AI-Based Detection and Identification of Fault Direction of High Impedance Ground Faults in Resonant-Grounded Grids

Conrad T, Bluhm JN, Gaube S, Kordowich G, Jäger J (2025)

Publication Language: English

Publication Type: Conference contribution, Conference Contribution

Publication year: 2025

Event location: Esslingen DE

ISBN: 978-3-8007-6651-2

Abstract

In this paper, the generalisation capability of an AI-based approach using neural networks is investigated for the
detection of high impedance ground faults and their direction in resonant-grounded grids. The approach is to
train a neural network using simulated data of a generic station model. Data from several real substations with
multiple feeders was collected over a three-year period to generate a large dataset. This dataset includes both
manually labelled natural occurring faults and manually provoked ground faults. This resulted in a diverse dataset for validation and testing of the approach. Using this data, a neural network model with strong generalisation capabilities, with high sensitivity and specificity, was found. This allows for the recognition of ground faults up to 40 kOhm, while accurately identifying fault direction per feeder. This shows significant potential for widespread use in various power grids.

Authors with CRIS profile

Timon Conrad Lehrstuhl für Elektrische Energiesysteme Georg Kordowich Lehrstuhl für Elektrische Energiesysteme Johann Jäger Professur für Elektrische Energieversorgung

How to cite

APA:

Conrad, T., Bluhm, J.-N., Gaube, S., Kordowich, G., & Jäger, J. (2025). AI-Based Detection and Identification of Fault Direction of High Impedance Ground Faults in Resonant-Grounded Grids. In Proceedings of the STE 2025 - Sternpunktbehandlung in Netzen bis 110 kV (D-A-CH). Esslingen, DE.

MLA:

Conrad, Timon, et al. "AI-Based Detection and Identification of Fault Direction of High Impedance Ground Faults in Resonant-Grounded Grids." Proceedings of the STE 2025 - Sternpunktbehandlung in Netzen bis 110 kV (D-A-CH), Esslingen 2025.

BibTeX: Download