How can we make better use of energy resources? Over the past three years, INESC TEC set out to answer this question through the European GREEN.DAT.AI project, developing tools that combine Artificial Intelligence (AI), secure data sharing and distributed computing.
AI has become an increasingly important ally in the energy transition; the applications continue to expand as energy systems become more distributed and complex, from forecasting renewable energy generation and managing the smart charging of electric vehicles to enabling secure data sharing between organisations.
GREEN.DAT.AI focused on tackling these challenges. As part of the project, INESC TEC developed AI solutions for the energy and e-mobility sectors, leaving behind an experimental smart charging infrastructure for electric vehicles, alongside a range of other tools.
Gil Sampaio, INESC TEC researcher and project coordinator, said the project’s biggest challenge involved striking the right balance between algorithm performance and energy consumption, “especially since computational frugality – achieving strong results while using the minimum possible computational and energy resources – was one of the project’s core principles.”
“In a sector where energy efficiency is both part of the problem and part of the solution, producing accurate results is no longer enough. We also need to understand the cost of running an algorithm, the data it requires, the computing infrastructure it relies on, and how well it scales from the laboratory to an industrial environment. GREEN.DAT.AI brought together two dimensions that are often treated separately: the physical operation of power systems and the computational sustainability of AI tools,” he added.
INESC TEC’s work focused on energy systems and high-assurance software, delivering a range of results and demonstrators.
To improve renewable energy management, the team developed collaborative forecasting mechanisms for electricity market participants using real data from EDP’s wind power facilities. The aim was to explore new ways for organisations to work together without compromising the confidentiality of shared information by using federated learning approaches.
Concerning e- mobility, researchers developed forecasting, optimisation and smart charging services for electric vehicles. Using the microgrid at one of INESC TEC’s research facilities – the x-Energy Lab – together with the charging network at the Institute’s headquarters, the team created a chain of services capable of anticipating charging demand, forecasting renewable energy production, and continuously adapting charging profiles to match grid conditions and the availability of green energy, particularly from the building’s photovoltaic panels. As the number of electric vehicles expands, managing the power grid becomes increasingly challenging; charging several vehicles at the same time can create local voltage issues or overload the network, while photovoltaic generation does not always coincide with periods of highest demand. Sampaio explained that the solution involved “transferring recent AI innovations to the classic challenges of power system operation”, making it possible to improve the efficiency of existing resources. In this context, the team explored data-driven approaches to grid modelling, including digital twins built from historical data and updated in real time using measurements collected by smart meters.
“We worked with real-world data, representative low-voltage network operating conditions, and electric vehicle chargers designed and built by INESC TEC with advanced measurement, communication and control capabilities. This gave us a solid foundation for understanding how AI can support the integration of renewable energy and e-mobility without creating excessive dependence on fully centralised architectures,” the researcher added.
To assess the energy performance of the project’s tools, INESC TEC developed a tool for project partners to measure the energy consumption and computational resources required to run the AI algorithms and services created within GREEN.DAT.AI. “This tool supported our work on computational frugality by identifying solutions that required fewer computational and energy resources without compromising functional performance,” Sampaio said.
Another of INESC TEC’s contributions focused on the integration of Data Spaces, particularly their application in the energy sector, enabling secure, controlled and interoperable data sharing between organisations. This approach promotes more efficient use of information while protecting data sovereignty, privacy and the access rights defined by each organisation. As a result, it strengthens collaboration across the sector, supports the development of new services, and accelerates the transition towards smarter, more flexible and more sustainable energy systems.
“Sustainable AI is not simply a label; it is an engineering constraint. In industrial applications, where different algorithms may compete to solve the same problem, evaluating the energy cost of each solution becomes just as important as measuring their performance,” Sampaio added.
GREEN.DAT.AI reinforced INESC TEC’s research into the intersection of energy, software and data science, demonstrating that the energy transition increasingly depends on integrating physical infrastructure, intelligent algorithms and secure information-sharing mechanisms. Although the project has now ended, Sampaio pointed out that “the infrastructure we created, the smart chargers and the tools we developed will continue to support new research, demonstration and technology transfer projects in smart grids and e-mobility.”
News, current topics, curiosities and so much more about INESC TEC and its community!
