AINETUS

AINETUS logo

AI-based decision support for complex grid operations

AI for Safety-Critical Network Infrastructures. Open source AI components that augment power system operator decision-making — combining reinforcement learning, explainability, and human-AI interaction to improve situational awareness and manage complex grid conditions.

Visit Project Website

AINETUS (AI for Safety-Critical Network Infrastructures) is an open source project hosted at LF Energy that brings human-centric artificial intelligence into power grid operations. It provides AI components designed to augment operator decision-making in real-time and for operational planning, improving situational awareness, anticipating system risks, and delivering explainable, actionable recommendations to control room staff.

Project Lifecycle Stage
Sandbox

Project Special Interest Group
Artificial Intelligence SIG

Originating from the European research project AI4REALNET, AINETUS combines domain-informed AI agents, reinforcement learning, and uncertainty-aware advisory functions with hypervision-based human-AI interfaces. It integrates with existing open source LF Energy platforms, including Grid2Op and OperatorFabric, and can be incorporated into grid management platforms, contributing an AI-based decision layer to the LF Energy ecosystem.

The Challenge

Modern power grids are increasingly difficult to operate. High renewable energy integration introduces faster, less predictable system dynamics. Traditional operational tools, built for a more stable, dispatchable generation mix, are often limited in their ability to anticipate system conditions and support operators under rapidly changing circumstances.

Control room operators face growing complexity with the same fundamental constraints: decisions must be made quickly, consequences are safety-critical, and informational load continues to increase. The gap between what AI and machine learning can offer and what operators can practically rely on in production environments has remained wide.

The barriers are concrete:

  • Existing tools provide limited proactive support, reacting to events rather than anticipating them
  • Recommendations from AI systems are often opaque, making it difficult for operators to evaluate or act on them with confidence
  • AI components developed in research settings are rarely designed to integrate with production operational platforms and workflows
  • Human factors and operator trust are underserved in most AI-for-grid research, limiting real-world adoption

These gaps create real risks: slower response to emerging congestion problems, reduced ability to manage renewable variability, and increased operator cognitive load at precisely the moments when clear, reliable decision support matters most.

Key Features

AINETUS provides operator-facing tools designed around the hypervision concept: intuitive visualisations and interactive interfaces that allow human operators to understand system conditions and AI recommendations at a glance. Rather than replacing operator judgment, these tools are built to extend it, surfacing relevant information at the right moment and presenting AI outputs in forms that operators can interrogate and trust.

AINETUS delivers reinforcement learning agents that identify optimal operational strategies for power system management. Agents are capable of adapting to evolving grid conditions while respecting physical constraints and established engineering practice. The current primary use case is real-time grid operation support, with a focus on topology optimisation and redispatch for congestion management.

Planned future use cases include:

  • Dynamic security assessment
  • Voltage control
  • Smart alarm management

Operators need to understand not just what an AI system recommends, but why and how confident that recommendation is. AINETUS provides explainability tools and uncertainty estimation as core components, not optional additions. This supports informed decision-making and builds the human operator trust that is a prerequisite for real-world deployment of AI in safety-critical environments.

AINETUS is designed to integrate with existing platforms and workflows. It leverages Grid2Op as the simulation environment for agent training and validation, integrates with OperatorFabric for operator notification and coordination workflows, and can be incorporated into SOGNO-based grid management platforms. This positions AINETUS as the AI decision layer within a broader open source operational stack, rather than a standalone tool requiring parallel infrastructure.

The human-centric AI approach developed in AINETUS is applicable to other safety-critical infrastructure domains, including railway operations and air traffic management. The architecture is designed to generalise across domains that share the core challenge: complex, fast-moving systems where operators require reliable, explainable AI support under time pressure.

Key Contributors

  • INESC TEC
  • Politecnico di Milano
  • Fraunhofer IEE
  • IRT SystemX
  • University of Amsterdam
  • Réseau de Transport d’Électricité (RTE)
  • TenneT
  • enliteAI
  • University of Kassel

Technical Foundation

AINETUS is built on established open source tools and research-validated AI methods to ensure reproducibility and integration with production-grade operational environments.

  • Grid2Op provides the simulation environment for training, validation, and benchmarking of reinforcement learning agents in realistic power system scenarios
  • Reinforcement learning, supervised learning, and human-AI interaction research underpin the core agent and interface components
  • Explainability techniques and uncertainty quantification are embedded throughout, ensuring recommendations are transparent and auditable

The project originates from the European research project AI4REALNET, funded by the European Union’s Horizon Europe, Grant Agreement no. 101119527.

Use Cases

  • Receive proactive, explainable recommendations for congestion management through topology optimisation and redispatch
  • Improve situational awareness under high-renewable operating conditions
  • Reduce cognitive load during rapidly changing system states
  • Evaluate AI recommendations with confidence through uncertainty-aware outputs and explainability interfaces

  • Deploy AI decision-support components within existing operational environments without replacing established tools and workflows
  • Integrate with OperatorFabric for structured coordination and notification across control centres
  • Validate agent behaviour in Grid2Op simulation environments before operational deployment

  • Use Grid2Op-compatible agent frameworks for research into power system AI
  • Contribute to explainability and uncertainty estimation methods for safety-critical AI applications
  • Integrate new use cases and AI-based decision agents
  • Extend the approach to adjacent safety-critical domains including railway and air traffic management

  • Integrate AINETUS components into grid management platforms as a modular AI decision layer
  • Build on open source human-AI interface components to develop operator-facing applications
  • Integrate new use cases and AI-based decision agents
  • Leverage the SOGNO integration pathway for distribution grid management applications

Project Roadmap

Current development focus is real-time grid operation support, covering topology optimisation and redispatch for congestion management. Planned future work spans dynamic security assessment, voltage control, and smart alarm management.

View the AI4REALNET project for the research context underpinning AINETUS’s development: https://ai4realnet.eu/.

AINETUS Roadmap
AINETUS Use Cases

Collaboration Opportunities

AINETUS welcomes participation from grid operators, transmission system operators, AI researchers, and platform developers working on:

  • AI and machine learning applications for power system operations
  • Human factors and operator interface design for safety-critical systems
  • Reinforcement learning or other AI methods (e.g., agentic AI, foundation models) for grid management and congestion resolution
  • Integration of AI decision-support into operational control room environments
  • Explainability and uncertainty quantification in production AI systems

Organizations seeking to validate AI decision-support tools in realistic grid environments, or to contribute operational requirements and use cases to the project, are encouraged to join the working group and contribute to real-world validation of the components.

AINETUS is building the shared, open AI infrastructure that grid operators need to manage increasingly complex systems, without each organization rebuilding the same foundational components independently.

To get involved:

Related Posts

LF Energy Welcomes New Members, Launches Three Projects, and Advances Open Source Grid Technology Across the Portfolio

LF Energy welcomes new members AZX, EcoPhi, and Empa; launches the AINETUS, URPX, and CUPID projects; advances Power Grid Model to Early Adoption; and highlights a TenneT case study demonstrating a 10x performance improvement using PowSyBl. Recent releases across EVerest, OperatorFabric, FlexMeasures, and other projects underscore continued momentum across the LF Energy ecosystem.