AI Grid Protection – Coordinated Power System Protection using Machine Learning

Coordinated power system protection using machine learning (Netzschutz-KI)

This project was funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) – Project number 535389056.

On this page:

• Abstract
• Data and Code Availability
  • Simulation Dataset
  • Code Repositories
• Publications
• Project Details

Abstract

To address these limitations, the project explores the replacement of traditional protection devices by learning-based protection agents that operate directly on voltage and current measurements. These agents are trained to perform key protection tasks such as fault detection, fault localization, and selectively coordinated fault clearance. By leveraging the nonlinear modeling capabilities and generalization properties of neural networks, the proposed methods aim to achieve robust protection behavior across a wide range of fault scenarios, network topologies, and operating points.

A central focus of the project lies on coordination: instead of isolated relay decisions, multiple protection agents are trained and evaluated in a coordinated setting, enabling system-wide decision-making across different voltage levels. The developed approaches are validated using high-fidelity simulation data and demonstrated in a real-time laboratory environment, with the long-term goal of enabling transparent, reliable, and universally applicable AI-based protection concepts for future power systems.

Data and Code Availability

The AI Grid Protection project follows the principles of reproducible and transparent research. All core datasets, evaluation frameworks, and reference implementations developed within the project are publicly available.

Simulation Dataset

The electromagnetic transient (EMT)-simulated voltage and current dataset used throughout the project is publicly available on Zenodo:

• EMT Protection Dataset
  DOI: 10.5281/zenodo.18418330

The dataset provides high-resolution time-domain measurements and comprehensive scenario metadata, enabling reproducible evaluation of machine-learning-based protection methods under controlled and well-documented conditions.

Code Repositories

Reference implementations, evaluation pipelines, and experiment configurations are available via the following public GitHub repositories:

• Protection Evaluation Framework
  https://github.com/julianoelhaf/protection-eval-framework
  Modular evaluation framework for benchmarking machine-learning models across fault detection, localization, and coordination tasks.
• PSCC 2026 Deep Learning Protection Benchmark
  https://github.com/julianoelhaf/pscc2026-dl-protection-benchmark
  Reproducible benchmark accompanying the PSCC 2026 publication, including data preprocessing, model training, and evaluation scripts.

Together, these resources enable independent validation of the reported results and support further research on learning-based and coordinated protection concepts for future power systems.

Publications

Project Details

2024

• Coordinated grid protection based on machine learning methods

  
  (Third Party Funds Single)
  Project leader: Andreas Maier, Johann Jäger, Siming Bayer
  Term: July 1, 2024 - June 30, 2027
  Acronym: Netzschutz-KI
  Funding source: DFG-Einzelförderung / Sachbeihilfe (EIN-SBH)
  
  →More information