Data-Driven Cyberphysical Anomaly Detection for Microgrids With GFM Inverters

Abstract

Microgrids (MGs) have gained significant attention considering their enhanced capability to integrate increasing distributed energy resources (DERs). The application of grid forming (GFM) inverters in a MG can control voltage/frequency, enable both islanded and grid-connected operation, and achieve 100% penetration. However, cyberphysical anomaly detection for a MG with GFM inverters has not been investigated before. In this article, the cyberphysical security of an ac MG with multiple GFM inverters is comprehensively assessed by considering short-circuit high-impedance faults (HIFs) as well as firstly exploiting False Data Injection Attacks (FDIAs) against centralized communication networks. Although the applied IEEE 1547-2018 based protection function could detect abnormal conditions, there exist cyberphysical anomalies could bypass it. In order to accomplish the detection and classification of such anomaly cases, a novel LSTM-based approach is proposed to identify the multi-class pattern regarding normal, cyberphysical threats during islanded and grid-connected by utilizing time series point of common coupling (PCC) frequency data as the paramount feature to effectively reflect the system operation status. The simulation is conducted in OPAL-RT real-time environment and the effectiveness of the proposed strategy is verified with an average detection accuracy of 94.72%.

Keywords

• Anomaly detection and classification
• cyberphysical security
• data-driven
• FDIAs
• GFM inverters
• HIFs