Computational Strategies for RTN Model for Supply Logistics of Carbon Dioxide for Carbon Capture and Storage

In this paper, we address the solution of a large-scale mixed-integer linear programing (MILP) model to maximize profit for shipping cryogenic carbon dioxide in Carbon Capture and Storage field management systems. The model is based on a discrete-time Resource Task Network as discussed in Guisso et al. (2024), where inventory levels of carbon dioxide at the ports are determined along with the decision variables determined at each time interval. To solve the resulting large-scale MILP model, decomposition techniques based on bilevel decomposition and two-stage optimization decomposition are first proposed for the simpler case where trips (or milk runs) between emitter ports are not considered. For the real-life case that allows milk runs, a Lagrangean decomposition is proposed with a shrinking time horizon strategy for the solution of subproblems for long time horizons. Numerical results are presented to illustrate the application of the proposed decomposition techniques, which show that significant computational savings up to 1 order of magnitude reduction can be achieved.