Source-linked AI summary
Opening the black box of energy modelling: Strategies and lessons learned
Stefan Pfenninger, Lion Hirth, Ingmar Schlecht, Eva Schmid, Frauke Wiese, Tom Brown, Chris Davis, Birgit Fais, Matthew Gidden, Heidi Heinrichs, Clara Heuberger, Simon Hilpert, Uwe Krien, Carsten Matke, Arjuna Nebel, Robbie Morrison, Berit Müller, Guido Pleßmann, Matthias Reeg, Jörn C. Richstein, Abhishek Shivakumar, Iain Staffell, Tim Tröndle, Clemens Wingenbach
TL;DR
Energy planning increasingly depends on models whose assumptions and data matter for contested policy decisions, yet practical and institutional barriers limit openness. Drawing on Openmod experience, the article provides a how-to guide for opening code and data through licensing, implementation, distribution, and community choices. It concludes that open modelling can improve research and policy practice, but broader institutional changes remain necessary.
Problem
Energy models and data are important to contested energy-policy decisions, but practical knowledge gaps, inertia, and restrictive or unclear data rights limit their transparency and reuse.
Method
The article synthesizes Openmod members’ practical experience into strategies covering intellectual property, licensing, modelling languages, distribution, support, and community building.
Results
Open energy-system modelling using open data and open-source models is associated by the authors with higher research quality, less duplication, greater credibility and legitimacy, and broader access to planning tools.
Takeaways & Limitations
Opening modelling black boxes requires coordinated choices by researchers and research groups, alongside institutional changes affecting data licensing and research incentives.
Takeaways & Limitations
Opening models can be delayed by documentation requirements and policymaking constraints, while aggregated datasets may remain only partly open when rights holders do not consent.
Abstract
from arXiv · showhide
The global energy system is undergoing a major transition, and in energy planning and decision-making across governments, industry and academia, models play a crucial role. Because of their policy relevance and contested nature, the transparency and open availability of energy models and data are of particular importance. Here we provide a practical how-to guide based on the collective experience of members of the Open Energy Modelling Initiative (Openmod). We discuss key steps to consider when opening code and data, including determining intellectual property ownership, choosing a licence and appropriate modelling languages, distributing code and data, and providing support and building communities. After illustrating these decisions with examples and lessons learned from the community, we conclude that even though individual researchers' choices are important, institutional changes are still also necessary for more openness and transparency in energy research.
1. Introduction
Energy-system changes make transparent and reproducible modelling increasingly important, while practical knowledge gaps and institutional inertia continue to impede openness. The article offers a how-to guide for opening model code and data.
- Changing energy-system requirements: Liberalised markets, emissions-reduction needs, and rapid wind, photovoltaic, and storage deployment demand models with greater spatial and temporal resolution.These changes complicate planning compared with a conventional energy system dominated by fewer players.
- Why openness matters: Open code and data can improve transparency, reproducibility, collaboration, and the quality of energy-policy analysis.The authors emphasize these benefits because energy policy is urgent and contested.
- Defining openness: Open artefacts must be both accessible and legally usable under commonly used licences that permit reuse without undue restrictions.This pragmatic definition applies to open code, open data, and open models.
- Article contribution: The article addresses limited openness by providing a practical guide based on experience with intellectual property, licensing, languages, distribution, and community building.It is not a review of energy-modelling methods or tools.
2. The energy modelling process
Energy modelling links diverse data inputs, processing choices, model code, and published results, but openness is constrained by data rights and undocumented assumptions. The modelling process therefore requires careful treatment of both technical transformations and legal conditions.
- Model types: Energy system models generate long-term deployment scenarios, whereas power system models focus on electricity and often represent grids in greater detail.Power system models may also be used operationally by utilities.
- Process scope: The article focuses on how open data and open-source code relate to the energy-modelling process, excluding publication and article-distribution conditions.Figure 1 situates these openness concepts within the modelling process.
- Data inputs: Input data comes from heterogeneous sources, and non-disclosure agreements or absent licences can limit whether models built on it can be opened.Unlicensed input data does not automatically grant legal permission to use or share it.
- Data processing: Three common processing tasks are time-series processing, geographic aggregation or disaggregation, and tabular-data transformation.Tabular processing can include assumptions for missing values and currency conversion.
- Transparency of transformations: Undocumented processing assumptions can affect model conclusions, making open processing code necessary for assessing results, interpretations, and policy implications.The authors specifically connect documentation with reproducibility and evaluation of processing impacts.
- Model implementation: Models combine code for reading data, constructing and solving equations with data parameters such as technology costs.Implementation translates conceptual equations and accompanying parameters into a computer program.
- Implementation choices: Implementations range from spreadsheets to commercial tools, mathematical-programming languages, and general-purpose languages such as Java, C++, or Python.Mathematical-programming languages keep model code at a higher level and focused on the mathematical formulation.
3. Key considerations when going open
Opening energy-model code and data requires coordinated legal, technical, and community decisions. The guidance emphasizes clarifying ownership and licensing, choosing suitable tools and release practices, and supporting reuse.
- Ownership: Researchers should establish who owns the relevant intellectual property before releasing code or data.Ownership is often held by the employer, so institutional technology-transfer or legal guidance may be needed.
- Ownership: A clear provenance record and contributor confirmations help ensure that released code is not encumbered by conflicting ownership claims.Provenance records the history of a codebase and its contributions.
- What and how much to publish: Complete reproducibility is difficult, but sharing partial code, data, or processing steps can still support reproduction and reuse.Workflow systems, provenance tools, and containerisation can help make research more reproducible.
- What and how much to publish: Documentation should match the target audience and explain interfaces, internals, or usage modes when collaboration and reuse are intended.Relevant considerations include graphical interfaces, library use, and application programming interfaces.
- Which licence to choose: Well-known licences clarify legal terms and interoperability, while permissive and copyleft choices differ in reuse restrictions and derivative-work requirements.Permissive licences generally allow integration into closed-source projects, whereas copyleft licences require derivative works to use the same or a compatible licence.
- Which licence to choose: Code-specific licences should be selected for software, while modelling languages and tools should reflect both model capabilities and the intended user group.Open-source languages can remove financial barriers, whereas tools such as Excel may be easier to access but harder to integrate into collaborative workflows.
4. Examples
The examples show that opening models and data involves legal, technical, organisational and community-building choices. They also show that open licensing can support collaboration, while retroactive opening and community coordination remain difficult.
- Starting open or closed and choosing a licence: Starting an open codebase simplifies licensing decisions, whereas retroactively opening closed models may require substituting or rewriting unreleasable code.Closed components may be restricted by copyright or have untraceable provenance.
- Starting open or closed and choosing a licence: Re-licensing existing code can require agreement from multiple copyright holders, motivating projects such as SciGRID and Calliope to choose permissive licences from the outset.The Apache 2.0 licence permits reuse in closed-source projects, subject to retaining original copyright notices.
- Starting open or closed and choosing a licence: UKTM illustrates how government-use requirements and documentation effort can delay the release of a previously closed model.Its release was postponed while the model had to meet government presentation, documentation and plausibility guidelines, and policymakers did not want it freely available during carbon-budget work.
- Aggregating and licensing data: Licensing aggregated data is complex when sources have different rights holders, unclear provenance, or commercial interests.For OPSD, only about half of rights holders had agreed to open licensing by 2017.
- Cross-pollination: PYPOWER and its descendants: Open licensing enabled PyPSA, pandapower and GridCal to share code and ideas despite initially fragmented development.Their projects pursued different directions, including multi-period optimisation, distribution-network modelling and new power-flow algorithms.
- Community building: OSeMOSYS and oemof: Community building requires sustained coordination, with OSeMOSYS emphasising outreach to researchers and policymakers while oemof first established inter-institutional development processes.OSeMOSYS used events, collaboration with United Nations agencies and newsletters; oemof later reimplemented renpass as renpassGIS.
5. Discussion and conclusion
The authors distil practical strategies for opening energy modelling, while arguing that researchers’ choices alone cannot deliver full openness. Institutional licensing, funding incentives and academic recognition also need to change.
- Discussion and conclusion: Open energy modelling can improve research quality, reduce duplicated work, increase credibility and legitimacy, and broaden access to planning tools.The authors describe benefits for researchers and government agencies, including those unable to afford commercial options.
- Discussion and conclusion: The recommended strategy covers intellectual-property consent, publishable components, licensing, programming choices, distribution channels and community building.These recommendations are based on the authors’ practical experience opening modelling black boxes.
- Discussion and conclusion: Researchers should complement open artefacts with conferences, accurate publications, concise methods and accessible, current documentation.Traditional academic dissemination remains useful for explaining models, data and methodology.
- Discussion and conclusion: Institutional openness requires data providers to offer open licences, funders to change incentives, and academic assessment to recognise software and data development.The authors argue that publicly funded research methods and results should ultimately be openly available.