Abstract
Hydrogen has come to the forefront of the global energy transition. It can support the integration of variable renewable energy sources and serve as a renewable feedstock in applications where direct electrification is not feasible. Despite its potential, hydrogen currently plays a minor role in energy systems due to technical challenges, high costs, and uncertain demand.
To analyze investments in future energy systems, mathematical optimization tools, such as capacity expansion models, are widely used in the academic literature. Unfortunately, modeling hydrogen within capacity expansion models remains challenging, as it requires a certain level of detail to represent its applications across different sectors and its interconnections with other energy carriers. To remain computationally feasible, capacity expansion models simplify real-world dynamics, which results in a trade-off between complexity and accuracy.
In the first part of the thesis, the current literature is synthesized, and five key modeling requirements are identified to accurately represent the unique characteristics of hydrogen. In this context, the underlying methodologies of 11 open-source models are discussed. The analysis shows that no single model can capture all dimensions simultaneously, and that some models are more effective at representing certain aspects of hydrogen.
In the second and third parts of the thesis, two methodologies are developed to better manage the trade-off between model detail and computational feasibility. First, the temporal dimension is improved with a novel clustering method for time-series aggregation in sector-coupled capacity expansion models. This method achieves a higher level of accuracy with lower computational cost compared to state-of-the-art solutions. Second, a machine learning-based surrogate modeling methodology is developed to reduce the cost of sensitivity analysis.
This thesis contributes to a better understanding of complex modeling decisions related to hydrogen. It also proposes two computationally efficient methodologies that can be applied in future case studies to better capture the complex dynamics of renewable energy systems with hydrogen.