Accurate and Energy-Efficient Detection of Cyberattacks Against Non-Linear AGC Systems

Automatic Generation Control (AGC) systems are widely used in interconnected power systems. However, AGC systems fully rely on communicated measurements, making them susceptible to cyber threats, particularly False Data Injection Attacks (FDIAs), which can destabilize the grid. This paper proposes a novel Spiking Neural Networks (SNNs)-based deep learning (DL) framework for detecting FDIAs against AGC systems considering their nonlinearities. The proposed framework combines FDIA-detection accuracy and energy-consumption efficiency by integrating Convolutional Neural Networks (CNNs) for feature extraction, Long Short-Term Memory (LSTM) networks for temporal modeling, attention mechanisms for adaptive focus, and SNNs for energy-efficient computation. Three different variations of the proposed framework—an ANN-based DL model, a fully-spiking model, and a hybrid spiking-ANN-based model—are implemented and analyzed under a high-fidelity set of diverse FDIA scenarios against an AGC system with nonlinearities. Results demonstrate that the proposed Hybrid-SNN DL model can accurately detect FDIAs on AGC systems while reducing DL model’s energy consumption by up to 56%. These findings highlight the potential of SNNs for enhancing cybersecurity in smart grids, offering a sustainable approach to reducing operational costs.