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

1 Introduction

The global demographic dynamics in the early twentieth century reflect a profound restructuring of human habitation, characterized by rapid urbanization and a corresponding decline in the relative share of rural dwellers. According to the United Nations Department of Economic and Social Affairs, cities currently accommodate approximately 45% of the estimated 8.2 billion people around the world, marking a substantial rise from about 20% in 1950 and underscoring the importance of urban centers in contemporary settlement patterns (United-Nations, 2025). Projections through 2050 indicate that urbanization will intensify further, with independent UN demographic assessments estimating that around 68% of the global population will reside in urban areas by the middle of the century. These trends have profound implications for sustainable development, infrastructure planning, and socioeconomic policy, as the expansion of urban agglomerations, along with the continued evolution of towns and peri-urban zones, reshapes the spatial distribution of global populations and heightens the need for integrated, equitable planning across all settlement categories (United-Nations, 2025).

Urban energy planning involves one main challenge for cities and governments, referring to creating sustainable and wellbeing environments in terms of social, environmental, and economic aspects. In addition, cities around the world are focusing on raising living standards by formulating their own sustainability goals. In this area, they identify energy generation and consumption as one of their main challenges. Energy planning aims to integrate and align objectives and policies in a top-down approach, which considers the realization of the government’s energy and climate objectives and activities. However, it is well known that to achieve these objectives, actions must follow a bottom-up approach, which includes the participation of different levels of government with the idea that actions are planned from the municipal and local levels (Stern et al., 2014). In addition, a sustainable energy transition requires the participation of multiple actors involved in the energy and social sectors.

Energy Transition usually makes use of different tools and methods such as scenario analysis, system optimization models, power system analysis, simulation, etc. However, due to the lack of transparency regarding data availability, experiment documentation, and the abuse of scenario-based approaches without a minimum of formal requirements when used (Asendorpf et al., 2013) generates a crisis with respect to reproducibility and replicability in scientific experiments. This concern is not specific to energy research but is increasingly documented within it: recent work argues that energy system models are too often applied and communicated as projection machines, when their evidentiary value lies instead in their use as diagnostic tools that expose assumptions, explore uncertainty, and support pluralistic, well-documented decision-making (Elder et al., 2026). Complementary quality-assessment methodologies, such as NUSAP and sensitivity auditing, have likewise been proposed to make explicit where model-based energy-policy conclusions carry more certainty than the underlying evidence can support (Piano et al., 2021). Thus, reproducibility refers to the ability to reproduce earlier results using published reports and data made available by the authors of the experiment. Replicability refers to the ability to obtain similar results using newly acquired data.

The goal of this work is to improve the approach to scenario-based modeling and data characterization to lay the foundation for transparency and traceability in energy system modeling while also pursuing robustness of reproducibility and replicability. The proposed knowledge management module formally and standardizes the characterization of:

• a system representation;
• a formal scenario approach; and
• data used in different use cases.

The module aims to integrate the data and analytical models (i.e., energy system modeling and planning tools) used to analyze energy strategies for the electricity grid. In addition, the framework considers the management of uncertainty and assumptions about the system and data, resulting in a documented, consistent, traceable, and formal script. Studying the future begins with accepting its uncertainties and converting them into sources of new ideas and opportunities.

1.1 Scope of the proposed framework

Energy system modeling spans a wide range of applications, including power system operation, long-term energy planning, optimization, forecasting, demand response, multi-energy systems, and sector coupling; this work does not attempt to formalize all of them at once. The scope of the Scenario Management Ontology, as applied together with the Common Energy System Domain Model (CESDM, Section 4), targets the scenario-based analyses most closely tied to physical network planning and operation: system planning and design, operational and security assessment (including optimal dispatch and power-flow profiles), and techno-economic analysis of markets, following the practice of the Research Center for Energy Networks. The illustrative case study (Section 3; Section 5) is built at the information-content level of the European (ENTSO-E) TYNDP scenarios and a Swiss regional application. Use cases such as demand-response market design or building-level behavioral modeling are not the primary target of this first application; rather, the ontology is designed to be paired with additional domain ontologies as such needs arise, so that the scenario layer does not need to be redefined every time the underlying energy-system domain model is extended.

The document is structured as follows: Section 2 shows a review of the literature related to strategic energy planning tasks, the main features of the technologies used in this work, namely, knowledge management and digital frameworks, and the main ideas on the scenario approach. Then, Section 2.4 presents the Data, System, and Scenario approach with the use case of an energy-planning tool. Next, Section 2.5 presents the methodology and development process of the scenario management ontology. Section 3 describes the use case. Section 4 presents the CESDM framework. In addition, Section 5 shows an example of an application with a presentation of the use case. Finally, Section 6 presents conclusions and observations.

2 Materials and methods

2.1 Review of the literature

This section presents important areas that lay the foundation for the scenario characterization approach proposed to integrate the energy plan strategy.

2.1.1 Energy system modeling

Energy system models guide strategic decision-making when planning the energy transition. Typically, models focus on illustrating different strategies to meet future demands, environmental goals, and economic investment (Heuberger et al., 2017). In addition, energy system models are often integrated with simulations, increasing analytical capabilities. Yliruka (Yliruka et al., 2022) concludes that energy system models explore different pathways to achieve environmental objectives, but most studies do not account for uncertainties in input parameters, and the above mentioned temporal/spatial aggregation has been shown to affect the optimal solution of energy system models. Then a more important question arises concerning the key parameters that model the system, which points to the number and precision of the assumptions made to coupling the uncertainty linked to the future behavior of the parameters (Gressel, 2005).

Energy system models are computational frameworks designed to represent the complex interactions of energy supply, conversion, distribution, and demand across multiple temporal and spatial scales. Over the last decade, these models have evolved from simple cost-minimization tools into sophisticated systems that incorporate multi-energy vectors, sector coupling, and uncertainty handling. This evolution supports the transition to low-carbon and resilient energy infrastructures (Fodstad et al., 2022). Modern energy system modeling captures the interdependence of electricity, heat, gas, transportation, and storage. This facilitates strategic decision-making for decarbonization pathways and just transitions in policy contexts. Contemporary research identifies four key challenges in modeling: temporal and spatial resolution, integration of multiple energy carriers, handling uncertainty, and representing energy behavior and transition dynamics. These challenges have influenced the development of model formulations and solution techniques in recent studies (Lonergan et al., 2023).

In the context of energy system modeling, sector coupling has emerged as a significant development, with the integration of multi-energy carriers and sector coupling being a prominent feature. This integration facilitates the simultaneous representation of electricity, heat, gas, fuels, and other vectors within unified frameworks. This integration is imperative for the formulation of decarbonized futures, in which electrification, power-to-heat, power-to-gas, and hydrogen pathways interact across various sectors. Sector-coupled models provide a more comprehensive understanding of inter-sectoral synergies and trade-offs, enabling planners to assess how changes in one sector (e.g., heating or transportation) affect the others. Recent studies have demonstrated the efficacy of these models in evaluating flexibility needs, infrastructure investments, and operational strategies in various sectors. These models have been shown to enhance cost-efficiency and emissions reduction, while also capturing interdependencies that are often overlooked by single-sector models.

Another central research Frontier pertains to uncertainty modeling within energy system models, driven by high shares of variable renewable energy (VRE) and long-term planning horizons. Contrary to earlier deterministic models, contemporary frameworks have begun to incorporate stochastic, robust, and scenario-based techniques. These techniques are designed to enhance the capture of variability in renewable output, demand response, technology costs, and policy trajectories. This shift in perspective is indicative of an emerging recognition that deep uncertainty regarding future conditions can exert a substantial influence on investment decisions and operational reliability. Consequently, contemporary models have been developed to simulate a broad spectrum of plausible future scenarios, frequently employing probabilistic time series or scenario ensembles as a means to do so. Nevertheless this progress, there are still significant gaps in quantifying the deep uncertainties associated with emerging technologies and socio-economic transitions.

Contemporary advances in energy system modeling are centered on the investigation of flexibility. Flexibility is defined as the ability of an energy system to adapt to temporal mismatches between supply and demand, an ability that sector-coupling technologies such as power-to-heat, power-to-gas, and vehicle-to-grid charging are increasingly designed to provide as variable renewable generation grows Ramsebner et al. (2021). This concept is now embedded in many models, driven by the increase in VRE integration and decentralized generation. A preponderance of studies indicates that storage technologies, demand response mechanisms, and sector coupling are the flexibility options most frequently modeled. In contrast, the representation of the flexibility of the transport sector (e.g., electric vehicles as mobile storage) and management of the industrial demand-side has historically been more limited, although recent process-level modeling of industrial demand-side management in the Swiss energy system shows that this flexibility can now be quantified explicitly, reporting total system cost reductions of up to 4.4% from industrial demand-side management alone Mayer et al. (2024). Robustness analyzes assess the performance of systems under variable conditions, highlighting the need to transition from cost-optimal designs to solutions that maintain performance under stress. This trend signifies a higher level of engagement with metrics that extend beyond cost and emissions, encompassing reliability, adaptability, and user behavior (Kachirayil et al., 2022).

Finally, the integration of human behavior and socio-technical transition dynamics represents a critical yet emerging area within energy system modeling. Contemporary behavioral factors, which include technology adoption rates, consumer preferences, and policy acceptance, have considerable influence on system outcomes. The objective is to elucidate human behavior and socio-technical dimensions to inform more nuanced understanding and informed policy design. Recent reviews have emphasized the dearth of interdisciplinary efforts to comprehensively integrate behavioral research with technical system models Huckebrink and Bertsch (2021), and more recent assessments confirm that this gap remains only partially closed: a systematic examination of how energy system models have represented technical, economic, and social aspects finds that socio-technical parameters are still addressed unevenly across the field Liegl et al. (2023), while a dedicated evaluation of consumer-adoption modeling shows that most models continue to rely on homogenized financial assumptions and under-represent the psychological, household, and contextual factors known to shape technology-adoption decisions Ball-Burack et al. (2025). It is imperative to address the existing gap by enhancing the realism of projections related to technology uptake, demand elasticity, and policy effectiveness. This enhancement is expected to ensure that models more closely align with real-world socio-technical transitions.

In summary, these advancements signify a persistent transition from static, single-objective optimization to dynamic, multi-objective frameworks that address the technical, economic, and social dimensions essential for the planning of future low-carbon energy systems.

2.1.1.1 Energy power systems

Power energy systems are defined as the electrical grid and its associated subsystems, including generation, transmission, distribution, and novel trends in decentralized mini-grids and micro-grids, which are based on distributed energy resources (DERs). In the last decade, the integration of renewable energy sources, the development of storage technologies, and the advancement of smart grid technologies have led to substantial challenges and opportunities in the analysis, control, and planning of power systems. The following functionalities are considered essential to the operations of energy power systems: stability and operational control in low-inertia systems, resilience and extreme event preparedness, and optimal power flow and advanced optimization techniques.

In the initial stage, the primary challenge pertains to the reduction of system inertia, as it relates to stability and operational control in low-inertia systems. The displacement of synchronous machines in favor of inverter-based renewable technologies poses a series of critical challenges with regard to maintaining stable levels of frequency and voltage. This issue is being addressed through improved control strategies, including predictive and adaptive controllers, grid-forming inverter designs, fast-acting storage, and advanced power electronics. These methodologies facilitate more responsive adjustments to disturbances and rapid changes in generation or load, thus enhancing the resilience of the system to transient events. These solutions go beyond conventional static stability criteria, incorporating real-time and predictive control mechanisms to ensure secure operation of low-inertia systems that are predominantly characterized by power electronics (Ilyushin et al., 2023).

Subsequently, a pivotal component previously mentioned pertains to the Optimal Power Flow (OPF), a foundational principle in the operation and planning of power systems. However, the classical deterministic OPF is evolving to manage uncertainty, renewable variability, and multi-objective goals. Contemporary research surveys indicate that modern OPF methods incorporate stochastic programming, robust optimization, and data-driven techniques to balance economic dispatch, voltage stability, and reserve adequacy under deep uncertainty (Chen, 2024). Advances in this field have also explored the integration of energy storage and flexible resources into OPF formulations, enabling enhanced operational performance in systems with high shares of variable renewable energy (VRE). These innovations are indicative of a broader trend toward the development of optimization frameworks capable of co-optimizing cost, reliability, and emissions in highly renewable grids.

It is imperative to acknowledge the paramount importance of the resilience of the power system in the context of climate change and the increasing prevalence of extreme weather events. These phenomena have led to a notable escalation in the frequency of high-impact, low-probability disruptions, underscoring the critical need for robust resilience measures in power systems. The concept of resilience encompasses more than mere reliability; it encompasses the system’s capacity to anticipate, withstand, and fast recover from disruptions such as hurricanes, wildfires, or cyberattacks (Ghanbari and Jiang, 2025). Recent research describes resilience metrics that differ from reliability measures, tracing their evolution across topological, probabilistic, and time-dependent formulations developed to capture preparedness, absorption, and recovery separately (Younesi et al., 2022), and investigates strategies such as distributed generation, microgrid formation, grid reinforcement, and dynamic resource reconfiguration to strengthen the robustness of the system. These strategies are designed to ensure the continuity of essential services, reduce energy consumption, and facilitate rapid restoration.

2.1.1.2 Energy planning

Energy planning or municipal government energy conservation planning (municipal facilities and operations) is used to develop energy goals and evaluate options for current energy use, identify energy problems, and assess the potential for energy efficiency and conservation at the municipal level (Dane, 2010). Energy planning uses energy system modeling and analytical tools to create scenarios that show the implications and possibilities for improved decision-making. The main objective of municipal energy planning is to define strategies for the energy transition, defining plausible scenarios and achievable goals (Johannsen et al., 2021). However, energy-planning systems at the municipal level involve strategies and objectives at the local level that often lack the appropriate tools and resources for the necessary analysis. Dane (2010) established a series of activities that comprise the energy planning development process:

• Definition of the current state of energy use and operations
• Definition of the main leader and the leading group for energy planning tasks.
• Set the timeline and the final goal.
• Involvement of key stakeholders and real commitment.
• Demonstrate the importance and relevance of energy planning for the municipality.
• Definition of energy plan scope and goal SMART (specific-measurable-ambitious-realistic-timed)
• Integrate and tie energy planning with other municipal plans, such as financial, strategic, or comprehensive.
• Development of the action plan to guide decision-making.

There is a clear need to obtain data to understand the historical behavior of energy use. Thus, having an accurate history of energy use is the first step in developing an energy use and cost management plan. This helps to establish and set energy targets and identify opportunities to define scenarios that include key characteristics that suggest trend lines. When developing an energy plan, municipalities need to decide on aspects to consider and characteristics to be monitored, as well as identify opportunities to address future scenarios that arise. The main energy issues are costs, availability, and reliability. First, cost often has a negative impact on municipal budgets, and the energy plan should define how to react to cost fluctuations. The second issue concerns the availability of energy resources and the development of scenarios for future outages. Finally, reliability comprises the power supply, reducing energy use and reducing peak demand.

2.2 Knowledge management

Semantic models, also known as knowledge models, are formally employed for the representation of real-world systems (Uschold and Gruninger, 2004). Ontology research has evolved from a domain dominated by expert-centric approaches to one that integrates human expertise with advanced computational methods, such as machine learning and automated reasoning. This paradigm shift signifies a significant advancement in ontology research, moving it from the realm of pure handcrafted, expert-centric approaches to a hybrid engineering strategy that combines human expertise with advanced computational methods. Ontologies are formal structures that facilitate the acquisition, maintenance, access, distribution, and reuse of data, information, and knowledge. They also enable its interpretation and dissemination through machine processes.

A predominant theme in contemporary ontology research pertains to methodologies and tools of ontology engineering. These methodologies and tools address fundamental challenges related to the accuracy, consistency, interoperability, and scalability of knowledge models. Systematic analysis of tooling practices indicates that ontology development remains a highly knowledge-intensive process that requires domain experts to specify concepts and relationships. In addition, supporting software tools have become increasingly essential to reduce manual burden and improve quality assurance across the engineering lifecycle (e.g., modeling, validation, documentation) of ontologies.

In addition to tools and learning methods, ongoing work examines semantic interoperability and alignment, emphasizing how ontology models facilitate integration across heterogeneous information systems. Such work frequently involves the integration of foundational ontologies and pattern-driven modeling methodologies to achieve harmonization of domain-specific vocabularies and enhancement of the robustness of ontology-driven systems. Recent studies further indicate that incorporation of semantic context in ontology models enhances their applicability in domains that require high degrees of semantic consistency (Stănescu and Oprea, 2025).

Knowledge management systems benefit from ontologies that semantically enrich information and precisely define the meaning of various information artifacts. Ontologies have been developed to facilitate the sharing and reuse of knowledge across disciplines and research communities, such as advances in medicine, knowledge engineering, natural language processing, cooperative information systems, information integration, and software agents, among others. In general, we can state that ontologies provide a common and shared understanding of a specific domain and context, an approach based on human- and machine-readable knowledge, and an explicit conceptualization that describes the meaning of the data.

In the domain of energy systems, which is characterized by heterogeneous data, multidisciplinary models, and complex interfaces between stakeholders, ontological models are used as semantic infrastructures that support data integration, model coupling, and reproducible analysis in pursuit of interoperability. A significant development in this area is the creation of domain-specific ontologies, such as the Open Energy Ontology (OEO), which facilitates a shared vocabulary and conceptual framework for energy systems analysis. The OEO’s primary function is to establish a unified representation of energy concepts that are evident across various models, scenarios, and datasets. This facilitates semantic annotation, data exchange, and interoperability of computational models Booshehri et al. (2021). Integration of ontologies with contemporary AI technologies facilitates the development of scalable, explainable, and maintainable knowledge systems, thus supporting advanced analytics, decision support, and system automation across complex and data-intensive domains.

Ontology engineering and knowledge-graph construction are, in fact, challenges shared well beyond the energy domain, and recent methodological work from adjacent fields is directly relevant to the approach followed here. The LOT framework structures ontology development around requirements specification, implementation, and publication in a way intended for industrial rather than purely academic adoption (Poveda-Villalón et al., 2022), while systematic reviews of knowledge-graph construction from semi-structured data (Ryen et al., 2022) and of the knowledge-graph development process more broadly Tamašauskaitė and Groth (2023) identify recurring bottlenecks in data integration, schema alignment, and quality control that are largely independent of the application domain; a complementary survey addresses how knowledge-graph quality should be assessed and maintained once a graph is in production use Whaley (2008). Tooling that lowers the barrier between conceptual modeling and formal implementation, such as the translation of UML-based ontology diagrams into OWL Chávez-Feria et al. (2022), reflects the same concern that motivates the Protégé-based, glossary-first methodology followed in Section 2.5: an ontology is only as useful as the ease with which domain experts, rather than only ontology engineers, can inspect, validate, and extend it.

Within the energy domain specifically, the Open Energy Ontology (OEO) is not the only recent attempt to formalize scenario- or simulation-related knowledge. A co-simulation scenario ontology has been proposed to describe the configuration of coupled energy-system simulations, including the models, interfaces, and data exchanges involved in a given simulation run Schwarz et al. (2025), Schwarz et al. (2026). At the building scale, an ontology-based modeling framework has been field-demonstrated to support interoperable digital twins across a portfolio of buildings Bjørnskov et al. (2025). A recent systematic review situates these and related efforts within the broader landscape of ontologies developed for local electricity markets and distributed energy resources Costa et al. (2026). Table 1 positions the Scenario Management Ontology proposed in this work relative to these three efforts, along the dimensions of primary purpose, support for scenario lineage and versioning, provenance and traceability, the semantic web standard used, and the form of domain coupling adopted. Unlike OEO, which formalizes the energy-system domain itself, and the co-simulation and digital-twin ontologies, which formalize a simulation or asset configuration, the Scenario Management Ontology formalizes the scenario as a first-class, versioned artifact that references a domain ontology (CESDM, Section 4) rather than embedding domain concepts directly. This is what allows a baseline scenario, its sub-scenarios, and their full derivation chain to be queried and audited independently of whichever domain ontology or analytical tool is attached to them, as demonstrated for the TYNDP case in Section 5.6.

TABLE 1. Comparison of the Scenario Management Ontology with related ontology-based frameworks for energy-system scenarios and simulations.

2.3 Scenario approach

Scenarios have been used to support strategic planning with a historical precedence for 5 decades (Whaley, 2008). The use of scenarios as a tool can be traced back to the early philosophers, including Plato, however, scenario as a tool, was used intensively for contingency planning during World War II (Gu et al., 2023). The use of strategic planning scenarios emerged during the 1973 oil price crisis, when it gained prominence and was recognized as a key tool (Mercer, 1995). The application of scenarios to activities within companies was proposed by the applications made by Royal Dutch Shell and General Electric, which have been followed by a wider use of this tool to support decision-making of scientific, social, and political interests (Bradfield et al., 2005). The subsequent section will present a series of definitions of the scenario:

• Scenarios are a set of hypothetical events set in the future constructed to clarify a possible chain of causal events as well as their decision points (Herman and Wiener, 1967).
• Scenarios are descriptions of a future situation and the course of events that allows one to move forward (Godet, 2000).
• Scenarios are based on the assumption that the business world is unpredictable but certain events are predetermined. No scenario can provide an accurate description of the future. Their function is to help decision-makers understand their task (Roubelat, 2000).
• Scenarios are part of strategic planning as tools for managing future uncertainties (Ringland, 2004).
• Scenarios generate a series of alternative futures rather than a naive extrapolation of a single historical datum and constitute scenarios when combined with judgments and narratives about events in these futures (Bradfield et al., 2005).
• Scenarios are hypothetical sequences of events constructed to focus attention on causal processes and decision-making (Burt, 2007).
• Scenarios are structurally different stories about how the future might unfold to deal with uncertainty (Gu et al., 2023).

The fundamental aim of a future scenario is to treat uncertainty. Scenarios are structurally different stories about how the future might develop that are believed to impact the field/business at focus (Rialland and Wold, 2009). Cordova and Rouwette (Cordova-Pozo and Rouwette, 2023) found that six components are the most commonly used elements in the definition of scenarios. The components comprise:

• a scenario is future-oriented,
• a scenario refers to external contexts,
• a scenario has a narrative form,
• a scenario is plausible,
• a scenario is part of a systematized set, and
• a scenario poses significantly different alternatives.

Over time, various scenario techniques applied to business planning have emerged, the United States and France being the main pioneering countries (Bradfield et al., 2005). Martelli (1993) concludes that scenario-based planning lacks several methodological shortcomings despite years of use in various organizations. One is the lack of a formal and accepted definition of scenarios (Godet, 2000; Chermack, 2004; Bradfield et al., 2005; Varum and Melo, 2010; Spaniol and Rowland, 2019). The second refers to using different theories, principles, and practical rules. Many different scenario approaches use their own techniques (Amer et al., 2013; Bishop et al., 2007). On the one hand, as an approach, the scenario lacks (Martelli, 1993). On the other hand, there is an ongoing debate in the literature on how to classify scenarios. Although some authors contrast qualitative and quantitative techniques to build scenarios, others prefer to distinguish forward-looking (qualitative) from predictive (quantitative) techniques, combining them with other more complex techniques such as artificial intelligence, such as the use of the Delphi three-round method to support scenario analysis (Bessa and Barbosa, 2025).

A popular classification of scenario methods divides the field into three schools: the intuitive logic method (ILM) using qualitative techniques, prospective thinking (La Prospective or LP) that contemplates the use of qualitative techniques and quantitative analysis, and modified probabilistic trends (PMT) from quantitative data (Chermack, 2004; Varum and Melo, 2010; Amer et al., 2013). Finally, the third shortcoming is the lack of evidence for the claimed effects of the scenarios. Each school employs a set of techniques, some of which have not been adequately tested and lack practical value or theoretical basis (Varum and Melo, 2010).

2.3.1 Heterogeneity across public scenario sources

The methodological shortcomings described above are compounded, in practice, by the fact that the organizations that publish the scenarios most widely used for energy and climate analysis do not describe them in a common way. Table 2 compares four major, actively maintained scenario sources with different institutional purposes and modeling traditions: the International Energy Agency’s World Energy Outlook International Energy Agency (2025), the International Renewable Energy Agency’s World Energy Transitions Outlook International Renewable Energy Agency (2024), the Intergovernmental Panel on Climate Change’s Sixth Assessment Report mitigation pathways Intergovernmental Panel on Climate Change (2022), and the Network for Greening the Financial System’s Climate Scenarios Network for Greening the Financial System (2023). These sources differ in at least four ways that are relevant to ontology design. Content heterogeneity concerns what a scenario reports as content: an IEA scenario emphasizes energy balances, technology deployment, and investment, whereas an NGFS scenario additionally represents GDP, asset risk, and transition-related financial losses, and IPCC pathways extend beyond the energy system to land use and non-CO 2 gases. Presentation heterogeneity concerns how that information is published, ranging from narrative text and static charts to structured, versioned databases. Detail heterogeneity concerns spatial and temporal resolution, from global and multi-decadal to national and hourly. Structural heterogeneity concerns how scenarios are constructed in the first place, whether as single exploratory narratives, normative target-driven pathways, or probabilistic ensembles.

TABLE 2. Comparison of major public scenario sources and associated heterogeneity dimensions.

This heterogeneity is precisely what motivates representing a scenario as a structured object rather than as a document. The dimensions in Table 2 map onto the classes and data properties already introduced in Section 2.5 and detailed in Tables Tables 3–8: the driving assumptions behind a narrative correspond to the assumption data property and to ScenarioDrivenFeature instances; the technologies, parameters, and constraints that content heterogeneity refers to correspond to ModelKeyFeature and the domain entities defined through CESDM (Section 4); detail heterogeneity is captured explicitly through the spatial_resolution and temporal_resolution data properties (Tables 7, 8); and structural heterogeneity, the distinction between an exploratory narrative, a normative pathway, and a derived variant, is what the baseline-scenario, reference-scenario, and sub-scenario triad (Section 2.4) is designed to express formally. In this sense, the ontology does not need a separate class for every possible scenario source; it needs the small set of concepts already formalized in Tables Tables 5–8 to be expressive enough to accommodate them, which Section 5 demonstrates for the TYNDP case.

TABLE 3. Scenario management ontology glossary of terms (Part 1/2).

TABLE 4. Scenario management ontology glossary of terms (Part 2/2).

TABLE 5. Classes defined in the Scenario Management Ontology, with superclass and definition.

TABLE 6. Object properties defined in the Scenario Management Ontology, with domain, range, and axioms.

TABLE 7. Data properties defined in the ScenarioManagement ontology, with domain, range, units and annotations (Part 1/2).

TABLE 8. Data properties defined in the ScenarioManagement ontology, with domain, range, units and annotations (Part 2/2).

2.4 Scenario knowledge-based approach

In this work, a scenario is defined as a strategic tool that creates narratives based on a formal representation of the system with explicit assumptions about the conception of the system. The scenario defines future assumptions and uncertainties to create short-, mid-, and long-term sub-scenarios. Furthermore, a scenario must be able to track data sources, quantify uncertainty, and define the probability of occurrence of certain assumptions. This approach considers the following three different types of scenario in relation to their temporal situation: prior-scenario, baseline-scenario, and future-scenario, as described in the following.

• Baseline-Scenario:. A baseline scenario represents the complete study system that includes all entities, attributes, and relationships. It is self-contained and independent, serving as the root of a scenario family tree. It includes assumptions and data sources used to define the system’s behavior. A baseline scenario is used as an initial system representation in analysis activities and is the starting point for scenario variations to create multiple sub-scenarios. In addition, the scenario describes the analysis to be performed and the solution method applied. Together, all of this provides a clear idea of the context of the goals and objectives to create future scenarios.
• Scenario Reference:. A reference scenario can be either a baseline or a sub-scenario that serves as the starting point for creating a new sub-scenario. This enables scenario chaining.
• Sub-Scenario:. A sub-scenario derives from a reference scenario and contains only the modifications (deltas), such as, differences: additions, modifications, deletions, relative to that reference scenario, placing assumptions that support those modifications. Changes may be related to factors such as policy changes, technological developments, sensitivity analyses, or simply hypothetical scenarios. The assumptions must formally describe the sources that support them. Finally, the probability of occurrence places a weight on the degree of uncertainty of the assumption. It references a parent scenario that can be a baseline or another sub-scenario. In this way a sub-scenario maintains traceability to baseline scenario, meaning that we always can find the main source of the system modified. One of the main objectives of the sub-scenario is to have a more compact and easier to review asset.

This work aims at transparency and a clear definition of the elements required of a scenario, as shown in Figure 1. The defined elements are described below.

• Scenario: The scenario generates narratives based on an explicit and formal representation of the system. The scenario presents a description to have a general idea about the scenario and main assumptions. The scenario contains metadata to explicit information like id, name, author, institution, time horizon, creation date, date of update, context, tool used, coverage of the physical-region, sector covered, covered technology, key parameters, driven parameters, and assumptions. Formality is achieved by using domain ontologies to describe the system and its parts. Ontological classes with data properties represent parameters or attributes that serve as the system’s building blocks and can vary through different future-scenarios. The scenario should highlight the most significant uncertainties regarding the system’s behavior.
• Analytical Activity: This feature relates to the approach, method, or tool chosen for the analysis activity. These activities include planning, scheduling, controlling, monitoring, designing, process improvement, and any other activity necessary for decision-making. Any uncertainty or assumptions about how the chosen method drives the activity can be documented.
• Scope of analysis: A brief description of the specific part of the system that is important or taken into account in the objective analysis. This includes the definition of the temporal and physical resolution that delimits the scope of a specific analysis. Any uncertainties related to insufficient information to cover the objectives of the activity must be adequately defined, as well as any assumptions that address these uncertainties.
• System characterization: The characterization of the system defines its key elements and key performance indicators. In the proposed approach, the system’s characterization is based on domain ontologies, which provide a formal semantic model. There is uncertainty regarding the representation of the system based on knowledge related to specific activities. In addition, a robust system characterization can be used for various analytical activities.
• Key features: These are the parameters and variables necessary to represent the system and perform the analytical activity for which the scenario was designed. Any uncertainties or assumptions related to the behavior of each parameter or variable must be defined. If applicable, the data pre-processing performed to obtain a specific parameter should also be defined.
• Data: The lower level refers to data representing the values of variables and parameters, as well as their respective sources of information and, potentially, calculation methods. Ideally, uncertainties, assumptions, calculation methods, data sources, and references to the theorems that define them should be specified.

In terms of traceability, the scenario management approach is data-centric. It considers a set of artifacts that capture information related to the data representing the different elements of the system from which the scenarios will be created. Figure 2 shows the system characterization templates, input data, data processing (ELT), and output data. This work proposes a bottom-up approach with respect to the way scenarios are defined. Thus, Figure 3 shows the proposed procedure to manage scenarios, uncertainties, and assumptions.

In addition, the development of the scenario is managed as a dynamic process through the support of semantic knowledge models, as presented in Figure 4. This means that taking a scenario as a basis, there will be an iterative process, as assumptions are modified and scenarios change based on observed behaviors. Finally, this approach aims to capture and define the assumed system, its uncertainties, and assumptions, as well as to show the unknown parts of the system for improvement.

2.4.1 Relation to optimization and operations-research concepts

Because the elements described above are also the ingredients of most energy-system optimization studies, it is useful to make the correspondence with standard operations-research terminology explicit. Parameters, the fixed inputs of an optimization model, correspond to ModelKeyFeature instances and their associated data properties, such as the constrained numerical attributes shown in Tables 7, 8 (for example, energy_conversion_efficiency bounded to [ 0,1 ], or nominal_power_capacity bounded to non-negative values in the underlying CESDM schema, Section 4). Decision variables and the alternative technology or policy choices that distinguish one plausible future from another correspond to ScenarioDrivenFeature instances, whose values are precisely what changes across a family of sub-scenarios derived from the same baseline (Section 2.4; Section 5.4). Constraints on these variables are the type and range restrictions already enforced by schema validation (Section 4), so that a candidate model input which would be an infeasible or physically meaningless choice is rejected before it ever reaches a solver. The objective function itself, and the solution method used to optimize it, are recorded through the analytical_activity_performed, analytical_activity_scope, and analytical_model_used data properties of the Scenario class (Table 7; Section 2.4), keeping the choice of optimization approach traceable but external to the ontology, consistent with the framework’s focus on structure rather than solution behavior (Section 4). Finally, uncertainty and stochastic data are handled by the dedicated Uncertainty and Risk classes and their uncertainty_measurement and risk_measurement data properties (Table 8), which allow a scenario to attach a quantified probability, rather than only a point estimate, to a given assumption. This mapping does not implement a specific stochastic or robust optimization method; rather, it gives modelers who work with such methods Kotzur et al. (2021), Bohlayer et al. (2021), Neumann and Brown (2021), Seljom et al. (2021), Frey et al. (2023), Frey et al. (2024) a shared, machine-checkable place to record the parameters, variables, constraints, and uncertainty ranges that their formulations already require, so that two studies using different solvers can still be compared at the level of what they assumed.

2.5 Scenario management ontology development

2.5.1 Methodology

The development of a Scenario Management Ontology is based on a formal ontology expressed using Semantic Web standards (RDF/OWL). The ontology defines domain energy system concepts, their properties and relationships in a machine-interpretable manner. The Scenario Management Ontology has been developed following the ontological development methodology based on the improvement cycle (Plan, Do, Check, Act, and Re-plan) presented in Muñoz Mata (2011). This methodology focuses on dedicating the appropriate amount of time to each phase and activity of ontological development, which is essential to achieve a smooth and agile quality improvement process. The elements of the cycle involve a reflective process based on the scientific method and ensure that improvement efforts are carried out based on the development and application of the model to specific use cases.

2.5.1.1 Glossary of terms

The glossary of terms identifies and defines useful and potential items for modeling a specific domain. It also provides the correct meaning of each term, clarifying the context. The glossary of terms is also used to extract relations between concepts (properties) and attributes that guide the creation of rules and axioms. The glossary of terms is presented in Tables Tables 3, 4.

2.5.1.2 Classes, object properties, and data properties

The following tables summarize the Classes shown in Table 5, Object Properties shown in Table 6, and Data Properties shown in Tables Tables 7, 8, defined in the Scenario Management Ontology (http://www.fen-ethz.ch/ontology/2026/ScenarioManagement#), together with their associated annotation properties (rdfs:definition, rdfs:comment, and the custom annotation property ScenarioManagement:units).

2.5.1.3 Model development (protègè)

The ontology was instantiated and implemented using the Protégé ontology editor, adhering to OWL DL specifications.

3 Use case

In this work, the Ten-Year Network Development Plan (TYNDP) scenarios were taken as a proof of concept to test our framework. The joint developed by the European Network of Transmission System Operators for Electricity (ENTSO-E) and Gas (ENTSOG) provides a long-term analytical foundation for European energy infrastructure planning. Rather than forecasting the future, these scenarios explore plausible pathways for the evolution of the European energy system under different technological, economic, and policy assumptions. Their primary objective is to ensure that investments in electricity, hydrogen, and gas infrastructure remain consistent with the European Union’s climate neutrality objectives, while maintaining security of supply and supporting market integration.

3.1 TYNDP 2024 scenarios

The TYNDP 2024 framework (European Network of Transmission System Operators for Electricity (ENTSO-E) and European Network of Transmission System Operators for Gas (ENTSOG, 2025a; ENTSOG, 2025b) builds on previous scenario cycles and combines bottom-up national policy assumptions with alternative long-term decarbonization pathways. The framework includes six scenarios covering various time horizons, specifically 2030, 2040, and 2050. These scenarios are grouped into two categories: “National Trends Plus” (NT+) scenarios and the deviation scenarios, namely, “Distributed Energy” (DE) and “Global Ambition” (GA) (European Network of Transmission System Operators for Electricity (ENTSO-E) and European Network of Transmission System Operators for Gas (ENTSOG, 2025c).

The National Trends Plus (NT+) scenario represents the baseline trajectory and is based primarily on Member States’ national energy and climate plans (NECPs), long-term national strategies, hydrogen strategies, and European climate targets. It reflects the expected implementation of policies under current commitments, while remaining consistent with the EU’s decarbonization goals. Consequently, NT + serves as the primary reference case for the assessment of infrastructure.

The Distributed Energy (DE) scenario examines a future characterized by significant decentralization of the energy system. It is based on the assumption of widespread deployment of distributed renewable generation, high levels of energy efficiency, greater consumer participation, local flexibility resources, and significant behavioral changes. Energy production becomes increasingly localized, reducing dependence on imported fuels while requiring extensive integration of distributed energy resources into the power grids.

In contrast, the Global Ambition (GA) scenario is based on a more centrally coordinated transition, driven by large-scale technology deployment and stronger policy intervention. This pathway relies on accelerated investment in industrial-scale renewable energy generation, hydrogen production, large-scale energy transmission projects, and cross-border energy exchanges. Compared to the DE scenario, the GA scenario emphasizes the coordinated expansion of infrastructure and greater integration of European energy markets.

Across all scenarios, TYNDP 2024 incorporates a fully integrated modeling framework linking electricity, methane, and hydrogen systems. This represents an important methodological advancement compared with earlier TYNDP cycles, enabling the assessment of interactions between different energy carriers and providing a more comprehensive evaluation of future infrastructure requirements. The framework also increases transparency through extensive stakeholder consultation, updated modeling methodologies, and alignment with the revised TEN-E Regulation.

3.2 Evolution towards the TYNDP 2026 Scenario Framework

The TYNDP 2026 Scenario Framework (European Network of Transmission System Operators for Electricity (ENTSO-E) and European Network of Transmission System Operators for Gas (ENTSOG, 2026) represents a significant evolution in scenario development at the European level following the implementation of European Union Agency for the Cooperation of Energy Regulators’ (ACER) Scenario Framework Guidelines. Rather than maintaining multiple alternative transition pathways, the new framework simplifies the structure of the scenarios while strengthening consistency with European legislation and national policy commitments. The central element of the new framework is a single “Central Scenario,” which replaces the previous NT + scenario as the primary reference for infrastructure planning. This scenario continues to be based on the updated National Energy and Climate Plans, European climate legislation, and national policy commitments, ensuring that infrastructure assessments remain aligned with officially adopted policy trajectories.

Instead of the previous “Distributed Energy” and “Global Ambition” pathways, the 2026 framework introduces two “Economic Variants”:

• High economic scenario (HEV): represents higher economic growth, increased industrial activity, and increased energy demand.
• Low economic scenario (LEV): represents slower economic development and comparatively lower energy demand.

Unlike the DE and GA scenarios, these variants do not aim to describe fundamentally different decarbonization strategies. Rather, they serve as sensitivity analyses that assess how variations in macroeconomic conditions influence future energy demand, infrastructure utilization, and investment needs, while maintaining the same underlying climate policy objectives. This shift reflects a methodological transition from comparing alternative transition narratives to assessing uncertainty within a common policy trajectory. Another major advancement in TYNDP 2026 is the continued development of integrated cross-sector modeling. Electricity, methane, and hydrogen networks are simultaneously assessed using increasingly sophisticated optimization techniques, allowing a more detailed representation of sector coupling, electrification, hydrogen deployment, renewable integration, and system flexibility. The framework also strengthens stakeholder participation and improves methodological transparency through expanded consultation processes and closer alignment with the requirements of the revised TEN-E Regulation and ACER’s guidelines.

Finally, another important advancement in the TYNDP 2026 is the continued development of integrated cross-sectoral models. The electricity, methane, and hydrogen networks are evaluated simultaneously using increasingly sophisticated optimization techniques, enabling a more detailed representation of sectoral coupling, electrification, hydrogen deployment, integration of renewable energy, and system flexibility. The framework also strengthens stakeholder engagement and improves methodological transparency through expanded consultation processes and greater alignment with the requirements of the revised TEN-E Regulation and ACER guidelines.

3.3 Comparison between TYNDP 2024 and TYNDP 2026

The transition from TYNDP 2024 to TYNDP 2026 reflects a broader evolution in the planning of European energy systems. Although the 2024 framework explored multiple transition scenarios representing different technological and social options, the 2026 framework focuses on a single policy-aligned transition pathway, supplemented by economic sensitivity analyses. This change improves consistency in infrastructure assessments and reduces the uncertainty associated with multiple conflicting transition scenarios. Furthermore, TYNDP 2026 places greater emphasis on integrated energy system planning, recognizing the growing interdependence between electricity, hydrogen, and gas networks. Greater cross-sectoral optimization, improved modeling methodologies, and greater regulatory harmonization are intended to provide a more robust analytical basis for identifying future infrastructure investments that support Europe’s goal of achieving climate neutrality by 2050, while maintaining system reliability and cost-efficiency.

4 The CESDM framework

The Common Energy System Domain Model (CESDM) is a schema-based general-purpose modeling framework, developed by the Research Center for Energy Networks group (FEN-ETHz), to be applied to the field of energy systems. Its guiding principle is that the structure of a system should be made explicit, rather than its behavior. In energy system networks, this means that an energy system is described through entities, the attributes that characterize them, and the relationships that connect them, and that the set of permissible descriptions is determined by a set of schemas. In addition, the framework is deliberately independent of the optimization or simulation logic. CESDM focuses on the components of an energy system, leaving it up to the tool that uses the model to decide how to solve it. The emphasis on structure stems from the practical difficulty in modeling complex and heterogeneous systems. Furthermore, without an explicit formal layer, energy system models tend to be ambiguous and internally inconsistent; they are difficult to follow and are not easily reusable across different tools. CESDM addresses this problem by treating a schema as a formal contract, that is, for a given entity type, it establishes what is permitted, what is required, and what is excluded, so that a model validated according to its schema is, by definition, structurally correct.

Every system modeled in CESDM can be broken down into three basic elements:

• Entity: denotes an object in the system, such as a region, an asset, or a transporter.
• Attribute: denotes an intrinsic property of an entity, such as installed capacity or conversion efficiency.
• Relationship: denotes an explicit, typed link between two entities, such as “has a location” or “has a conversion.”

Schemas formalize these basic elements for each entity type: they specify which attributes an instance can or must have, along with their data types and constraints (for example, an efficiency value restricted to the unit interval), and which relationships an instance can or must establish, along with the entity type with which each relationship is permitted to be associated. Another structural feature is schema inheritance: a derived entity type automatically inherits the attributes, relationships, and constraints of its parent entity, so that the shared structure is defined only once and progressively specialized, rather than duplicated. Conversion technologies illustrate this pattern well. A base type EnergyTechnology contains common metadata; EnergyConversionTechnology limits it to conversion processes in general; and EnergyConversion Technology1x1 further restricts the type to those technologies that convert exactly one input vector into exactly one output vector and feed the result into a single node in the network. In the illustrative example of Switzerland (CH) and its neighboring countries used in the framework documentation, this entity type encompasses electricity generation from natural gas, run-of-river hydropower (water to electricity), and nuclear power generation (uranium to electricity).

Since schemas can be verified by a machine, a model can be validated automatically rather than manually inspected. Attribute validation confirms that required attributes are present, have the correct type, fall within any permitted numerical range and, where applicable, are drawn from a permitted set of enumerated values. Relationship validation confirms that a relationship is defined in the schema, that its target exists in the model, and that the target is of the entity type specified by the schema. Constraints inherited from parent schemas are applied along with the rest. An efficiency value greater than one, a conversion technology with no declared input carrier, or a technology that runs out of output nodes is rejected because it violates an explicit structural rule.

In the energy domain, CESDM provides a compact set of entity types sufficient to collectively describe electricity, gas, heat, and multi-energy systems. An EnergySystemModel acts as a root container in which all other entities are instantiated. An EnergyDomain groups components that operate within a common physical medium, while an EnergyCarrier represents that medium itself, distinguishing between domain, fuel, and resource carriers, and including attributes such as cost and emissions intensity. A GeographicalRegion and an EnergyNode (or GridNode) locate the components spatially and electrically; the latter records properties such as nominal voltage, along with its domain and region. An EnergyDemand associates exogenous consumption with a node. Conversion is primarily represented by Energy Conversion Technology1x1, described above, and by CombinedHeatAndPowerPlant for the combined generation of electricity and heat from a single fuel input, which tracks electrical and thermal efficiencies and capacities separately and maintains distinct connection points for fuel, electricity, and heat. EnergyStorageTechnology represents charging and discharging behavior and the temporal shift of energy, with separate capacities and efficiencies for each direction and an initial state of charge. Finally, NetTransferCapacity limits the power flow between two nodes in each direction, representing the transmission limits within the network. Finally, the same conceptual division applies to all these types of entities. Schemas encode domain knowledge and impose constraints; models translate that knowledge into scenario-specific data and are validated based on that data. This separation allows CESDM schemas to remain stable and be shared among different modeling tools, while the models built from them vary freely from one study to another.

5 Results

5.1 From TYNDP storylines to a machine-actionable scenario framework

ENTSO-E’s Ten-Year Network Development Plan (TYNDP) publishes its future outlooks as three narrative storylines, such as National Trends, Global Ambition, and Distributed Energy, each unfolding across several horizon years under a fixed weather year. As released, these storylines are descriptive: tables and narrative text rather than a validated, queryable dataset. The proposed scenario manager described here takes on the job of turning that narrative material into a set of CESDM-conformant scenario instances, connected to one another through an explicit lineage rather than left as a set of disconnected spreadsheets. The master_structure.yaml is the index that records the result of that transformation: one baseline scenario and seven sub-scenarios, each pointing to its own file and each declaring, through a reference_scenario_id and a baseline_scenario_id, where it sits in the overall structure.

5.2 A single baseline, three diverging storylines

Only one scenario in the index is a baseline: SC_NT_SY_2030_WY_2009, the 2030 horizon of the National Trends storyline. Every other scenario carries a baseline_scenario_id pointing back to this one node, so that no matter how far a storyline has evolved, its ultimate structural ancestor remains traceable. What changes from scenario to scenario is the reference_scenario_id: the immediate predecessor against which a given file’s content is expressed. Reading these two fields together across all eight files reproduces the storyline structure directly, without needing to consult anything outside the index itself:

• SC_NT_SY_2030_WY_2009 (baseline,National Trends 2030 )
• |
• +-- SC_NT_SY_2040_WY_2009 (National Trends 2040 )
• |
• +-- SC_GA_SY_2030_WY_2009 (Global Ambition 2030 )
• | +-- SC_GA_SY_2040_WY_2009 (Global Ambition 2040 )
• | +-- SC_GA_SY_2050_WY_2009 (Global Ambition 2050 )
• |
• +-- SC_DE_SY_2030_WY_2009 (Distributed Energy 2030 )
• +-- SC_DE_SY_2040_WY_2009 (Distributed Energy 2040 )
• +-- SC_DE_SY_2050_WY_2009 (Distributed Energy 2050 )

National Trends 2040 is a single step away from the baseline. Global Ambition and Distributed Energy each branch off the baseline at their 2030 horizon and then proceed as a chain, with each later horizon referencing the one before it rather than the baseline directly. baseline_scenario_id is therefore a constant anchor to the root of the tree, while reference_scenario_id encodes the order in which the underlying content actually has to be resolved.

5.3 The baseline as a complete reference snapshot

The baseline file, baseline_SC_NT_SY_2030_WY_2009.yaml, is the only scenario in the set that is fully self-contained: it instantiates every entity type the energy-domain schema defines, populated to the scale needed to represent a pan-European system. Next, Table 9 summarizes this population.

TABLE 9. Entity population of the National Trends 2030 baseline scenario.

Every entity in this file carries the attributes and relations its schema requires (a conversion technology, for instance, declares its nominal power_capacity and energy_conversion_efficiency alongside hasInputEnergyCarrier, hasOutputEnergyCarrier and isOutputNodeOf relations), so the file is directly validated against the schema on its own, independent of any other scenario. This is what qualifies it to serve as the baseline: everything downstream can be expressed as a difference from it, because it is the one place where nothing is left implicit.

5.4 Sub-scenarios: transformation expressed only as change

None of the seven remaining scenarios repeats this full population. Each is stored as a change-set against its reference scenario, split into three parts: added_components, modified_components, and deleted_components, with entities grouped under each part by entity type. This is the sense in which a sub-scenario is expressed “only with modifications”: it says what is new, what has changed, and what has been dropped, and leaves everything else to be inherited unchanged from the reference.

For example, the SC_DE_SY_2030_WY_2009. yaml, the Distributed Energy 2030 horizon, referencing the baseline directly, and can illustrate what that transformation looks like in practice. Table 10 gives the size of each change-set by entity type.

TABLE 10. Change-set applied by the Distributed Energy 2030 sub-scenario relative to the National Trends 2030 baseline.

He additions are consistent with what the storyline’s own assumptions describe: new EnergyDemand entities for electric-vehicle and prosumer load and for electrolysis demand, together with new conversion and storage technologies that did not exist in the baseline mix. The modifications are dominated by attribute-level updates, an annual_energy_demand revised upward for electrified load, a capacity or efficiency revised for a technology that already existed, recorded in the file as the new value of the affected attribute, without restating the rest of the entity. The deletions are concentrated in older, centralized thermal technologies (aging oil, lignite, hard-coal, and gas units), which is the expected counterpart of a storyline built around distributed, prosumer-centred generation rather than centralized plant.

Two things follow from this:

• First, the sub-scenario file is markedly smaller than the baseline it modifies, roughly half its length, because it stores a difference, not a duplicate.
• Second, that compactness does not mean the change is small: of the baseline’s 598 conversion technologies, 515 (86%) are either modified in place or removed outright, and 31 further technologies are introduced as new entities on top of that. That is a concrete way of showing that “Distributed Energy” is a structural reshaping of the generation mix rather than a minor parameter sweep, even though it is represented economically as a patch.

Because reference_scenario_id can point to a sub-scenario rather than the baseline, resolving a scenario such as SC_DE_SY_2050_WY_2009 is a recursive operation: the set of changes for 2050 is applied to the already resolved 2040 state, which in turn is the set of changes for 2030 applied to the baseline. This makes it possible to maintain three independent narrative threads, each spanning up to three time horizons, for a total of eight files, rather than repeated complete snapshots of a system with more than a thousand entities.

5.5 One validated model, three ready-to-use formats

Because each entity in a resolved scenario already has a set of typed and constrained attributes, as well as a set of typed relationships, the same underlying model can be transferred to an analysis tool in more than one physical format without having to be remodeled for each one.

• YAML format: The model is exactly as shown above: a human-readable, version-controlled serialization, which is very useful for inspection and for the sets of differential changes that make the sub-scenario mechanism practical in the first place.
• CSV format: The model is converted into a table with virtually no effort. Since the schema defines a common set of attributes and relationships for a given entity type, each instance of that type can be represented as a row in a shared table, with one column per attribute (and an associated unit-of-measure column when applicable) and one column per relationship, containing the identifier of the related entity. The 598 resolved instances of EnergyConversion Technology1x1 in the baseline, for example, are simplified into a single table with the columns nominal_power_capacity, energy_conversion_efficiency, hasInputEnergyCarrier, hasOutputEnergy Carrier, and isOutputNodeOf, in a format that both spreadsheet tools and energy system optimization frameworks can use directly.
• Frictionless Data Package format: This same schema serves a third purpose. The information about types and constraints already declared for each attribute (string or number, minimum and maximum, required or optional) corresponds directly to a “Frictionless” table schema, so that each CSV table can be accompanied by a machine-readable description of its own columns, and the complete set of tables in a scenario, along with their metadata, can be grouped into a single datapackage.json. This package is self-descriptive and can be validated and loaded by any tool that understands the “Frictionless” specification, without the need for prior knowledge of CESDM itself.

Therefore, the three formats are not maintained separately: the schema that validates the YAML instance is the same one that defines the CSV table layout and generates the “Frictionless” resource description, so the three remain consistent with one another by their very nature.

5.6 Scenario metadata as a SPARQL-queryable data catalog

In addition, each scenario file opens with a scenario metadata block, containing: scenario_id, scenario_name, author, created_at, target_year, weather_year, geographic_region, spatial_resolution, temporal_resolution, version, description, main_assumptions, together with reference_scenario_id and baseline_scenario_id. This block is more than descriptive front matter. Its fields correspond directly to the classes and properties of the ScenarioManagement ontology: Scenario, BaselineScenario, and SubScenario as classes; target_year, base_year, domain_of_application, spatial_resolution, temporal_resolution, and assumption as data properties; hasBaselineScenario and derivedFrom/derivesIn as the object properties linking a scenario to its baseline and to its reference. Because of that correspondence, a scenario metadata block can be mapped mechanically into RDF: scenario_id becomes the subject IRI, each descriptive field becomes a literal-valued triple through the matching data property, and reference_scenario_id and baseline_scenario_id become object-valued triples pointing at the IRIs of other scenario nodes.

Applied across the whole of master_structure.yaml, this produces a small knowledge graph: one BaselineScenario node and seven SubScenario nodes, joined by hasBaselineScenario edges into a common root and by derivedFrom edges into the three storyline chains described above, with every edge additionally carrying the target_year/weather_year pair that situates the scenario in time. This is precisely what a data catalog is meant to hold: not the energy-system content itself, but the identifying, provenance, and scope information needed to find the right scenario before loading any of its content.

Finally, as the scenario manager creates Turtle (TTL) format and stores it as a knowledge graph through the use of Apache Jena Fuseki, it is loaded into a triple store, so that catalog is queryable with SPARQL rather than by searching through YAML files by hand. Questions that are awkward to answer over a folder of files become single queries over the metadata graph: which scenarios share a given target year across storylines, what the full derivation chain from the baseline to a given horizon looks like, which scenarios were produced under the same weather_year, or who authored a given scenario and when. Because the metadata graph is small, eight nodes and their connecting edges, against well over a thousand content entities per resolved scenario, it can be kept resident and queried cheaply, while the underlying content, in whichever of the three formats above is convenient, is retrieved only for the scenario or scenarios a query has already identified. The metadata layer functions, in effect, as a lightweight index in front of the content-heavy scenario files, in the same spirit as the separation between a dataset’s description and its distribution familiar from data-catalog vocabularies such as DCAT.

6 Discussion

Activities related to power grids, such as analysis and decision-making, are subject to a high degree of uncertainty. Specifically, power flow, grid dynamics, overall maintenance planning, etc., are based on a modeled representation of the system in question, and the system’s future behavior is projected based on additional assumptions related to best estimates. Therefore, it is necessary to have tools that formally define the assumptions covering the various elements that make up the system, the analytical activities, the analytical methods, and the scope and limits of the study, as well as the characteristics of the model and the available data. The multi-scenario approach involves compiling descriptions of the present and projections into the future. This approach implements a formal and standardized characterization of (i) a system, (ii) a scenario, and (iii) the data used in different use cases. The Ten-Year Network Development Plan (TYNDP) scenarios were taken as a proof of concept to test our framework. The joint developed by the European Network of Transmission System Operators for Electricity (ENTSO-E) and Gas (ENTSOG) provides a long-term analytical foundation for European energy infrastructure planning. Furthermore, it allows for the management of evolving assumptions regarding the system and the data, resulting in a documented, consistent, traceable, and formal approach. The approach defines the activities involved, the characteristics, and the scope of support for energy strategy, including the evolution of scenarios and the assumptions underpinning them. This approach has been applied to definition of energy system scenarios for proper documentation, reproducibility, and replicability. Thus, there are several other hurdles, including data tracking, assumption quantification, and scenario definition. As a result, comparing the results of analytical activities is a challenging task, even when applied to the same scenario. Therefore, this work proposes a solution based on a modular approach supported by semantics.

In general, the current work involves the use of semantic models in a comprehensive tool for creating, managing, and configuring energy system scenarios using the CESDM framework, supporting both baseline and sub-scenarios with full change tracking. This approach considers addressing the dynamic task of managing data in real-time. The semantic data infrastructure involves semantic mapping, process cataloging, and a robust query process. Finally, it is worth mentioning that this framework aims to manage current and future knowledge related to scenarios and assumptions.

Although the case study in this paper is limited to transmission-level TYNDP scenarios, the same scenario-versus-domain separation underlying the Scenario Management Ontology is, in principle, applicable to less centralized settings such as Renewable Energy Communities (RECs). A REC study still needs a baseline description of its members, distributed energy resources, consumption profiles, and market or regulatory constraints, and still produces variants of that baseline under different membership, tariff, or technology assumptions; nothing in the Scenario, BaselineScenario, or SubScenario classes of Table 5 assumes a transmission-level domain. What would need to be supplied is a domain ontology for the REC itself, playing the role CESDM plays for the physical network in this work, so that entities such as community members, shared assets, and consumption or generation profiles are formalized with the same rigor as EnergyNode or EnergyConversionTechnology1x1 are here (Section 4). A systematic comparison against emerging standardized data representations for local electricity markets and distributed energy resources ? Would be needed to establish exactly which REC-specific concepts such a domain ontology should contain; this benchmarking exercise is left as future work rather than claimed here, but the architecture itself does not present an obvious obstacle to it.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Footnotes

Edited by: Gabriela Patricia Henning, National Institute of Standards and Technology (NIST), United States

Reviewed by: Dimitri Thomopulos, University of Pisa, Italy

Beril Alpagut, Demir Enerji, Türkiye