Decentralized Detection and Mitigation of False Data Injection Attacks in DC Microgrids Using Artificial Neural Network

Cooperative and distributed control strategies for direct current (DC) microgrids have made significant advancements in recent years. However, integrating a cyber layer to enhance resilience, scalability, and reliability also exposes the system to potential cyberattacks. This paper leverages data-driven methods to propose a novel approach for detecting cyberattacks in the secondary control layer. The proposed decentralized framework employs multilayer perceptron (MLP) neural networks to detect false data injection attacks (FDIA) on DC bus voltage while simultaneously mitigate their adverse effects on system operation. In this framework, the MLP neural networks are trained offline using local data under various conditions and are subsequently deployed online within the distributed generator units for fault detection and mitigation. Resilience is achieved through a pinning-node consensus-based secondary control strategy, which enables attack detection via MLP on other units regardless of whether all units are pinned. To demonstrate the effectiveness of the proposed method, simulation studies are conducted in MATLAB/Simulink on a DC microgrid across eight scenarios. The results and comparisons demonstrate that the proposed approach achieves a high detection accuracy of 99.96% under ideal conditions and 99.4% under sensor noise, enhancing security while reducing communication overhead.