Dynamic Real-Time Optimisation of an Amine-Based Post-Combustion CO2 Capture Facility using Single-Level Nonlinear Model Predictive Control
Master thesis
Permanent lenke
http://hdl.handle.net/11250/2560386Utgivelsesdato
2018Metadata
Vis full innførselSamlinger
Sammendrag
A complete model of a CO2 capture facility has been optimised with the aid of Dynamic Real-Time Optimisation (DRTO) utilising single-level, Nonlinear Model Predictive Control to merge regulatory and economic objectives. The goal has been to, during 24 hours, minimise the cost related to the energy consumption in the reboiler by variable solvent regeneration, whilst achieving a specified accumulated, or overall, capture ratio of CO2 at the end of the simulation horizon. An hourly varying price of energy with a period of 24 hours have been included in the optimisation problem.
The complete model is based on a previous model from Cybernetica AS, the original model, with model reductions suggested by Hotvedt (2017) for the absorber, desorber and heat exchanger. The suggestions included modelling using molar amounts as state variables for each substance in the facility and discretising the unit models in space using control volumes. The complete reduced model has been validated against the original model in addition to instrumental measurements from an existing test facility at Tiller in Trondheim. It was found the reduced model yielded adequate behaviour although with deviations from both the original model responses and instrumental measurements. Introduction of simple estimator; bias updating, removed the deviations significantly. Eigenvalue analysis of the original and the reduced model were performed, and results show that the reduced model yielded only minor reductions in stiffness. On the other hand, the reductions decreased the dimension of the state space with 225 states, resulting in a simulation time reduction of 73%.
The DRTO was designed using the infeasible soft-constraint method where constraints on the energy costs have been set infeasible. Results from simulation show that the DRTO is able to achieve the reference accumulated capture ratio after 24 hours in addition to utilise the time varying price of energy to minimise cost. The performance was compared to a basic case where the accumulated capture ratio of CO2 was forced constant during the prediction horizon, obtaining a constant solvent regeneration, and a cost reduction of 13.0% and 10.9% was found using a reference value for the accumulated capture ratio of 85% and 91% respectively. The DRTO was further tested for robustness by firstly inducing a step change in inlet conditions of the flue gas, secondly by abruptly increasing the price of energy and lastly by applying stricter constraints on the reboiler duty. The DRTO accomplished the capture goal in all cases, except for a step in inlet conditions close to the end of the simulation horizon. Lastly, the optimal solution resulted in unnecessary use of reboiler duty analysing a simulated plant replacement model, and consequently was bias updating introduced to enhance cost minimisation.