Detection and Mitigation of False Data Injection Attacks in DC Microgrids: A UIO, Luenberger Observer, and CNN-Based Approach for 6DGUs

In this paper, a real-time, hybrid detection and mitigation framework for false data injection attacks in DC microgrids is presented. It is then evaluated on six distributed generation units where it combines a Luenberger Observer, an Unknown-Input Observer and a lightweight CNN to monitor voltage and control signals and swiftly detect coordinated, delayed or stealthy attacks. Layered scheme simulations in MATLAB show that the scheme maintains voltage regulation and current sharing in multiple FDIA scenarios and, as a result, brings back system stability without too much computational overhead. The defense combines classical observer theory and the latest machine learning by providing a practical and scalable solution for protecting critical power systems. Such a method will be extended to other microgrid architectures and be adapted to changing cyber-threat patterns in future work.