Optimizing carbon capture and storage infrastructure including physics-based reservoir modelling

The deployment of carbon capture and storage (CCS) requires a potentially complex infrastructure to transport and store CO 2 underground. Its optimal roll-out is key to limiting costs and enabling a timely deployment in line with ambitious mitigation scenarios. However, identifying the optimal design of the infrastructure is computationally challenging. Within this framework, mixed-integer linear programming (MILP) offers a computationally effective solution, which, however, requires linear models. Not surprisingly, geological reservoirs are typically represented as static sinks with constant injection rates and storage capacities as parameters. This approach neglects the dynamic properties of CO 2 injection, such as the reservoir pressure evolution over time, limiting the ability to evaluate their impact on the CCS chain design. In this work, we propose a novel MILP model that integrates physics-based reservoir modeling into CCS chain optimization. Extending existing work on reduced order modeling of reservoirs, we combine proper orthogonal decomposition and trajectory piecewise linearization to obtain a precise, yet computational efficient MILP model for the dynamic behavior of CO 2 injection. Compared to full-scale reservoir simulations, the model computes the pressure around the injection well with ± 5% accuracy and significant computational speed-ups (500-1800 times faster). We demonstrate its application in a full chain MILP model with an illustrative case study optimizing the decarbonization of a small industrial cluster through CCS, highlighting the model’s ability to couple optimal operation of capture technologies with varying injection rates and to ensure the reservoir safety constraints while designing the chain. • Novel MILP model integrating dynamic reservoir behavior in CCS chain optimization. • Combines POD and piecewise linearization for efficient reservoir pressure modeling. • Achieves ± 5% accuracy with 500-1800 × speed-up over full-scale simulations. • Applied to an industrial cluster CCS case study including safety constraints.