Data sharing to support the training and development of AI foundation models in the energy sector

Expected Outcome

Projects are expected to contribute to all the following expected outcomes:

• Effective and innovative methods for gathering, sharing and using large data sets in energy applications for the purpose of training AI models while ensuring privacy and security.
• Advanced and - wherever possible - open-source AI foundation models to support the digitalisation of the energy system, through improved grid observability, forecasting of supply and demand, advanced storage and renewables integration, demand side flexibility and energy efficiency.
• Enhanced cooperation, knowledge sharing and interoperability among energy system actors for secure and seamless data exchange.
• Advanced methodologies for AI model development ensuring that the results are FAIR (Findable, Accessible, Interoperable, Reusable) beyond the project ending, open source and accommodating for technology evolution.

Scope

The quality of AI models depends strongly on the amount, quality, and representativeness of the data used for their training. Projects are expected to focus on developing, training and testing AI foundation models, using extensive data sets to accelerate the energy transition in key focus areas of the energy sector. Projects are expected to:

• Building on the results of previously funded projects on the Common European energy data space[1] as well as the ongoing work within the Data 4 Energy expert group, projects are expected to demonstrate innovative methods and data governance strategies for sharing data among various energy actors. Projects are expected to demonstrate that they have access to large relevant datasets at the time of submitting the proposal.
• Projects are expected to develop innovative strategies for sharing data in an effective, secure and transparent way to support the development of AI foundation models. An indicative non exhaustive list of strategies that can be explored is the use of data space connectors or federated training of models to allow the integration of confidential data; the creation of synthetic data as an intermediate step to fill data gaps or improve the quality of the data could also be explored. Projects could also explore other innovative data sharing strategies, such as an ‘AI gym’. The data governance strategies that will be developed in the project should be designed in a way to be scalable and operational even after the end of the project.
• Projects are expected to develop foundation models tailored for the energy sector, addressing at least one -or more- of the following use cases: planning and operation of electricity grids (including static power flow modelling and dynamic EMT modelling), forecasting, congestion management, anomaly detection, fault diagnosis, predictive maintenance, flexibility management, demand-side energy efficiency, smart and bidirectional charging of EVs or other use cases that contribute to the objective of digitalisation of the energy system.
• Projects are expected to perform in-depth testing and validation of developed foundation models in lab demonstrators and real-life pilots, building on the results and using the facilities of previously funded projects on “AI testing and experimentation facilities (TEFs)”[2]. Regulatory sandboxes could also be considered for real-life pilot implementations.
• Projects are expected to bring together a wide group of stakeholders including data owners (for example energy utilities, grid operators, asset owners), application developers (for example startups, SMEs, hyperscalers) and model deployers (for example industry, grid operators, equipment manufacturers) providing a space of cooperation and collaborative data exchange.
• To scale up the training and development of the models, projects can benefit from the computing capacity of the AI factories that were recently announced by the European Commission[3], in particular the three AI factories that aim to develop AI applications for the energy sector.
• AI models are expected to respect ethical, safety and security principles; furthermore, transparency, explainability, and accountability should be embedded by design in the model development. Open-source development practices should be pursued wherever feasible. Efforts should be made to avoid biases, ensuring that the data produced is representative of diverse populations.

Selected projects are expected to contribute to the BRIDGE initiative[4] and actively participate in its activities.

Technology Readiness Level - Technology readiness level expected from completed projects

Activities are expected to achieve TRL 7-8 by the end of the project – see General Annex B. Activities may start at any TRL.

[1] Establish the grounds for a common European energy data space

[2] AI Testing and Experimentation Facility (TEF) for the energy sector – bringing technology to the market

[3] AI Factories

[4] https://bridge-smart-grid-storage-systems-digital-projects.ec.europa.eu/