Counteracting Unobserved Cyber-Attacks in AGC and HVDC Schemes Using a Multi-Agent Framework

A novel approach to safeguard the AGC circlet and supremacy modulation controller of a high voltage direct current system from cyber-attacks is discussed in this research employing a multi-agent-based architecture. To achieve this, we propose adding a Master Agent (MA) and many Slave Agents (SAs) into the current AGC layer. For intrusion detection, the wide-area signals of the MA are processed employing a OCC. Used in the problem domain to minimize attacks by the SAs, are SVR models that utilize local knowledge. The suggested method moves to another training method for OCC which excludes the use of instances that are contaminated by noise in favor of protected data. The next step is to propagate a structure that the AGC system may employ for setting the online SVR models. When evaluating the tested solution’s performance, we consider a set of test scenarios that reflect the secured case, which includes normal and stressed system behavior; the attacked case, which pertains to various types of threats targeting data integrity and availability, as well as grid states. Interestingly, there is decent accuracy in identifying the assaults using the OCC, specifically the SVDDM. It is also observed that the SVR models operate well and optimally in terms of performance and at the same time effectively counter cyber-attacks.