Researchers in Korea have developed an artificial intelligence-guided framework that sharply reduces the computational work needed to optimise solid oxide electrolysis cells (SOECs), a leading technology for producing green hydrogen from steam. The method, published in Applied Thermal Engineering, uses machine learning to guide high-fidelity simulations toward the most informative operating conditions, enabling faster identification of operating regimes that improve hydrogen yield while preserving thermal stability.
What the team did
Solid oxide electrolysis cells operate at high temperatures and are among the most efficient technologies for producing hydrogen from water. However, engineers typically need thousands of computationally intensive simulations to explore the many variables that determine performance and longevity. The Korean researchers combined computational fluid dynamics (CFD) with an AI-driven active learning strategy so that the model learns from each completed simulation and predicts which new operating conditions will be most informative.
Rather than exhaustively evaluating every parameter combination, the framework focuses computational resources on the regions of operation most likely to improve understanding and performance. The published paper was made available online on 25 June 2026 and appears in Volume 303, Part 1, of Applied Thermal Engineering on 1 August 2026.
Balancing competing objectives
Importantly, the framework does not seek a single numerical optimum. Instead it identifies a Pareto-optimal operating region — a set of operating conditions that represent the best trade-offs between two competing objectives:
- Maximising electrochemical performance (higher hydrogen production efficiency)
- Minimising temperature differences across the cell (which can cause material degradation)
By highlighting a region of acceptable trade-offs rather than one operating point, engineers retain flexibility to choose conditions that suit particular priorities, such as short-term efficiency versus long-term durability.
Why this matters for South Africa
South Africa is actively exploring green hydrogen as part of industrial decarbonisation and economic development strategies. Technologies that improve the efficiency and longevity of electrolysers can lower costs and speed deployment. The new AI framework can:
- reduce development time and computational expense when designing SOEC systems;
- help equipment developers and plant operators choose operating regimes that balance output and component life;
- support faster iteration of designs, potentially attracting investment into local manufacturing and demonstration projects.
These gains are especially relevant where the capital cost of devices and the cost of simulation time are bottlenecks to innovation.
Limitations and next steps
The study demonstrates a methodological advance rather than a finished commercial product. The framework relies on the quality of the CFD models and the representativeness of the simulated operating space. Real-world SOEC systems face additional uncertainties — such as manufacturing variability, impurities in feedstock, and operational transients — that must be validated with experiments and pilot plants. The researchers present the method as a way to guide and reduce the number of costly experiments needed, not to replace them.
Future work typically follows two paths: adding more physics and real-world constraints into simulation models, and linking the AI-guided optimisation to laboratory and pilot-scale testing so that predictions can be validated and refined.
| Objective | Reason |
|---|---|
| Maximise electrochemical performance | Higher hydrogen production efficiency |
| Minimise temperature differences | Reduce material degradation and extend device life |
The Korean team's approach is an example of how machine learning can make high-fidelity modelling more practical by prioritising the most informative simulations. For South African researchers, industry and policymakers, the advance offers a tool to accelerate mature technology development while keeping an eye on longevity — a balance that will influence the cost and feasibility of large-scale green hydrogen projects.
