The European Union has set out to become carbon neutral by 2050. To reach this goal, the fossil-fuel-dominated energy system must be transformed into a low-carbon renewable-based energy system. One method proposed to reduce fossil-fuel dependence is to establish a renewable hydrogen supply chain between Africa and the European Union. This work provides a model framework that enables an economic analysis of a renewable hydrogen supply chain between northern Africa and the EU. The aim of this work is to gain a better understanding of large-scale intercontinental renewable hydrogen production by investigating the factors that impact the economic performance of this system.
A mixed-integer linear programming model that describes the supply chain is built, and modeled in Python and a case study defining the overall energy system under consideration and applied to this model. The system must provide 780 TWh of hydrogen demand for the EU in 2050. Hydrogen is produced by water electrolysis, using renewable electricity from solar photovoltaics. The water required for electrolysis is supplied through reverse osmosis of seawater. Seasonal storage of hydrogen is enabled through the inclusion of hydrogen storage in salt caverns.

The resulting annual costs of operating the supply chain are 36.55 B$, corresponding to a levelized cost of hydrogen of 1.56 $kg^-1. The main contributors to the cost of the supply chain are solar photovoltaics (50\%), alkaline electrolysis (22%), and transportation (26%). Through a sensitivity analysis on price uncertainty, it is found that the system is most sensitive to photovoltaics prices (59%), electrolysis prices (21%), and hydrogen pipeline prices (18%). A sensitivity analysis of the interest rate on capital investment points towards a significant impact of the interest rate on the total annual cost.

The findings of the model and applied case study in this work are then compared to projected costs in other works. From this analysis, it is found that hydrogen production using electrolysis will be cheaper than fossil-based low-carbon hydrogen production alternatives in the form of steam methane reforming or coal gasification with carbon capture technology. Predictions of the selected works regarding electrolysis-based hydrogen production using renewable electricity show an expected LCOH of between 1.66 - 2.39 $kg^-1.

This work indicates that a renewable hydrogen supply chain between Morocco and the EU in 2050 is both technically and economically feasible, able to compete with alternative hydrogen production methods, and able to supply 35% of the projected hydrogen demand. Moreover, this work developed a general hydrogen supply-chain model that allows for the implementation of additional features as well as the analysis of different case studies through the adjustment of case-specific parameters.

Recently, ReLU neural networks have been modelled as constraints in mixed integer linear programming (MILP) enabling surrogate-based optimisation in various domains as well as efficient solution of machine learning verification problems. However, previous works have been limited to multilayer perceptrons (MLPs). The Graph Convolutional Neural Network (GCN) model and the GraphSAGE model can learn from non-euclidean data structures efficiently. We propose a bilinear formulation for ReLU GCNs and a MILP formulation for ReLU GraphSAGE models. We compare our formulations to a Genetic Algorithm (GA) by comparing solution times and optimality gaps while modelling a dataset of boiling points of different molecules. Our method guarantees to solve optimisation problems with trained GNNs embedded to global optimality. Between our two formulations the GraphSAGE neural network achieves similar model accuracy, and achieves faster solving times when embedded as a surrogate model in an MILP problem. Finally, we present a computer aided molecular design (CAMD) case study where the formulations of the trained GNNs are used to find molecules with optimal boiling points.