This preprint examines data infrastructure requirements for hydrogen energy system planning in the context of the UK Department for Energy Security and Net Zero (DESNZ) 2026 call for evidence on energy datasets for artificial intelligence applications. While current policy discussions emphasise artificial intelligence primarily for prediction and pattern recognition, hydrogen infrastructure planning depends equally on mathematical optimisation methods used for long-term infrastructure design. The paper argues that effective hydrogen planning requires integrated data systems capable of supporting both machine learning and optimisation models simultaneously. It proposes a minimal three-layer data architecture comprising physics, operations, and economics data. These layers must be interoperable and governed through common identifiers, unit conventions, version control, and uncertainty protocols to enable consistent modelling across analytical approaches. The study further discusses five policy choices that would shape the development of such infrastructure: dedicated funding for energy data systems, early interoperability standards, creation of a trusted data intermediary, confidentiality and security governance, and cross-disciplinary workforce development. Hydrogen infrastructure planning is used as a stress test for energy data architecture because of its capital intensity, spatial complexity, cross-sector coupling, and high demand uncertainty. The paper contributes a conceptual framework for integrating machine learning and optimisation in hydrogen system planning and highlights the importance of treating data architecture as a core policy design problem rather than a secondary modelling issue.
Gordhan Das (Wed,) studied this question.