DETECTION OF ANOMALIES IN POWER SYSTEMS: USING THE ISOLATION FOREST MODEL TO IDENTIFY CYBER THREATS

• Published in: Bezopasnostʹ informat͡s︡ionnykh tekhnologiĭ Vol. 32; no. 1; pp. 112 - 121
• Main Authors: Kochergin, Sergey V., Artemova, Svetlana V., Bakaev, Anatoly A., Maksimova, Elena А., Mityakov, Evgenii S., Vegera, Zhanna K.
• Format: Journal Article
• Language: English
• Published: Joint Stock Company "Experimental Scientific and Production Association SPELS 01.02.2025
• Subjects: power system anomalies, cybersecurity, risk theory, isolation forest, time series analysis, fractal analysis, power spectral density, machine learning, anomaly detection, cyberattacks
• ISSN: 2074-7128; 2074-7136
• DOI: 10.26583/bit.2025.1.07

Anomalies in the electrical signals of power systems play a critical role in identifying potential cyber threats, highlighting the need to integrate anomaly detection methods with risk theory in the context of cybersecurity. This study presents an approach to detecting anomalies in the voltage of an electrical grid using the Isolation Forest model. For the modeling, synthetic data were generated to simulate real operating conditions of the grid with a nominal voltage of 10 kV and an average load of 500 kVA. The anomalous data included artificially induced sharp voltage changes at specific time intervals, which could be associated with cyberattacks or other external interventions. The Isolation Forest model was trained on normal data and successfully applied to classify anomalies, effectively identifying critical moments related to potential threats. The study results demonstrate the high effectiveness of the proposed approach, enabling its use to enhance the resilience and security of power systems in the face of growing cyber threats.