Accurate Joint Detection of False Data Injection Attacks on Islanded PV Output Power and State of Health Estimation of Lithium-Ion Batteries

With the widespread adoption of islanded DC micro-grids incorporating photovoltaic (PV) systems and battery energy storage systems (BESSs), safeguarding the cyber-physical security of PV-BESS configurations becomes crucial. While numerous false data injection attack (FDIA) detection techniques have been presented in existing literature, these methods are often tailor-made for specific attack types. This paper introduces an innovative approach for concurrently detecting FDIAs on islanded $P V$ power plants and BESSs. A bespoke two-layer random forest model, meticulously crafted and trained, is deployed for identifying FDIAs instigated by malicious adversaries. Our method uniquely caters to three distinct cyber-attack scenarios: tampered attacks, poisoned attacks, and replay attacks. The efficacy of our FDIA detection approach is rigorously evaluated utilizing real-world datasets and synthetically generated attacks. The results underscore the method’s impressive performance in identifying a spectrum of cyber-attacks on both PV system output and the state of health output from the battery data.