LF Energy Summit Europe 2026

During the last years, the PowSyBl ecosystem has expanded in many directions around the core module dedicated to grid data management, allowing the collaboration of an increasing number of organisations. As the workhorse of power systems, Open Load Flow is leading the adoption cycle with live projects for load flow, security analysis or sensitivity analysis at many TSOs and Coordination Centers. The advanced optimization tool OpenRAO is also a leading operational remedial action optimizer in Europe. Beyond computation modules, many users also adopted PowSyBl thanks to its easiness of use in day-to-day scripting through the python interface, and to grid and substation visualization features. The technical ecosystem is also widening with projects like the dynamic simulation module Dynawo and the grid study platform GridSuite. These last projects are open-source and are an open invitation to start cooperation. The session will provide an overview of the ecosystem, providing, for each module, the main features and use cases, the maturity status, and the roadmap. Eventually, through some examples, the way PowSyBl governance allows to start new cooperations will be presented.

This session presents Artelys' vision of a sustainable business model around the PowSyBl open-source framework for power system modelling.

The approach relies on a clear separation between strategic governance, technical steering and community contributions, enabling transparency and alignment across stakeholders (TSOs, RCCs, utilities, vendors). The roadmap is not imposed by a vendor but co-defined by contributors and users, who can influence priorities through governance bodies and structured collaboration.

From a business perspective, value is created through services: integration, custom developments, and especially long-term support & maintenance of shared modules. This model leverages mutualization: several users co-finance core components, ensuring sustainability of complex software while avoiding vendor lock-in.

The session will illustrate how open-source approach enables both flexibility for individual projects and efficiency at ecosystem scale.

Capacity screening studies help US generation and large-load developers evaluate project size against available network capacity. Once a project size is selected, system impact studies determine project-specific impacts and upgrade costs. Developers often need to run tens or hundreds of sensitivities to understand risk, compare sites, and navigate the interconnection process, making scale, speed, and accuracy critical. In the US, these studies are most commonly performed with commercial planning software. While open-source alternatives exist, they are still rarely used in industry-grade workflows.

We will show how we are incorporating PowSyBl into our power system analysis platform as an alternative engine for US studies. We will cover why we selected PowSyBl after evaluating other open-source options, what it took to adapt it to US planning conventions, how its results compare with commercial tools, and lessons learned bringing open-source power-system analysis into production workflows. We will also discuss where we plan to extend its use next, including broader study coverage, deeper workflow integration, and contributions back to the ecosystem.

Transmission system operators face structural challenges: sovereignty, security, safety, efficiency, digitalization …
In this session, Sopra Steria shares a practitioner’s journey of embedding LF Energy’s PowSyBl into a real TSO production program for RTE. We cover two concrete contributions shipped to the open source community: decoupling the network-viewer component from any frontend framework (React/Angular agnostic), enabling simultaneous dual-value display (P active and Q reactive) on grid lines. Beyond the code, we discuss how an IT company structures its OSPO, open source governance (Licence Rules Book), and the cultural shift required to move from “consumer” to “contributor” in a critical infrastructure context.

The power grid is changing fast, and the systems behind it are under pressure. More data, tighter timeframes, increasing congestion, and regulatory requirements are pushing traditional grid security tools beyond their limits. At TenneT Netherlands, legacy tooling was simply no longer fast and adaptable enough to keep up with this new reality.

By combining the open source PowSyBl framework with a cloud-native orchestration platform (ReFlow), TenneT rebuilt its grid security analysis from the ground up. The result is a step change in performance and capability: reducing calculation runtimes from minutes to seconds and enabling a fundamentally different way of working.

This session will show how we:

- Moved from monolithic legacy tooling to a modular, scalable architecture
- Used open source to accelerate delivery and avoid vendor lock-in
- Established what it takes to run open source in a mission-critical environment
- Strengthened operator support by providing an integrated view across analyses through modular design

Predico is a collaborative forecasting platform developed by INESC TEC that turns energy forecasting into a daily market with two sides: buyers procure forecasts for the assets they operate, and multiple forecast suppliers compete to provide them, rewarded based on their accuracy. Their submissions are combined into a single, sharper forecast by Predico's open-source forecasting engine. This session introduces the platform and shares what we have learned from more than a year of running Predico in real operational pilots.

We bring LF Energy's PowSyBl into the control room as a decision support tool, combining grid analysis with Reinforcement Learning to solve post-contingency problems immediately. Increasing complexity amplifies overloads, demanding corrective actions beyond what manual operator heuristics and existing tools can deliver in time. Open-source RL agents backed by PowSyBl learn curative topology reconfigurations for the Simbench EHV network in Node-Breaker representation. Agents observe bus voltages, line loadings, injection profiles, and switching states. The action space controls breakers and disconnectors via valid action masks for node splitting, busbar re-assignment, and line (dis)connection. Remedial actions are injected as Operator Strategy actions inside PowSyBl's Security Analysis, faithfully reproducing N-1 assessment in training. We discuss reward design, bonuses, penalties, and action-masking. The talk ends with a live demo: a Leaflet map with colour-coded loadings, interactive single-line diagrams, forecast plots, and a control panel triggering Security Analysis and visualising corrective actions via PowSyBl's Operator Strategy in real time.

This talk will present the LF Energy project AINETUS, which implements AI-based decision-support (AI assistant) capabilities for power system operations, addressing the increasing complexity of modern grids driven by high levels of renewable energy integration. Traditional tools are often limited in their ability to anticipate system dynamics and support operators under rapidly changing conditions. AINETUS tackles this gap by combining data-driven techniques with domain knowledge to provide proactive, explainable, and operator-centric support.
It will show a set of components designed to work together or independently within existing operational environments, in particular: (1) Human-AI interaction tools that enhance situational awareness. (2) Reinforcement learning agents that identify optimal operational strategies, capable of adapting to evolving grid conditions while respecting physical constraints.
(3) Explainability tools and uncertainty estimation that help to provide transparent recommendations to support informed decision-making.
AINETUS is designed to integrate with existing platforms and workflows, leveraging environments such as Grid2Op for training and validation.

This lightning talk presents the integration of the PowSyBl framework into the datakit of GridFM, two flagship LF Energy projects, to bridge foundation models and industrial-grade power system simulation. GridFM relies on large-scale synthetic data generation, such as load flows, to train a foundation model, where training data quality directly impacts model performance during inference.
The integration of PowSyBl brings several key benefits. It extends GridFM to realistic and industrial-grade network models by supporting multiple formats (CIM, CGMES, PSS/E), enabling the use of richer and more representative datasets. It provides access to an industrial-grade load flow solver through Open Load Flow, improving the physical consistency and accuracy of generated data. Finally, it enables the generation of dynamic simulation data through Dynawo, opening the way to foundation models trained on both static and dynamic grid behaviours.
Overall, this integration is a key step toward the industrialization of grid foundation models, enabling their application to real-world power system studies.

Last year, the LF Energy project OpenSynth expanded with the launch of D-GITT, an open data hub featuring RTE7K: a dataset of the whole French transmission grid, including topological changes over three years. Striking the interest of other actors in opening their own dataset, RTE7K demonstrated its values as a realistic, open benchmark for grid analysis.

Now, RTE, CRESYM, eRoots and IRT SystemX joined forces to launch the RTE7K Challenges, a series of challenges centered around the dataset. The first one, kicking off at this summit, focuses on tackling the "missing piece" for running powerflows on the RTE7K dataset: power injection data. Participants are tasked with producing realistic, synthetic injections using whatever public data and methodology they can come up with. As well as being an enticing challenge for participants, the outcome will be a critical extension to RTE7K that will enable further users to test and validate tools using data comparable to what was previously exclusive to TSOs.

Future challenges will explore topics like AI-driven grid topology management, power flow analysis and more, pushing the boundaries of what’s possible with open grid data.

Last year, we introduced a basic Grid Foundation Model (GridFM) for power flow. This year we released GENCO – the GEometric Neural Corrective Optimizer, a neural solver for steady-state transmission-grid analysis. It unifies three grid-analysis tasks that are typically handled by separate pipelines: power flow (PF), optimal power flow (OPF), and state estimation (SE). GENCO represents the grid as a heterogeneous graph and iteratively refines solutions using task-specific physical decoders, box constraints, and explicit power-balance residuals. GENCO achieves state-of-the-art performance against specialized model baselines and classical solvers, with 5× and 200× speedup for AC-PF and AC-OPF respectively, as well as improved robustness in SE under sparse observability conditions. Beyond the model, we released gridfm-datakit for realistic grid data generation, gridfm-graphkit for low-code training and evaluation, and benchmark datasets spanning PF and OPF instances. By unifying model architecture, data generation, training, inference, and benchmarking, this end-to-end framework shortens innovation cycles and reduces integration efforts for the LFE community.

We present GridSFM, an open-source dataset and foundation model that enables building and operating transmission-scale power grid models from open data to machine learning inference.

Starting from publicly available sources in the United States (OpenStreetMap, U.S. EIA, and U.S. Census), our pipeline constructs geographically grounded, AC-OPF-solvable U.S. power grid models at continental scale. Importantly, our modeling pipeline can be extended to other locations where similar open data is available.

On top of this data layer, GridSFM introduces a physics-informed foundation model trained across a diverse mix of grid topologies and operating scenarios to predict AC-OPF solutions in milliseconds, enabling fast feasibility screening and solver acceleration.

This session will walk through the full stack: from open data ingestion and model construction to ML training and deployment. We will share lessons learned building reproducible, large-scale grid datasets and integrating physics constraints into learning systems.

EnerGNN is an open-source python package that aims at bridging the gap between research on Graph Neural Networks (GNNs) and industrial operation of energy networks.

This package includes:
- A faithful and adaptive data representation tailored for large scale industrial systems made of interacting entities.
- A modular library of GNN implementations designed to handle such data.
- A clear interface for business-driven use-case.
- A modular training loop for training GNNs on your specific use-case.

It is the result of 8+ years of research at RTE (Réseau de Transport d'Électricité) and academical partners (Université Paris-Saclay & Université de Liège). It is currently being used for multiple use-cases at RTE, and has shown its ability to scale to the full HV-EHV French system, and to output meaningful and high-precision predictions.

This presentation aims at showing the potential of our open-source package EnerGNN, and to attract potential users and contributors to help us on the following topics :
- GNN acceleration, robustification and improvement,
- Application of EnerGNN to new use cases,
- Extension of EnerGNN to other industrial domains.