Cristin-resultat-ID: 2158157
Sist endret: 26. juni 2023, 14:52
Resultat
Vitenskapelig foredrag
2023

An Extended Fully-Implicit Hybrid Model for Simulating CO2 Storage and Migration in Vertically Heterogeneous Formations

Bidragsytere:
  • Vetle Nevland
  • Odd Andersen og
  • Halvor Møll Nilsen

Presentasjon

Navn på arrangementet: Trondheim Carbon Capture & Storage Conference (TCCS) 12
Sted: Trondheim
Dato fra: 20. juni 2023
Dato til: 23. juni 2023

Arrangør:

Arrangørnavn: SINTEF

Om resultatet

Vitenskapelig foredrag
Publiseringsår: 2023

Beskrivelse Beskrivelse

Tittel

An Extended Fully-Implicit Hybrid Model for Simulating CO2 Storage and Migration in Vertically Heterogeneous Formations

Sammendrag

Numerical simulation of CO2 migration in geological formations is computationally challenging for multiple reasons, including accurate representation of the particular physics of the problem, requirements on grid resolution, and the large spatial and temporal scales typically involved. Due to the high aspect ratio of typical aquifer models and the strong density contrast between injected CO2 and aquifer brine, models based on vertical equilibrium (VE) have proven successful in the past to accurately model CO2 migration while reducing computational requirements by orders of magnitude [1]. These models work by reducing the dimensionality of the system from 3D to 2D, with post-hoc reconstruction of the full 3D solution based on the VE assumption. While this approach is often ideal for high-permeable aquifers with limited vertical heterogeneity, it cannot be directly applied on multilayered systems with interbedded impermeable or low-permeable layers, as the assumption of vertical equilibrium does not globally hold. However, both actual and prospective geological storage formations often exhibit internal layering [2,3]. To address this and similar issues, hybrid models have been proposed [3,4,5,6]. Such models provide a middle-ground between VE and full-dimensional models, using VE in high-permeable regions while retaining full dimensionality in low-permeable interbeddings and other regions where the VE assumption does not hold. In the work presented here, we demonstrate the hybrid modeling approach to study migration and trapping of CO2 in storage systems with multiple interspersed low-permeable layers, using the open-source MATLAB Reservoir Simulation Toolbox (MRST). MRST provides a framework for simulator prototyping that works on industry-standard grids and complex geometries, based on a black-oil formulation of the flow equations and fully implicit time discretization. In layered storage systems, diffuse leakage, where CO2 buildup breaches the capillary seal of internal barriers and slowly permeates upwards, needs to be considered in the model. To study long-term CO2 migration and construct trapping inventories, it is also important to properly represent residual trapping. Support for capillary entry pressure and residual saturation was therefore added to the hybrid model. In a VE setting, the modeling of residual saturation in conjunction with diffuse leakage must be carefully handled, as the slow, gradual drainage through low-permeable internal layers violate the VE assumption. We discuss some of the modeling challenges, and present results from applying our enhanced hybrid model on several test cases of varying complexity. We start by discussing some conceptual cases particularly designed to demonstrate diffuse leakage and the handling of residual saturation in a hybrid setting, where we benchmark the hybrid against the full 3D solutions. We also apply the model to study a synthetic case of upward CO2 migration through a very thick formation interspersed with many internal barriers. Lastly, we test the approach on the SPE 11 benchmark case [7]; a multilayered laboratory-scale reservoir representative of the geology of the Norwegian Continental Shelf. In contrast to subsurface reservoirs, this benchmark case includes complete geometrical and petrophysical data, facilitating ensemble modeling. Whereas full-dimensional models are often too expensive for this purpose, hybrid models are feasible alternatives, substantiating the importance of developing novel hybrid models.

Bidragsytere

Vetle Nevland

  • Tilknyttet:
    Forfatter
    ved Universitetet i Oslo

Odd Andersen

  • Tilknyttet:
    Forfatter
    ved Mathematics and Cybernetics ved SINTEF AS

Halvor Møll Nilsen

  • Tilknyttet:
    Forfatter
    ved Norges teknisk-naturvitenskapelige universitet
1 - 3 av 3