Minimum-regret hydrogen supply chain strategies to foster the energy transition of European hard-to-abate industries (2407.05988v1)

Abstract: Low-carbon hydrogen (H2) is envisioned to play a central role in decarbonizing European hard-to-abate industries, such as refineries, ammonia, methanol, steel, and cement. To enable its widespread use, H2 supply chain (HSC) infrastructure is required. Mature and economically viable low-carbon H2 production pathways include steam methane reforming (SMR) of natural gas coupled with carbon dioxide capture and storage (CCS), water-electrolysis from renewable electricity, biomethane reforming, and biomass gasification. However, uncertainties surrounding demand and feedstock availabilities hamper their proliferation. Here, we investigate the impact of uncertainty in H2 demand and biomass availability on the optimal HSC design. The HSC is modeled as a network of H2 production and consumption sites that are interconnected with H2 and biomass transport technologies. A CCS supply chain is modeled alongside the HSC. The cost-optimal HSC design is determined based on a linear optimization problem that considers a regional resolution and a multi-year time horizon (2022-2050). We adopt a scenario-based uncertainty quantification approach and define discrete H2 demand and biomass availability scenarios. Applying a minimum-regret strategy, we show that sufficiently large low-carbon H2 production capacities (about 9.6 Mt/a by 2030) are essential to flexibly scale up HSCs and accommodate H2 demands of up to 35 Mt/a by 2050. While biomass-based H2 production emerges as the most cost-efficient low-carbon H2 production pathway, investments are not recommended unless the availability of biomass feedstocks is guaranteed. Instead, investments in SMR-CCS and electrolysis often offer greater flexibility. In addition, we highlight the importance of CCS infrastructure, which is required across scenarios.

• The paper introduces a scenario-based, linear optimization framework using a minimum-regret strategy to design low-carbon hydrogen supply chains for hard-to-abate European industries.
• The study assesses 15 distinct scenarios, revealing that a medium hydrogen demand with no biomass availability yields the most robust investment pathway.
• The analysis emphasizes the critical role of coordinated CO2 capture, transport, storage, and DAC infrastructure to meet net-zero emissions targets by 2050.

This paper (2407.05988) presents a detailed optimization-based framework for designing and rolling out low-carbon hydrogen supply chains (HSCs) in Europe for hard-to-abate industries like refineries, ammonia, methanol, steel, and cement. The core challenge addressed is the deep uncertainty surrounding future hydrogen demand and the availability of biomass feedstock, which significantly impacts the optimal infrastructure design and investment strategy. The authors propose a scenario-based uncertainty quantification approach combined with a linear optimization model and a minimum-regret strategy to identify robust investment pathways from 2022 to 2050.

System Description and Modeling Approach

The hydrogen supply chain is modeled as a network of nodes (representing NUTS 2 regions in Europe) and edges (connecting regions). The model considers various technologies and carriers:

• Feedstocks/Energy: Natural gas, grid electricity, renewable electricity (wind, solar), and biomass (dry and wet).
• Hydrogen Production: Steam methane reforming (SMR) of natural gas or biomethane, water electrolysis from electricity, and biomass gasification. Technologies can be coupled with CO2 capture and storage (CCS).
• Transport: Trucks (compressed gas, liquid hydrogen) and pipelines for hydrogen. Trucks (liquid CO2) and pipelines for CO2. Trucks (dry biomass) and trucks/gas grid (biomethane) for biomass and its derivatives.
• Demand: Existing and projected hydrogen demand from hard-to-abate industries across European regions.
• CO2 Infrastructure: Integrated modeling of CO2 capture (from H2 production), transport, and storage at designated sites. Direct air capture (DAC) is also considered for CO2 removal to achieve net-zero emissions.

The model is formulated as a linear optimization problem implemented in the ZEN-garden framework and solved using Gurobi. It minimizes the total net present cost of the entire supply chain (H2, CO2, biomass) over the 2022-2050 horizon, subject to satisfying hydrogen demands, respecting technology performance limits, ensuring mass balances, and meeting annual CO2 emission targets (a linear decrease to net-zero by 2050).

Handling Uncertainty

Uncertainty is explicitly addressed through discrete scenarios:

• Hydrogen Demand: Five scenarios (minimum, low, medium, high, maximum) are derived from a review of 28 literature sources, capturing the wide range of projections for 2050 demand (2.4 Mt/a to 40 Mt/a). These scenarios define the temporal evolution of demand by industry and region.
• Biomass Availability: Three scenarios are defined based on sustainable biomass potential: a reference case, a reduced availability case (26% of reference), and a no biomass case, reflecting competition with other sectors.

Combining these uncertainties results in 15 distinct scenarios for which the optimal HSC design is initially determined (design scenarios).

Minimum-Regret Strategy

To identify a robust strategy under deep uncertainty, the authors employ an out-of-sample approach. For each design scenario, the optimal infrastructure built under that specific set of demand and biomass assumptions is then evaluated against all 14 other scenarios. Crucially, the model includes technology expansion constraints, limiting the annual growth rate of installed capacity based on existing capacity stock and historical growth rates. This simulates the real-world limitation on how quickly the infrastructure can adapt if the future unfolds differently than planned.

The performance under these out-of-sample scenarios is assessed based on the Levelized Cost of Hydrogen (LCOH) and the ability to meet the annual CO2 emission targets (feasibility). A design scenario is deemed feasible only if it can meet the CO2 targets across all other scenarios within the limits imposed by the expansion constraints. The minimum-regret solution is the feasible design scenario that results in the lowest LCOH in its worst-performing out-of-sample scenario (a min-max cost criterion).

Key Findings and Practical Implications

1. Technology Portfolio: Biomass-based hydrogen production (biomethane reforming, gasification with CCS) is identified as the most cost-efficient low-carbon pathway, especially early on. However, its deployment heavily depends on guaranteed biomass availability. SMR with CCS and electrolysis are alternative options, with electrolyzers becoming more competitive over time as costs decrease and if high renewable electricity shares are available. The increasing share of electrolyzers for higher demand scenarios reduces the need for large CO2 transport networks compared to SMR-CCS dominant scenarios.
2. Infrastructure Needs: The required scale of H2 and CO2 transport infrastructure correlates with H2 demand. However, relying on biomass necessitates a more geographically widespread H2 transport network to connect dispersed resources to demand centers. CCTS infrastructure (capture, transport, storage, and DAC) is universally required across scenarios to meet net-zero targets, either for capturing emissions from SMR/biomass gasification or for removing residual/upstream emissions via DAC.
3. Role of CCTS: CCTS is central to decarbonization. The capacity of existing/planned CO2 storage sites (estimated at 132 Mt/a) may be fully utilized by 2050 under high demand scenarios, particularly without significant biomass availability, requiring substantial investment in CO2 removal technologies (DAC). This highlights the urgent need for accelerated development and deployment of CCTS infrastructure.
4. Impact of Uncertainty & Robustness: While scenarios with high biomass availability offer lower LCOH in ideal conditions, they are less robust. If biomass is less available than expected, it becomes difficult to quickly scale up alternative technologies like electrolyzers and CO2 removal within the expansion constraints, potentially leading to failure to meet emission targets.
5. Minimum-Regret Strategy Characteristics: The minimum-regret solution is found to be the design based on medium hydrogen demand and no biomass availability. This strategy involves investing in a mix of SMR-CCS and electrolyzers, along with necessary CCTS infrastructure. This initial investment, while potentially not the lowest cost in optimistic biomass scenarios, provides sufficient flexibility and existing capacity in key technologies (electrolyzers, CO2 removal) to scale up adequately if high demands materialize or biomass is scarce, ensuring feasibility across a wide range of futures. The model suggests planning for about 9.6 Mt/a of low-carbon H2 production capacity by 2030 is essential for future flexibility.
6. Policy Implications: The paper emphasizes the need for clear European strategies regarding the role of biomass (directing it to specific sectors) and coordinated development of CCTS infrastructure. Incentivizing early investment in low-carbon H2 production capacity, particularly technologies with lower residual emissions and faster expansion potential (like electrolyzers and DAC), is crucial to navigate uncertainty and ensure timely decarbonization.

Implementation Considerations

The model is a complex linear program, requiring powerful solvers (like Gurobi). The scenario-based approach handles deep uncertainty but involves solving the optimization problem for multiple scenarios and evaluating out-of-sample performances, which can be computationally intensive. The inclusion of technology expansion constraints adds realism but limits the adaptability of initial designs, highlighting potential bottlenecks in infrastructure rollout. Spatial resolution at NUTS 2 allows for regional specificity but increases model size. The data for techno-economic parameters and future projections inherently carries uncertainty, and sensitivity analyses (e.g., on technology expansion rates, or optimistic cost scenarios for specific technologies) are important to test the robustness of the findings. The paper makes code and data available on Zenodo for reproducibility.