In the LuxHyVal project, partners, including the Luxembourg Institute of Science and Technology (LIST), adopt a model-based approach. By using explicit, executable models, they simulate and analyse the hydrogen valley from the very first design steps. This method supports better decision-making for LuxHyVal and provides a transferable blueprint for future hydrogen valleys across Europe.
LuxHyVal, a five-year-long project funded by the European Commission Horizon Europe Programme, aims to contribute to the achievement of the European target on decarbonisation of the industrial sector before 2030 by launching a flagship hydrogen valley in Luxembourg to boost the penetration of hydrogen.
Model-Driven Digital Twins: Rethinking Digitalisation for Hydrogen Valleys
Digital twins have become a cornerstone of energy system digitalisation. However, most existing implementations are still overwhelmingly data-driven, relying on sensor measurements, operational logs and historical datasets, often combined with machine-learning techniques. In large green hydrogen initiatives such as LuxHyVal, this paradigm reveals a fundamental weakness: at design time, the system does not yet exist, and therefore no operational data is available.
In the LuxHyVal project, this concept takes on renewed importance as partners develop digital tools to accelerate Europe’s hydrogen transition and support the rollout of hydrogen valleys. This article introduces an alternative approach: Model-Driven Digital Twins (MDDT), in which explicit models form the backbone of the digital twin from the very earliest project phases.
Beyond data-centric digital twins
Traditional digital twins are typically built as data platforms, aggregating measurements and events and extracting behavioural patterns from them. Industrial solutions such as AWS IoT TwinMaker or Azure Digital Twins illustrate this vision, with a strong emphasis on connectivity, data ingestion and real-time processing. While effective once systems are operational, this approach largely sidelines explicit models that describe structure, intent and expert knowledge.
In a Model-Driven Digital Twin, these models become first-class artefacts. They shape both the design and the operation of the system, providing structure when data is incomplete or entirely absent. For hydrogen valleys – where infrastructure, markets, regulations and technologies evolve simultaneously – this model-centric perspective is essential to understand interactions that raw data alone cannot reveal.
Designing the twin before the system exists
The proposed methodology deliberately starts long before any physical asset is installed, effectively creating a “virtual digital twin” of the future hydrogen valley. Design-time models are treated as the embryo of the future operational twin, rather than as disposable planning artefacts.
Key modelling steps include:
- Identifying stakeholders and defining system boundaries, such as which mobility segments are included in the first deployment.
- Developing high-level architectural views that combine enterprise architecture (actors, value chains, services) with system and manufacturing architecture (technical components and physical assets).
- Modelling the operational context, including weather patterns, demand evolution, urban development and policy incentives.
These models enable early simulations that resemble digital-twin “what-if” analyses, even though the physical system does not yet exist. This virtual twin allows project partners to detect design flaws, compare alternatives and reduce technical and economic risks well before construction begins.
Simulating reality before it happens
A central element of the MDDT approach is the use of executable models to simulate system behaviour from the outset. In LuxHyVal, this means combining estimates of renewable electricity production with projected hydrogen demand from public transport and industry, while accounting for meteorological constraints.
Such simulations help calibrate critical parameters, including electrolyser sizing, storage capacity and the coupling between renewable power availability and hydrogen output. At this stage, the twin is entirely model-driven. Importantly, the distinction between design models and the future run-time digital twin is intentionally blurred, ensuring that design assumptions, constraints and rationale are preserved as the system transitions into operation.
This continuity is particularly valuable for green hydrogen projects, which typically have long lifecycles and operate in regulatory environments that evolve over time. In other words, MDDTs help teams make better design choices even before breaking ground, while keeping a clear digital trace of why those choices were made.
Structuring data collection and physical system alignment
Model-Driven Digital Twins also play a key role in preparing data collection and system control once the hydrogen valley becomes operational. During later design phases, architecture models are annotated to specify which data should be collected – whether from sensors, operational systems or external databases – and which actions can be triggered through actuators or control systems.
As parts of the system go live, the MDDT enters an Alignement phase, where real-world measurements are compared with model-based predictions and/or simulations. Deviations may reveal sensor faults, unexpected operating conditions or the need to refine the models themselves. Once alignment reaches an acceptable level, the Exploitation phase enables full digital-twin services: monitoring, prediction, optimisation and exploratory what-if analyses for operators and decision-makers.
Enabling transfer to other hydrogen valleys
One of LuxHyVal’s key ambitions is to serve as a blueprint for hydrogen valleys beyond Luxembourg. The MDDT is a central enabler of this transferability. By explicitly modelling contextual factors – such as mobility demand, industrial structure, public subsidies and local renewable resources – the twin can be adapted to different national or regional settings.
In a new country, contextual models are adjusted to local conditions, turning the MDDT into a decision-support tool for policymakers and investors assessing hydrogen penetration, infrastructure needs and investment priorities. Using a consistent, executable modelling approach across contexts supports systematic transfer rather than ad-hoc replication, amplifying the strategic impact of LuxHyVal within the broader green hydrogen ecosystem.
The key message is simple: digital twins for hydrogen valleys should not be late-stage, data-only add-ons. They should be model-driven companions, evolving continuously from the very first design workshop to large-scale deployment and replication.
This article is based on the conference paper titled “How to Leverage Digital Twin for System Design?” presented at the 12th International Conference on Smart Cities and Green ICT Systems (SMARTGREENS 2025). The conference paper was co-authored by Jean-Sébastien Sottet, Pierre Brimont, Cédric Pruski, and Faima Abbasi from the Luxembourg Institute of Science and Technology and the University of Luxembourg, exploring the use of MDDTs with the Luxembourg hydrogen valley as a case study.
The full conference paper is available here.
The LuxHyVal project has received funding from the European Union’s Horizon Europe research and innovation programme and is co-funded by the Clean Hydrogen Joint Undertaking.
