The research, carried out by Edrisi Munoz and Turhan Demiray at the Research Center for Energy Networks of ETH Zurich, addresses a problem that has shadowed energy system modeling for decades. Cities now house roughly 45 percent of the world’s estimated 8.2 billion people, up from about 20 percent in 1950, and United Nations projections suggest around 68 percent of humanity will live in urban areas by 2050. Deciding how those cities will be powered requires models that explore pathways to a sustainable future, yet those models embed countless assumptions about future conditions that are rarely documented in a form anyone can audit. When two studies claim to analyze the same system, there is often no way to verify that they actually represent the same input parameters and variables.

The consequences go beyond academic inconvenience. The authors point to a growing reproducibility and replicability crisis in energy research, echoing recent arguments that energy system models are too often applied and communicated as projection machines when their real evidentiary value lies in their use as diagnostic tools that expose assumptions, explore uncertainty, and support well-documented decision-making. Reproducibility, in this context, means the ability to reproduce earlier results using published reports and data, while replicability means obtaining similar results from newly acquired data. Without a formal record of what a scenario assumed, neither is achievable. Complementary quality-assessment methods such as NUSAP and sensitivity auditing have been proposed to flag where model-based energy-policy conclusions carry more certainty than the underlying evidence can support, but the field has lacked a structural fix.

The Zurich team’s answer is a digital framework called the Scenario Management Ontology, expressed in the standard semantic web languages RDF and OWL and developed in the Protégé ontology editor. An ontology, in this sense, is a formal, machine-readable vocabulary that defines the concepts of a domain, their properties, and the relationships between them. Rather than treating a scenario as a document, the ontology treats it as a first-class, versioned information object. The framework distinguishes a baseline scenario, which is a complete and self-contained representation of the system under study, from sub-scenarios, which are stored only as the changes, or deltas, they introduce relative to a reference scenario. Explicit properties such as hasBaselineScenario, derivedFrom, and derivesIn chain every variant back to a shared root, so the full derivation history of any scenario can be queried and audited.

The ontology defines classes for the essential ingredients of any modeling exercise: the defined system, model key features, scenario-driven features, input and output data, system boundaries, uncertainty, and risk. Data properties capture metadata that is usually scattered across reports, including base year, target year, time horizon, spatial resolution, temporal resolution, assumptions, and a universal scenario code analogous to the universal product code used in retail. Uncertainty and risk each receive dedicated classes with quantified measurement properties, allowing a scenario to attach a probability to an assumption rather than merely a point estimate. The authors also map these elements onto standard operations-research terminology: fixed parameters correspond to model key features, decision variables to scenario-driven features, and constraints to the type and range restrictions enforced by schema validation, so that a physically meaningless input, such as a conversion efficiency greater than one, is rejected before it ever reaches a solver.

To describe the energy system itself, the framework builds on the Common Energy System Domain Model, or CESDM, a schema-based framework developed at the same research center. CESDM’s guiding principle is that the structure of a system should be made explicit rather than its behavior. An energy system is described through entities such as regions, nodes, and conversion technologies; attributes such as installed capacity and conversion efficiency; and typed relationships connecting them. Schemas act as formal contracts, specifying what each entity type must, may, and cannot contain, and schema inheritance lets specialized types, such as a one-input one-output conversion technology, inherit structure from their parents without duplication. Because schemas are machine-verifiable, a model can be validated automatically instead of inspected by hand, and the schemas remain stable and shareable across modeling tools while the models built from them vary freely between studies.

The proof of concept is ambitious: the Ten-Year Network Development Plan scenarios published jointly by the European Network of Transmission System Operators for Electricity and Gas, which provide the long-term analytical foundation for European energy infrastructure planning. As released, the TYNDP 2024 storylines, National Trends Plus, Distributed Energy, and Global Ambition, exist as narrative text, tables, and dashboards rather than validated, queryable datasets. The researchers converted this material into eight CESDM-conformant scenario files organized in a single index. Only one, the 2030 horizon of National Trends, is a full baseline, instantiating every entity type the schema defines: ten energy carriers, 47 geographical regions, 70 energy nodes, 97 demands, 598 conversion technologies, 128 storage technologies, and 131 network transfer capacities, among others. Every other scenario is stored as a change-set against its reference, split into added, modified, and deleted components.

The compactness of that representation is deceptive. The Distributed Energy 2030 sub-scenario, for example, adds 131 new demand entities for electric-vehicle, prosumer, and electrolysis load, modifies 331 of the baseline’s conversion technologies, and deletes 184, concentrated among aging centralized oil, lignite, hard-coal, and gas units. In total, 515 of the baseline’s 598 conversion technologies, some 86 percent, are either modified or removed, demonstrating that Distributed Energy represents a structural reshaping of the generation mix even though it is stored economically as a patch. Because each sub-scenario references its immediate predecessor rather than the baseline directly, resolving a 2050 scenario is a recursive operation, applying the 2050 changes to the resolved 2040 state, which itself applies 2030 changes to the baseline. Three narrative threads spanning three time horizons are thus maintained in eight files instead of repeated full snapshots of a system with more than a thousand entities.

Interoperability is where the semantic machinery pays off. Each scenario file opens with a metadata block whose fields map directly onto the ontology’s classes and properties, so the blocks can be mechanically converted into RDF and loaded into the Apache Jena Fuseki triple store as a small knowledge graph of one baseline node and seven sub-scenario nodes. Questions that would be awkward to answer by searching through files, such as which scenarios share a target year across storylines or what the full derivation chain from baseline to a given horizon looks like, become single SPARQL queries. The same validated model can then be exported in three consistent formats: human-readable YAML for inspection and version control, CSV tables whose columns are defined by the schema, and Frictionless Data Packages that carry machine-readable descriptions of their own structure, allowing any compliant tool to load the data without prior knowledge of CESDM.

The work arrives amid a broader shift in how the energy field handles knowledge. The Open Energy Ontology has established a shared vocabulary for energy systems analysis, and separate efforts have formalized co-simulation configurations and building digital twins, but the Zurich framework is distinctive in formalizing the scenario itself, with lineage, versioning, and provenance, while referencing an external domain ontology rather than embedding domain concepts. The authors note that the same scenario-versus-domain separation could in principle extend to less centralized settings such as renewable energy communities, provided a suitable domain ontology is supplied, and they leave a systematic benchmarking against emerging standards for local electricity markets as future work. The researchers also trace the evolution toward the TYNDP 2026 framework, which replaces multiple alternative transition pathways with a single Central Scenario plus high and low economic variants, a methodological shift from comparing transition narratives to assessing uncertainty within a common policy trajectory. For a discipline whose conclusions shape billion-euro infrastructure decisions, the message is that the assumptions behind the numbers deserve the same rigor as the numbers themselves.

Subject of Research: A scenario management ontology for formalizing assumptions and improving traceability in energy system modeling

Article Title: Improving decision-making in energy system modeling by formalizing the scenario-based approach using knowledge management tools

Article References: Munoz, E., & Demiray, T. (2026). Improving decision-making in energy system modeling by formalizing the scenario-based approach using knowledge management tools. Frontiers in Chemical Engineering, 8, Article 1932769. https://doi.org/10.3389/fceng.2026.1932769

Image Credits: AI Generated

DOI: 10.3389/fceng.2026.1932769

Keywords: energy system modeling, scenario analysis, ontology, semantic web, knowledge management, TYNDP, CESDM, reproducibility, energy planning, interoperability, uncertainty, power grids