Detecting Cyberattacks on Electrical Storage Systems through Neural Network Based Anomaly Detection Algorithm

Area of Science:

• Power Systems Engineering
• Cybersecurity in Energy Grids
• Artificial Intelligence for Grid Stability

Background:

• Distributed Energy Resources (DERs) are increasingly vital for modern power systems.
• Battery Electrical Storage Systems (BESS) are crucial for integrating unpredictable Renewable Energy Sources (RES).
• BESS are vulnerable to cyberattacks due to remote SCADA system control.

Purpose of the Study:

• To analyze the cybersecurity vulnerabilities of BESS.
• To propose an anomaly detection algorithm for BESS.
• To enhance the detection of dangerous working conditions in BESS.

Main Methods:

• Developed an anomaly detection algorithm utilizing an autoencoder neural network architecture.
• Focused on observing the physical behavior of BESS for anomaly identification.
• Compared the proposed autoencoder approach with the One Class Support Vector Machine (OCSVM) algorithm.

Main Results:

• The autoencoder-based algorithm effectively detects anomalies in BESS operation.
• Demonstrated superior performance compared to the traditional One Class Support Vector Machine (OCSVM).
• The method successfully identifies dangerous working conditions indicative of cyber threats.

Conclusions:

• The proposed autoencoder anomaly detection offers a robust solution for BESS cybersecurity.
• Monitoring physical behavior is a viable strategy to detect cyberattacks on BESS.
• This approach enhances the resilience and safety of power systems with integrated BESS.