A Long Short Term Memory Network-Based False Data Injection Cyberattack Detection and Mitigation in DC Microgrids

Integrating more renewable energy sources into dc microgrids necessitates the addition of control layers and communication channels, which can introduce vulnerabilities to cyberattacks. This paper presents an AI-based approach for detecting and countering False Data Injection Attacks (FDIAs) in these systems. The approach uses an Artificial Neural Network (ANN), specifically feedforward neural networks, to identify FDIAs and classify them as related to current or voltage. Additionally, a Long Short-Term Memory (LSTM) network is employed to mitigate the attacks by inferring and filtering out false voltage or current measurements. The method was validated on the Distributed Generator Unit (DGU) system using MATLAB/Simulink. To enhance the accuracy, the approach considers the impact on all grid lines and measurement points, supported by statistical analyses demonstrating the convergence and effectiveness of the response to cyberattacks. ANN successfully detected voltage attacks with an accuracy of 93% to 99% and identified current attacks with 99% accuracy. LSTM successfully overcame the false voltage injection attack with a 99.6% accuracy rate. LSTM effectively mitigates 90% of false current injection attacks.