Achieving a carbon-neutral, or “net zero” society requires a transition from fossil fuels to low-carbon solutions. However, fossil fuels supply approx. 80% of today’s worldwide energy demand and are projected to play an indispensable role in the immediate future [1]. Carbon capture may be the most effective way to decarbonize the fossil fuel-based energy sector. Carbon capture consists of two major fields: carbon capture and storage (CCS) as well as carbon capture and utilization (CCU). While CCS is more relevant to electricity production, CCU is compatible with the existing downstream processes of the oil and gas industry – the chemicals sector. CCU will be the focus in this thesis. Optimization is applied to explore the maximum performance of CCU. CCU contains multiple process options in both the capture and the utilization sections, eventually resulting in a large multi-process system. Optimizing such a multi-process system can be challenging because of the problem scale and its complexity. The problem scale is significantly larger than a single process and would be challenging to most existing optimization approaches; complexity comes from high-level interactions between sub-systems and the nonlinearity of the individual sub-systems. Optimizing a sub-system before extending it to the whole CCU system can lead to a sub-optimal solution due to the reduced decision space. Using one simulation result to represent a sub-system can neglect the complexity/nonlinearity of the individual processes. In this thesis, I intend to: (1) avoid sub-optimal solutions by simultaneously optimizing the CCU sub-systems, and (2) use surrogates to represent sub-systems to keep a certain complexity/nonlinearity of sub-systems.
This thesis is divided into two parts. Part I is methodology development, engaged in identifying suitable surrogate types for CCU sub-systems and how to obtain surrogates in an efficient way. The methodology development lays the foundation for an optimization framework for large multi-process systems. The optimization framework consists of three levels. Level 1 decomposes a large system into several sub-systems, which are digitalized by rigorous process models. Level 2 replaces rigorous process models with machine learning-based surrogates, as to efficiently evaluate mass and energy balances. Level 3 performs surrogate-based optimization. This optimization framework includes the interactions of sub-systems and optimizes sub-systems simultaneously. Part II is concerned with problem-solving, focusing on optimizing a CCU system (by the three-level optimization framework), where no renewables are involved. The result shows that CCU may be worse for greenhouse gas (GHG) emissions than the conventional (unabated gas) process, if operating conditions are not properly set. Single-objective optimization enables CCU to effectively reduce GHG emissions, and electrifying heating can further cut GHG emissions. Additionally, multi-objective optimization enables CCU to balance the competing criteria between environmental and economic aspects. The methodology developed in this thesis can be applied to other multi-process systems. In the long term, net zero needs various low-carbon pathways, which might integrate different sectors and form multi-process systems. While their decarbonization performances are enhanced by optimization, the overall progress of net zero will be accelerated.

[1] International Energy Agency, Net Zero by 2050: A roadmap for the global energy sector, 2021.