Battery Energy Storage Systems (BESSs) are critical to smart grid functioning but are exposed to mounting cybersecurity threats with their integration into IoT and cloud-based control systems. Current solutions tend to be deficient in proper multi-class attack classification, secure encryption, and full integrity and power quality features. This paper proposes a comprehensive framework that integrates machine learning for attack detection, cryptographic security, data validation, and power quality control. With the BESS-Set dataset for binary classification, Random Forest achieves more than 98.50% accuracy, while LightGBM attains more than 97.60% accuracy for multi-class classification on the resampled data. Principal Component Analysis and feature importance show vital indicators such as State of Charge and battery power. Secure communication is implemented using Elliptic Curve Cryptography and a hybrid Blowfish–RSA encryption method. Data integrity is ensured through applying anomaly detection using Z-scores and redundancy testing, and IEEE 519-2022 power quality compliance is ensured by adaptive filtering and harmonic analysis. Real-time feasibility is demonstrated through hardware implementation on a PYNQ board, thus making this framework a stable and feasible option for BESS security in smart grids.
BESS-Enabled Smart Grid Environments: A Comprehensive Framework for Cyber Threat Classification, Cybersecurity, and Operational Resilience
Cristian Randieri
2025-01-01
Abstract
Battery Energy Storage Systems (BESSs) are critical to smart grid functioning but are exposed to mounting cybersecurity threats with their integration into IoT and cloud-based control systems. Current solutions tend to be deficient in proper multi-class attack classification, secure encryption, and full integrity and power quality features. This paper proposes a comprehensive framework that integrates machine learning for attack detection, cryptographic security, data validation, and power quality control. With the BESS-Set dataset for binary classification, Random Forest achieves more than 98.50% accuracy, while LightGBM attains more than 97.60% accuracy for multi-class classification on the resampled data. Principal Component Analysis and feature importance show vital indicators such as State of Charge and battery power. Secure communication is implemented using Elliptic Curve Cryptography and a hybrid Blowfish–RSA encryption method. Data integrity is ensured through applying anomaly detection using Z-scores and redundancy testing, and IEEE 519-2022 power quality compliance is ensured by adaptive filtering and harmonic analysis. Real-time feasibility is demonstrated through hardware implementation on a PYNQ board, thus making this framework a stable and feasible option for BESS security in smart grids.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.