(214g) Distributed Model Predictive Control for UK CO2 Capture, Transport and Storage Network

The large scale deployment and operation of CO₂ capture, transport and storage (CCS) infrastructure will be extremely capital intensive. Although it is possible, using spatially and temporally explicit methods, to design a CO₂ capture and transport network which will minimise the lifetime costs associated with the network [1], it is important to realise that the CO₂ sources will themselves operate in a dynamic fashion. Indeed, the largest fixed point source emitters of CO₂, i.e., power stations, operate in a very dynamic fashion. Thus, efficient operation of the CCS network is key to minimising the costs associated with the operation of the network. For this, a state-of-the-art control technology is required.

The CCS network on which our current work is based on was optimally designed for deployment UK which and is comprised of 33 sources and 6 offshore sinks with three onshore terminals[2]. All sources, sinks and terminals are connected via pipeline. In this contribution, we develop a dynamic distributed model predictive control which concurrently maximises the CO₂ storage capacity and minimizes the operation cost whilst meeting infrastructure physical constraints such as availability, storage and pipeline capacity.

Owing to the high computational burden associated to such a dynamic optimization problem, the control framework utilized herein is based on the distribution of the overall network control effort among local agents presented in previous work[3]. Such an approach allows very large network simulation and management to be performed through parallel computation with little hardware requirements. As it is based on Lagrange-based distributed predictive control, the proposed algorithm uses a dynamic model of the CCS network taking explicitly into account system constraints while optimizing a multiobjective control function of the overall network, dividing this smart grid into subnetworks and assigning them to local control agents. Consequently, the computed set of local control moves is equivalent to those corresponding to the optimal solution of the overall network optimization problem. The dynamic model utilized is based on the so-called network hubs[4], which are defined as interfaces between input and output flows and the transportation infrastructure.This concept offers a number of very convenient features that makes it a very flexible and standard formulation in order to model and analyze general complex networks including multiple flows and storage.