Open-source AI for power grid operations, with humans in control.
AINETUS is an LF Energy Sandbox project. It grew out of the European research project AI4REALNET and brings its results together as reusable, open components for grid operators and researchers.
The current focus is congestion management: topology changes and redispatch, recommended by AI agents, explained to the operator, and decided by the operator.
Agents that propose actions for the grid.
ExpertOp4Grid ranks and simulates topology actions from expert rules, with no training required. Reinforcement learning agents and their training setups live in the Grid2Op and T2.x repositories.
Explainability and uncertainty so recommendations can be judged. GridExplainer, the explainability dashboard and the uncertainty prediction module (conformal prediction for line-loading forecasts).
Operator interfaces that put context, recommendations and human judgement in one place. InteractiveAI and its API, plus Co-Study4Grid.
Scenarios and evaluation for reproducible experiments. grid2op-scenario, domain shift KPIs and agent-as-a-service with trace visualisation.
Most components build on Grid2Op. Each repository documents its own setup. Several are research prototypes; check the README before relying on one.
- Website: ainetus.com
- Mailing list: ainetus-discussion
- Meetings: LF Energy calendar
- Demo: AI assistant with a hypervision-based HMI
Contributors include INESC TEC, RTE, TenneT, Fraunhofer IEE, IRT SystemX, Politecnico di Milano, University of Amsterdam, University of Kassel and enliteAI.