Securing DC Microgrids Against Cyber-Attacks: Hybrid Physics-Informed Neural Network Control Strategy with Real-Time Implementation

As the transition to green and sustainable energy accelerates, integrating renewable energy sources into DC microgrids (DC MGs) is becoming increasingly essential. However, ensuring efficient control performance while addressing cybersecurity threats remains a critical challenge. Communication links and sensors in DC MG are highly vulnerable to False Data Injection (FDI) attacks, yet cyber threat mitigation is often overlooked in control design. This study proposes a Hybrid Physics-Informed Neural Network (HPINN) methodology that integrates Linear Kalman Filters (LKF) with neural networks (NNs) to enhance cyber-attack detection and mitigation. Unlike conventional LKF, which lacks inherent attack mitigation capability, HPINN leverages NN-based correction to ensure robust state estimation and secure control operations. The proposed approach is validated on a supervisory-controlled 3-bus DC MG with a ring main connection, comprising a fuel cell and battery (node 1), a photovoltaic (PV) system and battery (node 2), and another PV system and battery (node 3). The performance is evaluated through MATLAB Simulink simulations and a real-time experimental setup, incorporating source changes, load variations, and FDI attack scenarios. Results demonstrate that HPINN effectively detects and mitigates cyber-attacks, ensuring accurate DC MG operation even under compromised sensor measurements. This work provides a scalable and secure control strategy, reinforcing the resilience of next-generation smart microgrids.