efficiently predict pressure distribution. By averaging properties over larger grid blocks, coarsened models significantly reduce computational demands, enabling faster and more manageable simulations. The approach focuses on preserving key physical properties, namely cell connectivity (transmissibilities) and storage capacity (pore volumes). The effectiveness of the coarsened model is demonstrated by applying the improved well location derived from the coarsened model to the full-resolution Troll aquifer model, resulting in improved pressure distribution. The coarsened model enabled the execution of approximately 100,000 simulations over five days using a local server with 144 CPUs, an effort that would have required around seven months using the original model. The coarsening methodology is implemented using pycopm, an open-source tool design to tailor geological models based on industry-standard grid formats, facilitating its application to other models and applications in reservoir engineering.